Add files via upload

demucs/__init__.py

@@ -0,0 +1,5 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.

demucs/__main__.py

@@ -0,0 +1,272 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+import json
+import os
+import sys
+import time
+from dataclasses import dataclass, field
+from fractions import Fraction
+
+import torch as th
+from torch import distributed, nn
+from torch.nn.parallel.distributed import DistributedDataParallel
+
+from .augment import FlipChannels, FlipSign, Remix, Shift
+from .compressed import StemsSet, build_musdb_metadata, get_musdb_tracks
+from .model import Demucs
+from .parser import get_name, get_parser
+from .raw import Rawset
+from .tasnet import ConvTasNet
+from .test import evaluate
+from .train import train_model, validate_model
+from .utils import human_seconds, load_model, save_model, sizeof_fmt
+
+
+@dataclass
+class SavedState:
+    metrics: list = field(default_factory=list)
+    last_state: dict = None
+    best_state: dict = None
+    optimizer: dict = None
+
+
+def main():
+    parser = get_parser()
+    args = parser.parse_args()
+    name = get_name(parser, args)
+    print(f"Experiment {name}")
+
+    if args.musdb is None and args.rank == 0:
+        print(
+            "You must provide the path to the MusDB dataset with the --musdb flag. "
+            "To download the MusDB dataset, see https://sigsep.github.io/datasets/musdb.html.",
+            file=sys.stderr)
+        sys.exit(1)
+
+    eval_folder = args.evals / name
+    eval_folder.mkdir(exist_ok=True, parents=True)
+    args.logs.mkdir(exist_ok=True)
+    metrics_path = args.logs / f"{name}.json"
+    eval_folder.mkdir(exist_ok=True, parents=True)
+    args.checkpoints.mkdir(exist_ok=True, parents=True)
+    args.models.mkdir(exist_ok=True, parents=True)
+
+    if args.device is None:
+        device = "cpu"
+        if th.cuda.is_available():
+            device = "cuda"
+    else:
+        device = args.device
+
+    th.manual_seed(args.seed)
+    # Prevents too many threads to be started when running `museval` as it can be quite
+    # inefficient on NUMA architectures.
+    os.environ["OMP_NUM_THREADS"] = "1"
+
+    if args.world_size > 1:
+        if device != "cuda" and args.rank == 0:
+            print("Error: distributed training is only available with cuda device", file=sys.stderr)
+            sys.exit(1)
+        th.cuda.set_device(args.rank % th.cuda.device_count())
+        distributed.init_process_group(backend="nccl",
+                                       init_method="tcp://" + args.master,
+                                       rank=args.rank,
+                                       world_size=args.world_size)
+
+    checkpoint = args.checkpoints / f"{name}.th"
+    checkpoint_tmp = args.checkpoints / f"{name}.th.tmp"
+    if args.restart and checkpoint.exists():
+        checkpoint.unlink()
+
+    if args.test:
+        args.epochs = 1
+        args.repeat = 0
+        model = load_model(args.models / args.test)
+    elif args.tasnet:
+        model = ConvTasNet(audio_channels=args.audio_channels, samplerate=args.samplerate, X=args.X)
+    else:
+        model = Demucs(
+            audio_channels=args.audio_channels,
+            channels=args.channels,
+            context=args.context,
+            depth=args.depth,
+            glu=args.glu,
+            growth=args.growth,
+            kernel_size=args.kernel_size,
+            lstm_layers=args.lstm_layers,
+            rescale=args.rescale,
+            rewrite=args.rewrite,
+            sources=4,
+            stride=args.conv_stride,
+            upsample=args.upsample,
+            samplerate=args.samplerate
+        )
+    model.to(device)
+    if args.show:
+        print(model)
+        size = sizeof_fmt(4 * sum(p.numel() for p in model.parameters()))
+        print(f"Model size {size}")
+        return
+
+    optimizer = th.optim.Adam(model.parameters(), lr=args.lr)
+
+    try:
+        saved = th.load(checkpoint, map_location='cpu')
+    except IOError:
+        saved = SavedState()
+    else:
+        model.load_state_dict(saved.last_state)
+        optimizer.load_state_dict(saved.optimizer)
+
+    if args.save_model:
+        if args.rank == 0:
+            model.to("cpu")
+            model.load_state_dict(saved.best_state)
+            save_model(model, args.models / f"{name}.th")
+        return
+
+    if args.rank == 0:
+        done = args.logs / f"{name}.done"
+        if done.exists():
+            done.unlink()
+
+    if args.augment:
+        augment = nn.Sequential(FlipSign(), FlipChannels(), Shift(args.data_stride),
+                                Remix(group_size=args.remix_group_size)).to(device)
+    else:
+        augment = Shift(args.data_stride)
+
+    if args.mse:
+        criterion = nn.MSELoss()
+    else:
+        criterion = nn.L1Loss()
+
+    # Setting number of samples so that all convolution windows are full.
+    # Prevents hard to debug mistake with the prediction being shifted compared
+    # to the input mixture.
+    samples = model.valid_length(args.samples)
+    print(f"Number of training samples adjusted to {samples}")
+
+    if args.raw:
+        train_set = Rawset(args.raw / "train",
+                           samples=samples + args.data_stride,
+                           channels=args.audio_channels,
+                           streams=[0, 1, 2, 3, 4],
+                           stride=args.data_stride)
+
+        valid_set = Rawset(args.raw / "valid", channels=args.audio_channels)
+    else:
+        if not args.metadata.is_file() and args.rank == 0:
+            build_musdb_metadata(args.metadata, args.musdb, args.workers)
+        if args.world_size > 1:
+            distributed.barrier()
+        metadata = json.load(open(args.metadata))
+        duration = Fraction(samples + args.data_stride, args.samplerate)
+        stride = Fraction(args.data_stride, args.samplerate)
+        train_set = StemsSet(get_musdb_tracks(args.musdb, subsets=["train"], split="train"),
+                             metadata,
+                             duration=duration,
+                             stride=stride,
+                             samplerate=args.samplerate,
+                             channels=args.audio_channels)
+        valid_set = StemsSet(get_musdb_tracks(args.musdb, subsets=["train"], split="valid"),
+                             metadata,
+                             samplerate=args.samplerate,
+                             channels=args.audio_channels)
+
+    best_loss = float("inf")
+    for epoch, metrics in enumerate(saved.metrics):
+        print(f"Epoch {epoch:03d}: "
+              f"train={metrics['train']:.8f} "
+              f"valid={metrics['valid']:.8f} "
+              f"best={metrics['best']:.4f} "
+              f"duration={human_seconds(metrics['duration'])}")
+        best_loss = metrics['best']
+
+    if args.world_size > 1:
+        dmodel = DistributedDataParallel(model,
+                                         device_ids=[th.cuda.current_device()],
+                                         output_device=th.cuda.current_device())
+    else:
+        dmodel = model
+
+    for epoch in range(len(saved.metrics), args.epochs):
+        begin = time.time()
+        model.train()
+        train_loss = train_model(epoch,
+                                 train_set,
+                                 dmodel,
+                                 criterion,
+                                 optimizer,
+                                 augment,
+                                 batch_size=args.batch_size,
+                                 device=device,
+                                 repeat=args.repeat,
+                                 seed=args.seed,
+                                 workers=args.workers,
+                                 world_size=args.world_size)
+        model.eval()
+        valid_loss = validate_model(epoch,
+                                    valid_set,
+                                    model,
+                                    criterion,
+                                    device=device,
+                                    rank=args.rank,
+                                    split=args.split_valid,
+                                    world_size=args.world_size)
+
+        duration = time.time() - begin
+        if valid_loss < best_loss:
+            best_loss = valid_loss
+            saved.best_state = {
+                key: value.to("cpu").clone()
+                for key, value in model.state_dict().items()
+            }
+        saved.metrics.append({
+            "train": train_loss,
+            "valid": valid_loss,
+            "best": best_loss,
+            "duration": duration
+        })
+        if args.rank == 0:
+            json.dump(saved.metrics, open(metrics_path, "w"))
+
+        saved.last_state = model.state_dict()
+        saved.optimizer = optimizer.state_dict()
+        if args.rank == 0 and not args.test:
+            th.save(saved, checkpoint_tmp)
+            checkpoint_tmp.rename(checkpoint)
+
+        print(f"Epoch {epoch:03d}: "
+              f"train={train_loss:.8f} valid={valid_loss:.8f} best={best_loss:.4f} "
+              f"duration={human_seconds(duration)}")
+
+    del dmodel
+    model.load_state_dict(saved.best_state)
+    if args.eval_cpu:
+        device = "cpu"
+        model.to(device)
+    model.eval()
+    evaluate(model,
+             args.musdb,
+             eval_folder,
+             rank=args.rank,
+             world_size=args.world_size,
+             device=device,
+             save=args.save,
+             split=args.split_valid,
+             shifts=args.shifts,
+             workers=args.eval_workers)
+    model.to("cpu")
+    save_model(model, args.models / f"{name}.th")
+    if args.rank == 0:
+        print("done")
+        done.write_text("done")
+
+
+if __name__ == "__main__":
+    main()

demucs/apply.py

@@ -0,0 +1,294 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+"""
+Code to apply a model to a mix. It will handle chunking with overlaps and
+inteprolation between chunks, as well as the "shift trick".
+"""
+from concurrent.futures import ThreadPoolExecutor
+import random
+import typing as tp
+from multiprocessing import Process,Queue,Pipe
+
+import torch as th
+from torch import nn
+from torch.nn import functional as F
+import tqdm
+import tkinter as tk
+
+from .demucs import Demucs
+from .hdemucs import HDemucs
+from .utils import center_trim, DummyPoolExecutor
+
+Model = tp.Union[Demucs, HDemucs]
+
+progress_bar_num = 0
+
+class BagOfModels(nn.Module):
+    def __init__(self, models: tp.List[Model],
+                 weights: tp.Optional[tp.List[tp.List[float]]] = None,
+                 segment: tp.Optional[float] = None):
+        """
+        Represents a bag of models with specific weights.
+        You should call `apply_model` rather than calling directly the forward here for
+        optimal performance.
+
+        Args:
+            models (list[nn.Module]): list of Demucs/HDemucs models.
+            weights (list[list[float]]): list of weights. If None, assumed to
+                be all ones, otherwise it should be a list of N list (N number of models),
+                each containing S floats (S number of sources).
+            segment (None or float): overrides the `segment` attribute of each model
+                (this is performed inplace, be careful if you reuse the models passed).
+        """
+        
+        super().__init__()
+        assert len(models) > 0
+        first = models[0]
+        for other in models:
+            assert other.sources == first.sources
+            assert other.samplerate == first.samplerate
+            assert other.audio_channels == first.audio_channels
+            if segment is not None:
+                other.segment = segment
+
+        self.audio_channels = first.audio_channels
+        self.samplerate = first.samplerate
+        self.sources = first.sources
+        self.models = nn.ModuleList(models)
+
+        if weights is None:
+            weights = [[1. for _ in first.sources] for _ in models]
+        else:
+            assert len(weights) == len(models)
+            for weight in weights:
+                assert len(weight) == len(first.sources)
+        self.weights = weights
+
+    def forward(self, x):
+        raise NotImplementedError("Call `apply_model` on this.")
+
+class TensorChunk:
+    def __init__(self, tensor, offset=0, length=None):
+        total_length = tensor.shape[-1]
+        assert offset >= 0
+        assert offset < total_length
+
+        if length is None:
+            length = total_length - offset
+        else:
+            length = min(total_length - offset, length)
+
+        if isinstance(tensor, TensorChunk):
+            self.tensor = tensor.tensor
+            self.offset = offset + tensor.offset
+        else:
+            self.tensor = tensor
+            self.offset = offset
+        self.length = length
+        self.device = tensor.device
+
+    @property
+    def shape(self):
+        shape = list(self.tensor.shape)
+        shape[-1] = self.length
+        return shape
+
+    def padded(self, target_length):
+        delta = target_length - self.length
+        total_length = self.tensor.shape[-1]
+        assert delta >= 0
+
+        start = self.offset - delta // 2
+        end = start + target_length
+
+        correct_start = max(0, start)
+        correct_end = min(total_length, end)
+
+        pad_left = correct_start - start
+        pad_right = end - correct_end
+
+        out = F.pad(self.tensor[..., correct_start:correct_end], (pad_left, pad_right))
+        assert out.shape[-1] == target_length
+        return out
+
+def tensor_chunk(tensor_or_chunk):
+    if isinstance(tensor_or_chunk, TensorChunk):
+        return tensor_or_chunk
+    else:
+        assert isinstance(tensor_or_chunk, th.Tensor)
+        return TensorChunk(tensor_or_chunk)
+
+def apply_model(model, mix, shifts=1, split=True, overlap=0.25, transition_power=1., static_shifts=1, set_progress_bar=None, device=None, progress=False, num_workers=0, pool=None): 
+    """
+    Apply model to a given mixture.
+
+    Args:
+        shifts (int): if > 0, will shift in time `mix` by a random amount between 0 and 0.5 sec
+            and apply the oppositve shift to the output. This is repeated `shifts` time and
+            all predictions are averaged. This effectively makes the model time equivariant
+            and improves SDR by up to 0.2 points.
+        split (bool): if True, the input will be broken down in 8 seconds extracts
+            and predictions will be performed individually on each and concatenated.
+            Useful for model with large memory footprint like Tasnet.
+        progress (bool): if True, show a progress bar (requires split=True)
+        device (torch.device, str, or None): if provided, device on which to
+            execute the computation, otherwise `mix.device` is assumed.
+            When `device` is different from `mix.device`, only local computations will
+            be on `device`, while the entire tracks will be stored on `mix.device`.
+    """
+    
+    global fut_length
+    global bag_num
+    global prog_bar
+    
+    if device is None:
+        device = mix.device
+    else:
+        device = th.device(device)
+    if pool is None:
+        if num_workers > 0 and device.type == 'cpu':
+            pool = ThreadPoolExecutor(num_workers)
+        else:
+            pool = DummyPoolExecutor()
+            
+    kwargs = {
+        'shifts': shifts,
+        'split': split,
+        'overlap': overlap,
+        'transition_power': transition_power,
+        'progress': progress,
+        'device': device,
+        'pool': pool,
+        'set_progress_bar': set_progress_bar,
+        'static_shifts': static_shifts,
+    }
+    
+    if isinstance(model, BagOfModels):
+        # Special treatment for bag of model.
+        # We explicitely apply multiple times `apply_model` so that the random shifts
+        # are different for each model.
+
+        estimates = 0
+        totals = [0] * len(model.sources)
+        bag_num = len(model.models)
+        fut_length = 0
+        prog_bar = 0
+        current_model = 0 #(bag_num + 1)
+        for sub_model, weight in zip(model.models, model.weights):
+            original_model_device = next(iter(sub_model.parameters())).device
+            sub_model.to(device)
+            fut_length += fut_length
+            current_model += 1
+            out = apply_model(sub_model, mix, **kwargs)
+            sub_model.to(original_model_device)
+            for k, inst_weight in enumerate(weight):
+                out[:, k, :, :] *= inst_weight
+                totals[k] += inst_weight
+            estimates += out
+            del out
+
+        for k in range(estimates.shape[1]):
+            estimates[:, k, :, :] /= totals[k]
+        return estimates
+
+    model.to(device)
+    model.eval()
+    assert transition_power >= 1, "transition_power < 1 leads to weird behavior."
+    batch, channels, length = mix.shape
+    
+    if shifts:
+        kwargs['shifts'] = 0
+        max_shift = int(0.5 * model.samplerate)
+        mix = tensor_chunk(mix)
+        padded_mix = mix.padded(length + 2 * max_shift)
+        out = 0
+        for _ in range(shifts):
+            offset = random.randint(0, max_shift)
+            shifted = TensorChunk(padded_mix, offset, length + max_shift - offset)
+            shifted_out = apply_model(model, shifted, **kwargs)
+            out += shifted_out[..., max_shift - offset:]
+        out /= shifts
+        return out
+    elif split:
+        kwargs['split'] = False
+        out = th.zeros(batch, len(model.sources), channels, length, device=mix.device)
+        sum_weight = th.zeros(length, device=mix.device)
+        segment = int(model.samplerate * model.segment)
+        stride = int((1 - overlap) * segment)
+        offsets = range(0, length, stride)
+        scale = float(format(stride / model.samplerate, ".2f"))
+        # We start from a triangle shaped weight, with maximal weight in the middle
+        # of the segment. Then we normalize and take to the power `transition_power`.
+        # Large values of transition power will lead to sharper transitions.
+        weight = th.cat([th.arange(1, segment // 2 + 1, device=device),
+                         th.arange(segment - segment // 2, 0, -1, device=device)])
+        assert len(weight) == segment
+        # If the overlap < 50%, this will translate to linear transition when
+        # transition_power is 1.
+        weight = (weight / weight.max())**transition_power
+        futures = []
+        for offset in offsets:
+            chunk = TensorChunk(mix, offset, segment)
+            future = pool.submit(apply_model, model, chunk, **kwargs)
+            futures.append((future, offset))
+            offset += segment
+        if progress:
+            futures = tqdm.tqdm(futures, unit_scale=scale, ncols=120, unit='seconds')
+        for future, offset in futures:
+            if set_progress_bar:
+                fut_length = (len(futures) * bag_num * static_shifts)
+                prog_bar += 1
+                set_progress_bar(0.1, (0.8/fut_length*prog_bar))
+            chunk_out = future.result()
+            chunk_length = chunk_out.shape[-1]
+            out[..., offset:offset + segment] += (weight[:chunk_length] * chunk_out).to(mix.device)
+            sum_weight[offset:offset + segment] += weight[:chunk_length].to(mix.device)
+        assert sum_weight.min() > 0
+        out /= sum_weight
+        return out
+    else:
+        if hasattr(model, 'valid_length'):
+            valid_length = model.valid_length(length)
+        else:
+            valid_length = length
+        mix = tensor_chunk(mix)
+        padded_mix = mix.padded(valid_length).to(device)
+        with th.no_grad():
+            out = model(padded_mix)
+        return center_trim(out, length)
+    
+def demucs_segments(demucs_segment, demucs_model):
+    
+    if demucs_segment == 'Default':
+        segment = None
+        if isinstance(demucs_model, BagOfModels):
+            if segment is not None:
+                for sub in demucs_model.models:
+                    sub.segment = segment
+        else:
+            if segment is not None:
+                sub.segment = segment
+    else:
+        try:
+            segment = int(demucs_segment)
+            if isinstance(demucs_model, BagOfModels):
+                if segment is not None:
+                    for sub in demucs_model.models:
+                        sub.segment = segment
+            else:
+                if segment is not None:
+                    sub.segment = segment
+        except:
+            segment = None
+            if isinstance(demucs_model, BagOfModels):
+                if segment is not None:
+                    for sub in demucs_model.models:
+                        sub.segment = segment
+            else:
+                if segment is not None:
+                    sub.segment = segment
+                    
+    return demucs_model

demucs/demucs.py

@@ -0,0 +1,459 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+import math
+import typing as tp
+
+import julius
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+from .states import capture_init
+from .utils import center_trim, unfold
+
+
+class BLSTM(nn.Module):
+    """
+    BiLSTM with same hidden units as input dim.
+    If `max_steps` is not None, input will be splitting in overlapping
+    chunks and the LSTM applied separately on each chunk.
+    """
+    def __init__(self, dim, layers=1, max_steps=None, skip=False):
+        super().__init__()
+        assert max_steps is None or max_steps % 4 == 0
+        self.max_steps = max_steps
+        self.lstm = nn.LSTM(bidirectional=True, num_layers=layers, hidden_size=dim, input_size=dim)
+        self.linear = nn.Linear(2 * dim, dim)
+        self.skip = skip
+
+    def forward(self, x):
+        B, C, T = x.shape
+        y = x
+        framed = False
+        if self.max_steps is not None and T > self.max_steps:
+            width = self.max_steps
+            stride = width // 2
+            frames = unfold(x, width, stride)
+            nframes = frames.shape[2]
+            framed = True
+            x = frames.permute(0, 2, 1, 3).reshape(-1, C, width)
+
+        x = x.permute(2, 0, 1)
+
+        x = self.lstm(x)[0]
+        x = self.linear(x)
+        x = x.permute(1, 2, 0)
+        if framed:
+            out = []
+            frames = x.reshape(B, -1, C, width)
+            limit = stride // 2
+            for k in range(nframes):
+                if k == 0:
+                    out.append(frames[:, k, :, :-limit])
+                elif k == nframes - 1:
+                    out.append(frames[:, k, :, limit:])
+                else:
+                    out.append(frames[:, k, :, limit:-limit])
+            out = torch.cat(out, -1)
+            out = out[..., :T]
+            x = out
+        if self.skip:
+            x = x + y
+        return x
+
+
+def rescale_conv(conv, reference):
+    """Rescale initial weight scale. It is unclear why it helps but it certainly does.
+    """
+    std = conv.weight.std().detach()
+    scale = (std / reference)**0.5
+    conv.weight.data /= scale
+    if conv.bias is not None:
+        conv.bias.data /= scale
+
+
+def rescale_module(module, reference):
+    for sub in module.modules():
+        if isinstance(sub, (nn.Conv1d, nn.ConvTranspose1d, nn.Conv2d, nn.ConvTranspose2d)):
+            rescale_conv(sub, reference)
+
+
+class LayerScale(nn.Module):
+    """Layer scale from [Touvron et al 2021] (https://arxiv.org/pdf/2103.17239.pdf).
+    This rescales diagonaly residual outputs close to 0 initially, then learnt.
+    """
+    def __init__(self, channels: int, init: float = 0):
+        super().__init__()
+        self.scale = nn.Parameter(torch.zeros(channels, requires_grad=True))
+        self.scale.data[:] = init
+
+    def forward(self, x):
+        return self.scale[:, None] * x
+
+
+class DConv(nn.Module):
+    """
+    New residual branches in each encoder layer.
+    This alternates dilated convolutions, potentially with LSTMs and attention.
+    Also before entering each residual branch, dimension is projected on a smaller subspace,
+    e.g. of dim `channels // compress`.
+    """
+    def __init__(self, channels: int, compress: float = 4, depth: int = 2, init: float = 1e-4,
+                 norm=True, attn=False, heads=4, ndecay=4, lstm=False, gelu=True,
+                 kernel=3, dilate=True):
+        """
+        Args:
+            channels: input/output channels for residual branch.
+            compress: amount of channel compression inside the branch.
+            depth: number of layers in the residual branch. Each layer has its own
+                projection, and potentially LSTM and attention.
+            init: initial scale for LayerNorm.
+            norm: use GroupNorm.
+            attn: use LocalAttention.
+            heads: number of heads for the LocalAttention.
+            ndecay: number of decay controls in the LocalAttention.
+            lstm: use LSTM.
+            gelu: Use GELU activation.
+            kernel: kernel size for the (dilated) convolutions.
+            dilate: if true, use dilation, increasing with the depth.
+        """
+
+        super().__init__()
+        assert kernel % 2 == 1
+        self.channels = channels
+        self.compress = compress
+        self.depth = abs(depth)
+        dilate = depth > 0
+
+        norm_fn: tp.Callable[[int], nn.Module]
+        norm_fn = lambda d: nn.Identity()  # noqa
+        if norm:
+            norm_fn = lambda d: nn.GroupNorm(1, d)  # noqa
+
+        hidden = int(channels / compress)
+
+        act: tp.Type[nn.Module]
+        if gelu:
+            act = nn.GELU
+        else:
+            act = nn.ReLU
+
+        self.layers = nn.ModuleList([])
+        for d in range(self.depth):
+            dilation = 2 ** d if dilate else 1
+            padding = dilation * (kernel // 2)
+            mods = [
+                nn.Conv1d(channels, hidden, kernel, dilation=dilation, padding=padding),
+                norm_fn(hidden), act(),
+                nn.Conv1d(hidden, 2 * channels, 1),
+                norm_fn(2 * channels), nn.GLU(1),
+                LayerScale(channels, init),
+            ]
+            if attn:
+                mods.insert(3, LocalState(hidden, heads=heads, ndecay=ndecay))
+            if lstm:
+                mods.insert(3, BLSTM(hidden, layers=2, max_steps=200, skip=True))
+            layer = nn.Sequential(*mods)
+            self.layers.append(layer)
+
+    def forward(self, x):
+        for layer in self.layers:
+            x = x + layer(x)
+        return x
+
+
+class LocalState(nn.Module):
+    """Local state allows to have attention based only on data (no positional embedding),
+    but while setting a constraint on the time window (e.g. decaying penalty term).
+
+    Also a failed experiments with trying to provide some frequency based attention.
+    """
+    def __init__(self, channels: int, heads: int = 4, nfreqs: int = 0, ndecay: int = 4):
+        super().__init__()
+        assert channels % heads == 0, (channels, heads)
+        self.heads = heads
+        self.nfreqs = nfreqs
+        self.ndecay = ndecay
+        self.content = nn.Conv1d(channels, channels, 1)
+        self.query = nn.Conv1d(channels, channels, 1)
+        self.key = nn.Conv1d(channels, channels, 1)
+        if nfreqs:
+            self.query_freqs = nn.Conv1d(channels, heads * nfreqs, 1)
+        if ndecay:
+            self.query_decay = nn.Conv1d(channels, heads * ndecay, 1)
+            # Initialize decay close to zero (there is a sigmoid), for maximum initial window.
+            self.query_decay.weight.data *= 0.01
+            assert self.query_decay.bias is not None  # stupid type checker
+            self.query_decay.bias.data[:] = -2
+        self.proj = nn.Conv1d(channels + heads * nfreqs, channels, 1)
+
+    def forward(self, x):
+        B, C, T = x.shape
+        heads = self.heads
+        indexes = torch.arange(T, device=x.device, dtype=x.dtype)
+        # left index are keys, right index are queries
+        delta = indexes[:, None] - indexes[None, :]
+
+        queries = self.query(x).view(B, heads, -1, T)
+        keys = self.key(x).view(B, heads, -1, T)
+        # t are keys, s are queries
+        dots = torch.einsum("bhct,bhcs->bhts", keys, queries)
+        dots /= keys.shape[2]**0.5
+        if self.nfreqs:
+            periods = torch.arange(1, self.nfreqs + 1, device=x.device, dtype=x.dtype)
+            freq_kernel = torch.cos(2 * math.pi * delta / periods.view(-1, 1, 1))
+            freq_q = self.query_freqs(x).view(B, heads, -1, T) / self.nfreqs ** 0.5
+            dots += torch.einsum("fts,bhfs->bhts", freq_kernel, freq_q)
+        if self.ndecay:
+            decays = torch.arange(1, self.ndecay + 1, device=x.device, dtype=x.dtype)
+            decay_q = self.query_decay(x).view(B, heads, -1, T)
+            decay_q = torch.sigmoid(decay_q) / 2
+            decay_kernel = - decays.view(-1, 1, 1) * delta.abs() / self.ndecay**0.5
+            dots += torch.einsum("fts,bhfs->bhts", decay_kernel, decay_q)
+
+        # Kill self reference.
+        dots.masked_fill_(torch.eye(T, device=dots.device, dtype=torch.bool), -100)
+        weights = torch.softmax(dots, dim=2)
+
+        content = self.content(x).view(B, heads, -1, T)
+        result = torch.einsum("bhts,bhct->bhcs", weights, content)
+        if self.nfreqs:
+            time_sig = torch.einsum("bhts,fts->bhfs", weights, freq_kernel)
+            result = torch.cat([result, time_sig], 2)
+        result = result.reshape(B, -1, T)
+        return x + self.proj(result)
+
+
+class Demucs(nn.Module):
+    @capture_init
+    def __init__(self,
+                 sources,
+                 # Channels
+                 audio_channels=2,
+                 channels=64,
+                 growth=2.,
+                 # Main structure
+                 depth=6,
+                 rewrite=True,
+                 lstm_layers=0,
+                 # Convolutions
+                 kernel_size=8,
+                 stride=4,
+                 context=1,
+                 # Activations
+                 gelu=True,
+                 glu=True,
+                 # Normalization
+                 norm_starts=4,
+                 norm_groups=4,
+                 # DConv residual branch
+                 dconv_mode=1,
+                 dconv_depth=2,
+                 dconv_comp=4,
+                 dconv_attn=4,
+                 dconv_lstm=4,
+                 dconv_init=1e-4,
+                 # Pre/post processing
+                 normalize=True,
+                 resample=True,
+                 # Weight init
+                 rescale=0.1,
+                 # Metadata
+                 samplerate=44100,
+                 segment=4 * 10):
+        """
+        Args:
+            sources (list[str]): list of source names
+            audio_channels (int): stereo or mono
+            channels (int): first convolution channels
+            depth (int): number of encoder/decoder layers
+            growth (float): multiply (resp divide) number of channels by that
+                for each layer of the encoder (resp decoder)
+            depth (int): number of layers in the encoder and in the decoder.
+            rewrite (bool): add 1x1 convolution to each layer.
+            lstm_layers (int): number of lstm layers, 0 = no lstm. Deactivated
+                by default, as this is now replaced by the smaller and faster small LSTMs
+                in the DConv branches.
+            kernel_size (int): kernel size for convolutions
+            stride (int): stride for convolutions
+            context (int): kernel size of the convolution in the
+                decoder before the transposed convolution. If > 1,
+                will provide some context from neighboring time steps.
+            gelu: use GELU activation function.
+            glu (bool): use glu instead of ReLU for the 1x1 rewrite conv.
+            norm_starts: layer at which group norm starts being used.
+                decoder layers are numbered in reverse order.
+            norm_groups: number of groups for group norm.
+            dconv_mode: if 1: dconv in encoder only, 2: decoder only, 3: both.
+            dconv_depth: depth of residual DConv branch.
+            dconv_comp: compression of DConv branch.
+            dconv_attn: adds attention layers in DConv branch starting at this layer.
+            dconv_lstm: adds a LSTM layer in DConv branch starting at this layer.
+            dconv_init: initial scale for the DConv branch LayerScale.
+            normalize (bool): normalizes the input audio on the fly, and scales back
+                the output by the same amount.
+            resample (bool): upsample x2 the input and downsample /2 the output.
+            rescale (int): rescale initial weights of convolutions
+                to get their standard deviation closer to `rescale`.
+            samplerate (int): stored as meta information for easing
+                future evaluations of the model.
+            segment (float): duration of the chunks of audio to ideally evaluate the model on.
+                This is used by `demucs.apply.apply_model`.
+        """
+
+        super().__init__()
+        self.audio_channels = audio_channels
+        self.sources = sources
+        self.kernel_size = kernel_size
+        self.context = context
+        self.stride = stride
+        self.depth = depth
+        self.resample = resample
+        self.channels = channels
+        self.normalize = normalize
+        self.samplerate = samplerate
+        self.segment = segment
+        self.encoder = nn.ModuleList()
+        self.decoder = nn.ModuleList()
+        self.skip_scales = nn.ModuleList()
+
+        if glu:
+            activation = nn.GLU(dim=1)
+            ch_scale = 2
+        else:
+            activation = nn.ReLU()
+            ch_scale = 1
+        if gelu:
+            act2 = nn.GELU
+        else:
+            act2 = nn.ReLU
+
+        in_channels = audio_channels
+        padding = 0
+        for index in range(depth):
+            norm_fn = lambda d: nn.Identity()  # noqa
+            if index >= norm_starts:
+                norm_fn = lambda d: nn.GroupNorm(norm_groups, d)  # noqa
+
+            encode = []
+            encode += [
+                nn.Conv1d(in_channels, channels, kernel_size, stride),
+                norm_fn(channels),
+                act2(),
+            ]
+            attn = index >= dconv_attn
+            lstm = index >= dconv_lstm
+            if dconv_mode & 1:
+                encode += [DConv(channels, depth=dconv_depth, init=dconv_init,
+                                 compress=dconv_comp, attn=attn, lstm=lstm)]
+            if rewrite:
+                encode += [
+                    nn.Conv1d(channels, ch_scale * channels, 1),
+                    norm_fn(ch_scale * channels), activation]
+            self.encoder.append(nn.Sequential(*encode))
+
+            decode = []
+            if index > 0:
+                out_channels = in_channels
+            else:
+                out_channels = len(self.sources) * audio_channels
+            if rewrite:
+                decode += [
+                    nn.Conv1d(channels, ch_scale * channels, 2 * context + 1, padding=context),
+                    norm_fn(ch_scale * channels), activation]
+            if dconv_mode & 2:
+                decode += [DConv(channels, depth=dconv_depth, init=dconv_init,
+                                 compress=dconv_comp, attn=attn, lstm=lstm)]
+            decode += [nn.ConvTranspose1d(channels, out_channels,
+                       kernel_size, stride, padding=padding)]
+            if index > 0:
+                decode += [norm_fn(out_channels), act2()]
+            self.decoder.insert(0, nn.Sequential(*decode))
+            in_channels = channels
+            channels = int(growth * channels)
+
+        channels = in_channels
+        if lstm_layers:
+            self.lstm = BLSTM(channels, lstm_layers)
+        else:
+            self.lstm = None
+
+        if rescale:
+            rescale_module(self, reference=rescale)
+
+    def valid_length(self, length):
+        """
+        Return the nearest valid length to use with the model so that
+        there is no time steps left over in a convolution, e.g. for all
+        layers, size of the input - kernel_size % stride = 0.
+
+        Note that input are automatically padded if necessary to ensure that the output
+        has the same length as the input.
+        """
+        if self.resample:
+            length *= 2
+
+        for _ in range(self.depth):
+            length = math.ceil((length - self.kernel_size) / self.stride) + 1
+            length = max(1, length)
+
+        for idx in range(self.depth):
+            length = (length - 1) * self.stride + self.kernel_size
+
+        if self.resample:
+            length = math.ceil(length / 2)
+        return int(length)
+
+    def forward(self, mix):
+        x = mix
+        length = x.shape[-1]
+
+        if self.normalize:
+            mono = mix.mean(dim=1, keepdim=True)
+            mean = mono.mean(dim=-1, keepdim=True)
+            std = mono.std(dim=-1, keepdim=True)
+            x = (x - mean) / (1e-5 + std)
+        else:
+            mean = 0
+            std = 1
+
+        delta = self.valid_length(length) - length
+        x = F.pad(x, (delta // 2, delta - delta // 2))
+
+        if self.resample:
+            x = julius.resample_frac(x, 1, 2)
+
+        saved = []
+        for encode in self.encoder:
+            x = encode(x)
+            saved.append(x)
+
+        if self.lstm:
+            x = self.lstm(x)
+
+        for decode in self.decoder:
+            skip = saved.pop(-1)
+            skip = center_trim(skip, x)
+            x = decode(x + skip)
+
+        if self.resample:
+            x = julius.resample_frac(x, 2, 1)
+        x = x * std + mean
+        x = center_trim(x, length)
+        x = x.view(x.size(0), len(self.sources), self.audio_channels, x.size(-1))
+        return x
+
+    def load_state_dict(self, state, strict=True):
+        # fix a mismatch with previous generation Demucs models.
+        for idx in range(self.depth):
+            for a in ['encoder', 'decoder']:
+                for b in ['bias', 'weight']:
+                    new = f'{a}.{idx}.3.{b}'
+                    old = f'{a}.{idx}.2.{b}'
+                    if old in state and new not in state:
+                        state[new] = state.pop(old)
+        super().load_state_dict(state, strict=strict)

demucs/filtering.py

@@ -0,0 +1,502 @@
+from typing import Optional
+import torch
+import torch.nn as nn
+from torch import Tensor
+from torch.utils.data import DataLoader
+
+def atan2(y, x):
+    r"""Element-wise arctangent function of y/x.
+    Returns a new tensor with signed angles in radians.
+    It is an alternative implementation of torch.atan2
+
+    Args:
+        y (Tensor): First input tensor
+        x (Tensor): Second input tensor [shape=y.shape]
+
+    Returns:
+        Tensor: [shape=y.shape].
+    """
+    pi = 2 * torch.asin(torch.tensor(1.0))
+    x += ((x == 0) & (y == 0)) * 1.0
+    out = torch.atan(y / x)
+    out += ((y >= 0) & (x < 0)) * pi
+    out -= ((y < 0) & (x < 0)) * pi
+    out *= 1 - ((y > 0) & (x == 0)) * 1.0
+    out += ((y > 0) & (x == 0)) * (pi / 2)
+    out *= 1 - ((y < 0) & (x == 0)) * 1.0
+    out += ((y < 0) & (x == 0)) * (-pi / 2)
+    return out
+
+
+# Define basic complex operations on torch.Tensor objects whose last dimension
+# consists in the concatenation of the real and imaginary parts.
+
+
+def _norm(x: torch.Tensor) -> torch.Tensor:
+    r"""Computes the norm value of a torch Tensor, assuming that it
+    comes as real and imaginary part in its last dimension.
+
+    Args:
+        x (Tensor): Input Tensor of shape [shape=(..., 2)]
+
+    Returns:
+        Tensor: shape as x excluding the last dimension.
+    """
+    return torch.abs(x[..., 0]) ** 2 + torch.abs(x[..., 1]) ** 2
+
+
+def _mul_add(a: torch.Tensor, b: torch.Tensor, out: Optional[torch.Tensor] = None) -> torch.Tensor:
+    """Element-wise multiplication of two complex Tensors described
+    through their real and imaginary parts.
+    The result is added to the `out` tensor"""
+
+    # check `out` and allocate it if needed
+    target_shape = torch.Size([max(sa, sb) for (sa, sb) in zip(a.shape, b.shape)])
+    if out is None or out.shape != target_shape:
+        out = torch.zeros(target_shape, dtype=a.dtype, device=a.device)
+    if out is a:
+        real_a = a[..., 0]
+        out[..., 0] = out[..., 0] + (real_a * b[..., 0] - a[..., 1] * b[..., 1])
+        out[..., 1] = out[..., 1] + (real_a * b[..., 1] + a[..., 1] * b[..., 0])
+    else:
+        out[..., 0] = out[..., 0] + (a[..., 0] * b[..., 0] - a[..., 1] * b[..., 1])
+        out[..., 1] = out[..., 1] + (a[..., 0] * b[..., 1] + a[..., 1] * b[..., 0])
+    return out
+
+
+def _mul(a: torch.Tensor, b: torch.Tensor, out: Optional[torch.Tensor] = None) -> torch.Tensor:
+    """Element-wise multiplication of two complex Tensors described
+    through their real and imaginary parts
+    can work in place in case out is a only"""
+    target_shape = torch.Size([max(sa, sb) for (sa, sb) in zip(a.shape, b.shape)])
+    if out is None or out.shape != target_shape:
+        out = torch.zeros(target_shape, dtype=a.dtype, device=a.device)
+    if out is a:
+        real_a = a[..., 0]
+        out[..., 0] = real_a * b[..., 0] - a[..., 1] * b[..., 1]
+        out[..., 1] = real_a * b[..., 1] + a[..., 1] * b[..., 0]
+    else:
+        out[..., 0] = a[..., 0] * b[..., 0] - a[..., 1] * b[..., 1]
+        out[..., 1] = a[..., 0] * b[..., 1] + a[..., 1] * b[..., 0]
+    return out
+
+
+def _inv(z: torch.Tensor, out: Optional[torch.Tensor] = None) -> torch.Tensor:
+    """Element-wise multiplicative inverse of a Tensor with complex
+    entries described through their real and imaginary parts.
+    can work in place in case out is z"""
+    ez = _norm(z)
+    if out is None or out.shape != z.shape:
+        out = torch.zeros_like(z)
+    out[..., 0] = z[..., 0] / ez
+    out[..., 1] = -z[..., 1] / ez
+    return out
+
+
+def _conj(z, out: Optional[torch.Tensor] = None) -> torch.Tensor:
+    """Element-wise complex conjugate of a Tensor with complex entries
+    described through their real and imaginary parts.
+    can work in place in case out is z"""
+    if out is None or out.shape != z.shape:
+        out = torch.zeros_like(z)
+    out[..., 0] = z[..., 0]
+    out[..., 1] = -z[..., 1]
+    return out
+
+
+def _invert(M: torch.Tensor, out: Optional[torch.Tensor] = None) -> torch.Tensor:
+    """
+    Invert 1x1 or 2x2 matrices
+
+    Will generate errors if the matrices are singular: user must handle this
+    through his own regularization schemes.
+
+    Args:
+        M (Tensor): [shape=(..., nb_channels, nb_channels, 2)]
+            matrices to invert: must be square along dimensions -3 and -2
+
+    Returns:
+        invM (Tensor): [shape=M.shape]
+            inverses of M
+    """
+    nb_channels = M.shape[-2]
+
+    if out is None or out.shape != M.shape:
+        out = torch.empty_like(M)
+
+    if nb_channels == 1:
+        # scalar case
+        out = _inv(M, out)
+    elif nb_channels == 2:
+        # two channels case: analytical expression
+
+        # first compute the determinent
+        det = _mul(M[..., 0, 0, :], M[..., 1, 1, :])
+        det = det - _mul(M[..., 0, 1, :], M[..., 1, 0, :])
+        # invert it
+        invDet = _inv(det)
+
+        # then fill out the matrix with the inverse
+        out[..., 0, 0, :] = _mul(invDet, M[..., 1, 1, :], out[..., 0, 0, :])
+        out[..., 1, 0, :] = _mul(-invDet, M[..., 1, 0, :], out[..., 1, 0, :])
+        out[..., 0, 1, :] = _mul(-invDet, M[..., 0, 1, :], out[..., 0, 1, :])
+        out[..., 1, 1, :] = _mul(invDet, M[..., 0, 0, :], out[..., 1, 1, :])
+    else:
+        raise Exception("Only 2 channels are supported for the torch version.")
+    return out
+
+
+# Now define the signal-processing low-level functions used by the Separator
+
+
+def expectation_maximization(
+    y: torch.Tensor,
+    x: torch.Tensor,
+    iterations: int = 2,
+    eps: float = 1e-10,
+    batch_size: int = 200,
+):
+    r"""Expectation maximization algorithm, for refining source separation
+    estimates.
+
+    This algorithm allows to make source separation results better by
+    enforcing multichannel consistency for the estimates. This usually means
+    a better perceptual quality in terms of spatial artifacts.
+
+    The implementation follows the details presented in [1]_, taking
+    inspiration from the original EM algorithm proposed in [2]_ and its
+    weighted refinement proposed in [3]_, [4]_.
+    It works by iteratively:
+
+     * Re-estimate source parameters (power spectral densities and spatial
+       covariance matrices) through :func:`get_local_gaussian_model`.
+
+     * Separate again the mixture with the new parameters by first computing
+       the new modelled mixture covariance matrices with :func:`get_mix_model`,
+       prepare the Wiener filters through :func:`wiener_gain` and apply them
+       with :func:`apply_filter``.
+
+    References
+    ----------
+    .. [1] S. Uhlich and M. Porcu and F. Giron and M. Enenkl and T. Kemp and
+        N. Takahashi and Y. Mitsufuji, "Improving music source separation based
+        on deep neural networks through data augmentation and network
+        blending." 2017 IEEE International Conference on Acoustics, Speech
+        and Signal Processing (ICASSP). IEEE, 2017.
+
+    .. [2] N.Q. Duong and E. Vincent and R.Gribonval. "Under-determined
+        reverberant audio source separation using a full-rank spatial
+        covariance model." IEEE Transactions on Audio, Speech, and Language
+        Processing 18.7 (2010): 1830-1840.
+
+    .. [3] A. Nugraha and A. Liutkus and E. Vincent. "Multichannel audio source
+        separation with deep neural networks." IEEE/ACM Transactions on Audio,
+        Speech, and Language Processing 24.9 (2016): 1652-1664.
+
+    .. [4] A. Nugraha and A. Liutkus and E. Vincent. "Multichannel music
+        separation with deep neural networks." 2016 24th European Signal
+        Processing Conference (EUSIPCO). IEEE, 2016.
+
+    .. [5] A. Liutkus and R. Badeau and G. Richard "Kernel additive models for
+        source separation." IEEE Transactions on Signal Processing
+        62.16 (2014): 4298-4310.
+
+    Args:
+        y (Tensor): [shape=(nb_frames, nb_bins, nb_channels, 2, nb_sources)]
+            initial estimates for the sources
+        x (Tensor): [shape=(nb_frames, nb_bins, nb_channels, 2)]
+            complex STFT of the mixture signal
+        iterations (int): [scalar]
+            number of iterations for the EM algorithm.
+        eps (float or None): [scalar]
+            The epsilon value to use for regularization and filters.
+
+    Returns:
+        y (Tensor): [shape=(nb_frames, nb_bins, nb_channels, 2, nb_sources)]
+            estimated sources after iterations
+        v (Tensor): [shape=(nb_frames, nb_bins, nb_sources)]
+            estimated power spectral densities
+        R (Tensor): [shape=(nb_bins, nb_channels, nb_channels, 2, nb_sources)]
+            estimated spatial covariance matrices
+
+    Notes:
+        * You need an initial estimate for the sources to apply this
+          algorithm. This is precisely what the :func:`wiener` function does.
+        * This algorithm *is not* an implementation of the "exact" EM
+          proposed in [1]_. In particular, it does compute the posterior
+          covariance matrices the same (exact) way. Instead, it uses the
+          simplified approximate scheme initially proposed in [5]_ and further
+          refined in [3]_, [4]_, that boils down to just take the empirical
+          covariance of the recent source estimates, followed by a weighted
+          average for the update of the spatial covariance matrix. It has been
+          empirically demonstrated that this simplified algorithm is more
+          robust for music separation.
+
+    Warning:
+        It is *very* important to make sure `x.dtype` is `torch.float64`
+        if you want double precision, because this function will **not**
+        do such conversion for you from `torch.complex32`, in case you want the
+        smaller RAM usage on purpose.
+
+        It is usually always better in terms of quality to have double
+        precision, by e.g. calling :func:`expectation_maximization`
+        with ``x.to(torch.float64)``.
+    """
+    # dimensions
+    (nb_frames, nb_bins, nb_channels) = x.shape[:-1]
+    nb_sources = y.shape[-1]
+
+    regularization = torch.cat(
+        (
+            torch.eye(nb_channels, dtype=x.dtype, device=x.device)[..., None],
+            torch.zeros((nb_channels, nb_channels, 1), dtype=x.dtype, device=x.device),
+        ),
+        dim=2,
+    )
+    regularization = torch.sqrt(torch.as_tensor(eps)) * (
+        regularization[None, None, ...].expand((-1, nb_bins, -1, -1, -1))
+    )
+
+    # allocate the spatial covariance matrices
+    R = [
+        torch.zeros((nb_bins, nb_channels, nb_channels, 2), dtype=x.dtype, device=x.device)
+        for j in range(nb_sources)
+    ]
+    weight: torch.Tensor = torch.zeros((nb_bins,), dtype=x.dtype, device=x.device)
+
+    v: torch.Tensor = torch.zeros((nb_frames, nb_bins, nb_sources), dtype=x.dtype, device=x.device)
+    for it in range(iterations):
+        # constructing the mixture covariance matrix. Doing it with a loop
+        # to avoid storing anytime in RAM the whole 6D tensor
+
+        # update the PSD as the average spectrogram over channels
+        v = torch.mean(torch.abs(y[..., 0, :]) ** 2 + torch.abs(y[..., 1, :]) ** 2, dim=-2)
+
+        # update spatial covariance matrices (weighted update)
+        for j in range(nb_sources):
+            R[j] = torch.tensor(0.0, device=x.device)
+            weight = torch.tensor(eps, device=x.device)
+            pos: int = 0
+            batch_size = batch_size if batch_size else nb_frames
+            while pos < nb_frames:
+                t = torch.arange(pos, min(nb_frames, pos + batch_size))
+                pos = int(t[-1]) + 1
+
+                R[j] = R[j] + torch.sum(_covariance(y[t, ..., j]), dim=0)
+                weight = weight + torch.sum(v[t, ..., j], dim=0)
+            R[j] = R[j] / weight[..., None, None, None]
+            weight = torch.zeros_like(weight)
+
+        # cloning y if we track gradient, because we're going to update it
+        if y.requires_grad:
+            y = y.clone()
+
+        pos = 0
+        while pos < nb_frames:
+            t = torch.arange(pos, min(nb_frames, pos + batch_size))
+            pos = int(t[-1]) + 1
+
+            y[t, ...] = torch.tensor(0.0, device=x.device, dtype=x.dtype)
+
+            # compute mix covariance matrix
+            Cxx = regularization
+            for j in range(nb_sources):
+                Cxx = Cxx + (v[t, ..., j, None, None, None] * R[j][None, ...].clone())
+
+            # invert it
+            inv_Cxx = _invert(Cxx)
+
+            # separate the sources
+            for j in range(nb_sources):
+
+                # create a wiener gain for this source
+                gain = torch.zeros_like(inv_Cxx)
+
+                # computes multichannel Wiener gain as v_j R_j inv_Cxx
+                indices = torch.cartesian_prod(
+                    torch.arange(nb_channels),
+                    torch.arange(nb_channels),
+                    torch.arange(nb_channels),
+                )
+                for index in indices:
+                    gain[:, :, index[0], index[1], :] = _mul_add(
+                        R[j][None, :, index[0], index[2], :].clone(),
+                        inv_Cxx[:, :, index[2], index[1], :],
+                        gain[:, :, index[0], index[1], :],
+                    )
+                gain = gain * v[t, ..., None, None, None, j]
+
+                # apply it to the mixture
+                for i in range(nb_channels):
+                    y[t, ..., j] = _mul_add(gain[..., i, :], x[t, ..., i, None, :], y[t, ..., j])
+
+    return y, v, R
+
+
+def wiener(
+    targets_spectrograms: torch.Tensor,
+    mix_stft: torch.Tensor,
+    iterations: int = 1,
+    softmask: bool = False,
+    residual: bool = False,
+    scale_factor: float = 10.0,
+    eps: float = 1e-10,
+):
+    """Wiener-based separation for multichannel audio.
+
+    The method uses the (possibly multichannel) spectrograms  of the
+    sources to separate the (complex) Short Term Fourier Transform  of the
+    mix. Separation is done in a sequential way by:
+
+    * Getting an initial estimate. This can be done in two ways: either by
+      directly using the spectrograms with the mixture phase, or
+      by using a softmasking strategy. This initial phase is controlled
+      by the `softmask` flag.
+
+    * If required, adding an additional residual target as the mix minus
+      all targets.
+
+    * Refinining these initial estimates through a call to
+      :func:`expectation_maximization` if the number of iterations is nonzero.
+
+    This implementation also allows to specify the epsilon value used for
+    regularization. It is based on [1]_, [2]_, [3]_, [4]_.
+
+    References
+    ----------
+    .. [1] S. Uhlich and M. Porcu and F. Giron and M. Enenkl and T. Kemp and
+        N. Takahashi and Y. Mitsufuji, "Improving music source separation based
+        on deep neural networks through data augmentation and network
+        blending." 2017 IEEE International Conference on Acoustics, Speech
+        and Signal Processing (ICASSP). IEEE, 2017.
+
+    .. [2] A. Nugraha and A. Liutkus and E. Vincent. "Multichannel audio source
+        separation with deep neural networks." IEEE/ACM Transactions on Audio,
+        Speech, and Language Processing 24.9 (2016): 1652-1664.
+
+    .. [3] A. Nugraha and A. Liutkus and E. Vincent. "Multichannel music
+        separation with deep neural networks." 2016 24th European Signal
+        Processing Conference (EUSIPCO). IEEE, 2016.
+
+    .. [4] A. Liutkus and R. Badeau and G. Richard "Kernel additive models for
+        source separation." IEEE Transactions on Signal Processing
+        62.16 (2014): 4298-4310.
+
+    Args:
+        targets_spectrograms (Tensor): spectrograms of the sources
+            [shape=(nb_frames, nb_bins, nb_channels, nb_sources)].
+            This is a nonnegative tensor that is
+            usually the output of the actual separation method of the user. The
+            spectrograms may be mono, but they need to be 4-dimensional in all
+            cases.
+        mix_stft (Tensor): [shape=(nb_frames, nb_bins, nb_channels, complex=2)]
+            STFT of the mixture signal.
+        iterations (int): [scalar]
+            number of iterations for the EM algorithm
+        softmask (bool): Describes how the initial estimates are obtained.
+            * if `False`, then the mixture phase will directly be used with the
+            spectrogram as initial estimates.
+            * if `True`, initial estimates are obtained by multiplying the
+            complex mix element-wise with the ratio of each target spectrogram
+            with the sum of them all. This strategy is better if the model are
+            not really good, and worse otherwise.
+        residual (bool): if `True`, an additional target is created, which is
+            equal to the mixture minus the other targets, before application of
+            expectation maximization
+        eps (float): Epsilon value to use for computing the separations.
+            This is used whenever division with a model energy is
+            performed, i.e. when softmasking and when iterating the EM.
+            It can be understood as the energy of the additional white noise
+            that is taken out when separating.
+
+    Returns:
+        Tensor: shape=(nb_frames, nb_bins, nb_channels, complex=2, nb_sources)
+            STFT of estimated sources
+
+    Notes:
+        * Be careful that you need *magnitude spectrogram estimates* for the
+        case `softmask==False`.
+        * `softmask=False` is recommended
+        * The epsilon value will have a huge impact on performance. If it's
+        large, only the parts of the signal with a significant energy will
+        be kept in the sources. This epsilon then directly controls the
+        energy of the reconstruction error.
+
+    Warning:
+        As in :func:`expectation_maximization`, we recommend converting the
+        mixture `x` to double precision `torch.float64` *before* calling
+        :func:`wiener`.
+    """
+    if softmask:
+        # if we use softmask, we compute the ratio mask for all targets and
+        # multiply by the mix stft
+        y = (
+            mix_stft[..., None]
+            * (
+                targets_spectrograms
+                / (eps + torch.sum(targets_spectrograms, dim=-1, keepdim=True).to(mix_stft.dtype))
+            )[..., None, :]
+        )
+    else:
+        # otherwise, we just multiply the targets spectrograms with mix phase
+        # we tacitly assume that we have magnitude estimates.
+        angle = atan2(mix_stft[..., 1], mix_stft[..., 0])[..., None]
+        nb_sources = targets_spectrograms.shape[-1]
+        y = torch.zeros(
+            mix_stft.shape + (nb_sources,), dtype=mix_stft.dtype, device=mix_stft.device
+        )
+        y[..., 0, :] = targets_spectrograms * torch.cos(angle)
+        y[..., 1, :] = targets_spectrograms * torch.sin(angle)
+
+    if residual:
+        # if required, adding an additional target as the mix minus
+        # available targets
+        y = torch.cat([y, mix_stft[..., None] - y.sum(dim=-1, keepdim=True)], dim=-1)
+
+    if iterations == 0:
+        return y
+
+    # we need to refine the estimates. Scales down the estimates for
+    # numerical stability
+    max_abs = torch.max(
+        torch.as_tensor(1.0, dtype=mix_stft.dtype, device=mix_stft.device),
+        torch.sqrt(_norm(mix_stft)).max() / scale_factor,
+    )
+
+    mix_stft = mix_stft / max_abs
+    y = y / max_abs
+
+    # call expectation maximization
+    y = expectation_maximization(y, mix_stft, iterations, eps=eps)[0]
+
+    # scale estimates up again
+    y = y * max_abs
+    return y
+
+
+def _covariance(y_j):
+    """
+    Compute the empirical covariance for a source.
+
+    Args:
+        y_j (Tensor): complex stft of the source.
+            [shape=(nb_frames, nb_bins, nb_channels, 2)].
+
+    Returns:
+        Cj (Tensor): [shape=(nb_frames, nb_bins, nb_channels, nb_channels, 2)]
+            just y_j * conj(y_j.T): empirical covariance for each TF bin.
+    """
+    (nb_frames, nb_bins, nb_channels) = y_j.shape[:-1]
+    Cj = torch.zeros(
+        (nb_frames, nb_bins, nb_channels, nb_channels, 2),
+        dtype=y_j.dtype,
+        device=y_j.device,
+    )
+    indices = torch.cartesian_prod(torch.arange(nb_channels), torch.arange(nb_channels))
+    for index in indices:
+        Cj[:, :, index[0], index[1], :] = _mul_add(
+            y_j[:, :, index[0], :],
+            _conj(y_j[:, :, index[1], :]),
+            Cj[:, :, index[0], index[1], :],
+        )
+    return Cj

demucs/hdemucs.py

@@ -0,0 +1,782 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+"""
+This code contains the spectrogram and Hybrid version of Demucs.
+"""
+from copy import deepcopy
+import math
+import typing as tp
+import torch
+from torch import nn
+from torch.nn import functional as F
+from .filtering import wiener
+from .demucs import DConv, rescale_module
+from .states import capture_init
+from .spec import spectro, ispectro
+
+def pad1d(x: torch.Tensor, paddings: tp.Tuple[int, int], mode: str = 'constant', value: float = 0.):
+    """Tiny wrapper around F.pad, just to allow for reflect padding on small input.
+    If this is the case, we insert extra 0 padding to the right before the reflection happen."""
+    x0 = x
+    length = x.shape[-1]
+    padding_left, padding_right = paddings
+    if mode == 'reflect':
+        max_pad = max(padding_left, padding_right)
+        if length <= max_pad:
+            extra_pad = max_pad - length + 1
+            extra_pad_right = min(padding_right, extra_pad)
+            extra_pad_left = extra_pad - extra_pad_right
+            paddings = (padding_left - extra_pad_left, padding_right - extra_pad_right)
+            x = F.pad(x, (extra_pad_left, extra_pad_right))
+    out = F.pad(x, paddings, mode, value)
+    assert out.shape[-1] == length + padding_left + padding_right
+    assert (out[..., padding_left: padding_left + length] == x0).all()
+    return out
+
+class ScaledEmbedding(nn.Module):
+    """
+    Boost learning rate for embeddings (with `scale`).
+    Also, can make embeddings continuous with `smooth`.
+    """
+    def __init__(self, num_embeddings: int, embedding_dim: int,
+                 scale: float = 10., smooth=False):
+        super().__init__()
+        self.embedding = nn.Embedding(num_embeddings, embedding_dim)
+        if smooth:
+            weight = torch.cumsum(self.embedding.weight.data, dim=0)
+            # when summing gaussian, overscale raises as sqrt(n), so we nornalize by that.
+            weight = weight / torch.arange(1, num_embeddings + 1).to(weight).sqrt()[:, None]
+            self.embedding.weight.data[:] = weight
+        self.embedding.weight.data /= scale
+        self.scale = scale
+
+    @property
+    def weight(self):
+        return self.embedding.weight * self.scale
+
+    def forward(self, x):
+        out = self.embedding(x) * self.scale
+        return out
+
+
+class HEncLayer(nn.Module):
+    def __init__(self, chin, chout, kernel_size=8, stride=4, norm_groups=1, empty=False,
+                 freq=True, dconv=True, norm=True, context=0, dconv_kw={}, pad=True,
+                 rewrite=True):
+        """Encoder layer. This used both by the time and the frequency branch.
+
+        Args:
+            chin: number of input channels.
+            chout: number of output channels.
+            norm_groups: number of groups for group norm.
+            empty: used to make a layer with just the first conv. this is used
+                before merging the time and freq. branches.
+            freq: this is acting on frequencies.
+            dconv: insert DConv residual branches.
+            norm: use GroupNorm.
+            context: context size for the 1x1 conv.
+            dconv_kw: list of kwargs for the DConv class.
+            pad: pad the input. Padding is done so that the output size is
+                always the input size / stride.
+            rewrite: add 1x1 conv at the end of the layer.
+        """
+        super().__init__()
+        norm_fn = lambda d: nn.Identity()  # noqa
+        if norm:
+            norm_fn = lambda d: nn.GroupNorm(norm_groups, d)  # noqa
+        if pad:
+            pad = kernel_size // 4
+        else:
+            pad = 0
+        klass = nn.Conv1d
+        self.freq = freq
+        self.kernel_size = kernel_size
+        self.stride = stride
+        self.empty = empty
+        self.norm = norm
+        self.pad = pad
+        if freq:
+            kernel_size = [kernel_size, 1]
+            stride = [stride, 1]
+            pad = [pad, 0]
+            klass = nn.Conv2d
+        self.conv = klass(chin, chout, kernel_size, stride, pad)
+        if self.empty:
+            return
+        self.norm1 = norm_fn(chout)
+        self.rewrite = None
+        if rewrite:
+            self.rewrite = klass(chout, 2 * chout, 1 + 2 * context, 1, context)
+            self.norm2 = norm_fn(2 * chout)
+
+        self.dconv = None
+        if dconv:
+            self.dconv = DConv(chout, **dconv_kw)
+
+    def forward(self, x, inject=None):
+        """
+        `inject` is used to inject the result from the time branch into the frequency branch,
+        when both have the same stride.
+        """
+        if not self.freq and x.dim() == 4:
+            B, C, Fr, T = x.shape
+            x = x.view(B, -1, T)
+
+        if not self.freq:
+            le = x.shape[-1]
+            if not le % self.stride == 0:
+                x = F.pad(x, (0, self.stride - (le % self.stride)))
+        y = self.conv(x)
+        if self.empty:
+            return y
+        if inject is not None:
+            assert inject.shape[-1] == y.shape[-1], (inject.shape, y.shape)
+            if inject.dim() == 3 and y.dim() == 4:
+                inject = inject[:, :, None]
+            y = y + inject
+        y = F.gelu(self.norm1(y))
+        if self.dconv:
+            if self.freq:
+                B, C, Fr, T = y.shape
+                y = y.permute(0, 2, 1, 3).reshape(-1, C, T)
+            y = self.dconv(y)
+            if self.freq:
+                y = y.view(B, Fr, C, T).permute(0, 2, 1, 3)
+        if self.rewrite:
+            z = self.norm2(self.rewrite(y))
+            z = F.glu(z, dim=1)
+        else:
+            z = y
+        return z
+
+
+class MultiWrap(nn.Module):
+    """
+    Takes one layer and replicate it N times. each replica will act
+    on a frequency band. All is done so that if the N replica have the same weights,
+    then this is exactly equivalent to applying the original module on all frequencies.
+
+    This is a bit over-engineered to avoid edge artifacts when splitting
+    the frequency bands, but it is possible the naive implementation would work as well...
+    """
+    def __init__(self, layer, split_ratios):
+        """
+        Args:
+            layer: module to clone, must be either HEncLayer or HDecLayer.
+            split_ratios: list of float indicating which ratio to keep for each band.
+        """
+        super().__init__()
+        self.split_ratios = split_ratios
+        self.layers = nn.ModuleList()
+        self.conv = isinstance(layer, HEncLayer)
+        assert not layer.norm
+        assert layer.freq
+        assert layer.pad
+        if not self.conv:
+            assert not layer.context_freq
+        for k in range(len(split_ratios) + 1):
+            lay = deepcopy(layer)
+            if self.conv:
+                lay.conv.padding = (0, 0)
+            else:
+                lay.pad = False
+            for m in lay.modules():
+                if hasattr(m, 'reset_parameters'):
+                    m.reset_parameters()
+            self.layers.append(lay)
+
+    def forward(self, x, skip=None, length=None):
+        B, C, Fr, T = x.shape
+
+        ratios = list(self.split_ratios) + [1]
+        start = 0
+        outs = []
+        for ratio, layer in zip(ratios, self.layers):
+            if self.conv:
+                pad = layer.kernel_size // 4
+                if ratio == 1:
+                    limit = Fr
+                    frames = -1
+                else:
+                    limit = int(round(Fr * ratio))
+                    le = limit - start
+                    if start == 0:
+                        le += pad
+                    frames = round((le - layer.kernel_size) / layer.stride + 1)
+                    limit = start + (frames - 1) * layer.stride + layer.kernel_size
+                    if start == 0:
+                        limit -= pad
+                assert limit - start > 0, (limit, start)
+                assert limit <= Fr, (limit, Fr)
+                y = x[:, :, start:limit, :]
+                if start == 0:
+                    y = F.pad(y, (0, 0, pad, 0))
+                if ratio == 1:
+                    y = F.pad(y, (0, 0, 0, pad))
+                outs.append(layer(y))
+                start = limit - layer.kernel_size + layer.stride
+            else:
+                if ratio == 1:
+                    limit = Fr
+                else:
+                    limit = int(round(Fr * ratio))
+                last = layer.last
+                layer.last = True
+
+                y = x[:, :, start:limit]
+                s = skip[:, :, start:limit]
+                out, _ = layer(y, s, None)
+                if outs:
+                    outs[-1][:, :, -layer.stride:] += (
+                        out[:, :, :layer.stride] - layer.conv_tr.bias.view(1, -1, 1, 1))
+                    out = out[:, :, layer.stride:]
+                if ratio == 1:
+                    out = out[:, :, :-layer.stride // 2, :]
+                if start == 0:
+                    out = out[:, :, layer.stride // 2:, :]
+                outs.append(out)
+                layer.last = last
+                start = limit
+        out = torch.cat(outs, dim=2)
+        if not self.conv and not last:
+            out = F.gelu(out)
+        if self.conv:
+            return out
+        else:
+            return out, None
+
+
+class HDecLayer(nn.Module):
+    def __init__(self, chin, chout, last=False, kernel_size=8, stride=4, norm_groups=1, empty=False,
+                 freq=True, dconv=True, norm=True, context=1, dconv_kw={}, pad=True,
+                 context_freq=True, rewrite=True):
+        """
+        Same as HEncLayer but for decoder. See `HEncLayer` for documentation.
+        """
+        super().__init__()
+        norm_fn = lambda d: nn.Identity()  # noqa
+        if norm:
+            norm_fn = lambda d: nn.GroupNorm(norm_groups, d)  # noqa
+        if pad:
+            pad = kernel_size // 4
+        else:
+            pad = 0
+        self.pad = pad
+        self.last = last
+        self.freq = freq
+        self.chin = chin
+        self.empty = empty
+        self.stride = stride
+        self.kernel_size = kernel_size
+        self.norm = norm
+        self.context_freq = context_freq
+        klass = nn.Conv1d
+        klass_tr = nn.ConvTranspose1d
+        if freq:
+            kernel_size = [kernel_size, 1]
+            stride = [stride, 1]
+            klass = nn.Conv2d
+            klass_tr = nn.ConvTranspose2d
+        self.conv_tr = klass_tr(chin, chout, kernel_size, stride)
+        self.norm2 = norm_fn(chout)
+        if self.empty:
+            return
+        self.rewrite = None
+        if rewrite:
+            if context_freq:
+                self.rewrite = klass(chin, 2 * chin, 1 + 2 * context, 1, context)
+            else:
+                self.rewrite = klass(chin, 2 * chin, [1, 1 + 2 * context], 1,
+                                     [0, context])
+            self.norm1 = norm_fn(2 * chin)
+
+        self.dconv = None
+        if dconv:
+            self.dconv = DConv(chin, **dconv_kw)
+
+    def forward(self, x, skip, length):
+        if self.freq and x.dim() == 3:
+            B, C, T = x.shape
+            x = x.view(B, self.chin, -1, T)
+
+        if not self.empty:
+            x = x + skip
+
+            if self.rewrite:
+                y = F.glu(self.norm1(self.rewrite(x)), dim=1)
+            else:
+                y = x
+            if self.dconv:
+                if self.freq:
+                    B, C, Fr, T = y.shape
+                    y = y.permute(0, 2, 1, 3).reshape(-1, C, T)
+                y = self.dconv(y)
+                if self.freq:
+                    y = y.view(B, Fr, C, T).permute(0, 2, 1, 3)
+        else:
+            y = x
+            assert skip is None
+        z = self.norm2(self.conv_tr(y))
+        if self.freq:
+            if self.pad:
+                z = z[..., self.pad:-self.pad, :]
+        else:
+            z = z[..., self.pad:self.pad + length]
+            assert z.shape[-1] == length, (z.shape[-1], length)
+        if not self.last:
+            z = F.gelu(z)
+        return z, y
+
+
+class HDemucs(nn.Module):
+    """
+    Spectrogram and hybrid Demucs model.
+    The spectrogram model has the same structure as Demucs, except the first few layers are over the
+    frequency axis, until there is only 1 frequency, and then it moves to time convolutions.
+    Frequency layers can still access information across time steps thanks to the DConv residual.
+
+    Hybrid model have a parallel time branch. At some layer, the time branch has the same stride
+    as the frequency branch and then the two are combined. The opposite happens in the decoder.
+
+    Models can either use naive iSTFT from masking, Wiener filtering ([Ulhih et al. 2017]),
+    or complex as channels (CaC) [Choi et al. 2020]. Wiener filtering is based on
+    Open Unmix implementation [Stoter et al. 2019].
+
+    The loss is always on the temporal domain, by backpropagating through the above
+    output methods and iSTFT. This allows to define hybrid models nicely. However, this breaks
+    a bit Wiener filtering, as doing more iteration at test time will change the spectrogram
+    contribution, without changing the one from the waveform, which will lead to worse performance.
+    I tried using the residual option in OpenUnmix Wiener implementation, but it didn't improve.
+    CaC on the other hand provides similar performance for hybrid, and works naturally with
+    hybrid models.
+
+    This model also uses frequency embeddings are used to improve efficiency on convolutions
+    over the freq. axis, following [Isik et al. 2020] (https://arxiv.org/pdf/2008.04470.pdf).
+
+    Unlike classic Demucs, there is no resampling here, and normalization is always applied.
+    """
+    @capture_init
+    def __init__(self,
+                 sources,
+                 # Channels
+                 audio_channels=2,
+                 channels=48,
+                 channels_time=None,
+                 growth=2,
+                 # STFT
+                 nfft=4096,
+                 wiener_iters=0,
+                 end_iters=0,
+                 wiener_residual=False,
+                 cac=True,
+                 # Main structure
+                 depth=6,
+                 rewrite=True,
+                 hybrid=True,
+                 hybrid_old=False,
+                 # Frequency branch
+                 multi_freqs=None,
+                 multi_freqs_depth=2,
+                 freq_emb=0.2,
+                 emb_scale=10,
+                 emb_smooth=True,
+                 # Convolutions
+                 kernel_size=8,
+                 time_stride=2,
+                 stride=4,
+                 context=1,
+                 context_enc=0,
+                 # Normalization
+                 norm_starts=4,
+                 norm_groups=4,
+                 # DConv residual branch
+                 dconv_mode=1,
+                 dconv_depth=2,
+                 dconv_comp=4,
+                 dconv_attn=4,
+                 dconv_lstm=4,
+                 dconv_init=1e-4,
+                 # Weight init
+                 rescale=0.1,
+                 # Metadata
+                 samplerate=44100,
+                 segment=4 * 10):
+        
+        """
+        Args:
+            sources (list[str]): list of source names.
+            audio_channels (int): input/output audio channels.
+            channels (int): initial number of hidden channels.
+            channels_time: if not None, use a different `channels` value for the time branch.
+            growth: increase the number of hidden channels by this factor at each layer.
+            nfft: number of fft bins. Note that changing this require careful computation of
+                various shape parameters and will not work out of the box for hybrid models.
+            wiener_iters: when using Wiener filtering, number of iterations at test time.
+            end_iters: same but at train time. For a hybrid model, must be equal to `wiener_iters`.
+            wiener_residual: add residual source before wiener filtering.
+            cac: uses complex as channels, i.e. complex numbers are 2 channels each
+                in input and output. no further processing is done before ISTFT.
+            depth (int): number of layers in the encoder and in the decoder.
+            rewrite (bool): add 1x1 convolution to each layer.
+            hybrid (bool): make a hybrid time/frequency domain, otherwise frequency only.
+            hybrid_old: some models trained for MDX had a padding bug. This replicates
+                this bug to avoid retraining them.
+            multi_freqs: list of frequency ratios for splitting frequency bands with `MultiWrap`.
+            multi_freqs_depth: how many layers to wrap with `MultiWrap`. Only the outermost
+                layers will be wrapped.
+            freq_emb: add frequency embedding after the first frequency layer if > 0,
+                the actual value controls the weight of the embedding.
+            emb_scale: equivalent to scaling the embedding learning rate
+            emb_smooth: initialize the embedding with a smooth one (with respect to frequencies).
+            kernel_size: kernel_size for encoder and decoder layers.
+            stride: stride for encoder and decoder layers.
+            time_stride: stride for the final time layer, after the merge.
+            context: context for 1x1 conv in the decoder.
+            context_enc: context for 1x1 conv in the encoder.
+            norm_starts: layer at which group norm starts being used.
+                decoder layers are numbered in reverse order.
+            norm_groups: number of groups for group norm.
+            dconv_mode: if 1: dconv in encoder only, 2: decoder only, 3: both.
+            dconv_depth: depth of residual DConv branch.
+            dconv_comp: compression of DConv branch.
+            dconv_attn: adds attention layers in DConv branch starting at this layer.
+            dconv_lstm: adds a LSTM layer in DConv branch starting at this layer.
+            dconv_init: initial scale for the DConv branch LayerScale.
+            rescale: weight recaling trick
+
+        """
+        super().__init__()
+        
+        self.cac = cac
+        self.wiener_residual = wiener_residual
+        self.audio_channels = audio_channels
+        self.sources = sources
+        self.kernel_size = kernel_size
+        self.context = context
+        self.stride = stride
+        self.depth = depth
+        self.channels = channels
+        self.samplerate = samplerate
+        self.segment = segment
+
+        self.nfft = nfft
+        self.hop_length = nfft // 4
+        self.wiener_iters = wiener_iters
+        self.end_iters = end_iters
+        self.freq_emb = None
+        self.hybrid = hybrid
+        self.hybrid_old = hybrid_old
+        if hybrid_old:
+            assert hybrid, "hybrid_old must come with hybrid=True"
+        if hybrid:
+            assert wiener_iters == end_iters
+
+        self.encoder = nn.ModuleList()
+        self.decoder = nn.ModuleList()
+
+        if hybrid:
+            self.tencoder = nn.ModuleList()
+            self.tdecoder = nn.ModuleList()
+
+        chin = audio_channels
+        chin_z = chin  # number of channels for the freq branch
+        if self.cac:
+            chin_z *= 2
+        chout = channels_time or channels
+        chout_z = channels
+        freqs = nfft // 2
+
+        for index in range(depth):
+            lstm = index >= dconv_lstm
+            attn = index >= dconv_attn
+            norm = index >= norm_starts
+            freq = freqs > 1
+            stri = stride
+            ker = kernel_size
+            if not freq:
+                assert freqs == 1
+                ker = time_stride * 2
+                stri = time_stride
+
+            pad = True
+            last_freq = False
+            if freq and freqs <= kernel_size:
+                ker = freqs
+                pad = False
+                last_freq = True
+
+            kw = {
+                'kernel_size': ker,
+                'stride': stri,
+                'freq': freq,
+                'pad': pad,
+                'norm': norm,
+                'rewrite': rewrite,
+                'norm_groups': norm_groups,
+                'dconv_kw': {
+                    'lstm': lstm,
+                    'attn': attn,
+                    'depth': dconv_depth,
+                    'compress': dconv_comp,
+                    'init': dconv_init,
+                    'gelu': True,
+                }
+            }
+            kwt = dict(kw)
+            kwt['freq'] = 0
+            kwt['kernel_size'] = kernel_size
+            kwt['stride'] = stride
+            kwt['pad'] = True
+            kw_dec = dict(kw)
+            multi = False
+            if multi_freqs and index < multi_freqs_depth:
+                multi = True
+                kw_dec['context_freq'] = False
+
+            if last_freq:
+                chout_z = max(chout, chout_z)
+                chout = chout_z
+
+            enc = HEncLayer(chin_z, chout_z,
+                            dconv=dconv_mode & 1, context=context_enc, **kw)
+            if hybrid and freq:
+                tenc = HEncLayer(chin, chout, dconv=dconv_mode & 1, context=context_enc,
+                                 empty=last_freq, **kwt)
+                self.tencoder.append(tenc)
+
+            if multi:
+                enc = MultiWrap(enc, multi_freqs)
+            self.encoder.append(enc)
+            if index == 0:
+                chin = self.audio_channels * len(self.sources)
+                chin_z = chin
+                if self.cac:
+                    chin_z *= 2
+            dec = HDecLayer(chout_z, chin_z, dconv=dconv_mode & 2,
+                            last=index == 0, context=context, **kw_dec)
+            if multi:
+                dec = MultiWrap(dec, multi_freqs)
+            if hybrid and freq:
+                tdec = HDecLayer(chout, chin, dconv=dconv_mode & 2, empty=last_freq,
+                                 last=index == 0, context=context, **kwt)
+                self.tdecoder.insert(0, tdec)
+            self.decoder.insert(0, dec)
+
+            chin = chout
+            chin_z = chout_z
+            chout = int(growth * chout)
+            chout_z = int(growth * chout_z)
+            if freq:
+                if freqs <= kernel_size:
+                    freqs = 1
+                else:
+                    freqs //= stride
+            if index == 0 and freq_emb:
+                self.freq_emb = ScaledEmbedding(
+                    freqs, chin_z, smooth=emb_smooth, scale=emb_scale)
+                self.freq_emb_scale = freq_emb
+
+        if rescale:
+            rescale_module(self, reference=rescale)
+
+    def _spec(self, x):
+        hl = self.hop_length
+        nfft = self.nfft
+        x0 = x  # noqa
+
+        if self.hybrid:
+            # We re-pad the signal in order to keep the property
+            # that the size of the output is exactly the size of the input
+            # divided by the stride (here hop_length), when divisible.
+            # This is achieved by padding by 1/4th of the kernel size (here nfft).
+            # which is not supported by torch.stft.
+            # Having all convolution operations follow this convention allow to easily
+            # align the time and frequency branches later on.
+            assert hl == nfft // 4
+            le = int(math.ceil(x.shape[-1] / hl))
+            pad = hl // 2 * 3
+            if not self.hybrid_old:
+                x = pad1d(x, (pad, pad + le * hl - x.shape[-1]), mode='reflect')
+            else:
+                x = pad1d(x, (pad, pad + le * hl - x.shape[-1]))
+
+        z = spectro(x, nfft, hl)[..., :-1, :]
+        if self.hybrid:
+            assert z.shape[-1] == le + 4, (z.shape, x.shape, le)
+            z = z[..., 2:2+le]
+        return z
+
+    def _ispec(self, z, length=None, scale=0):
+        hl = self.hop_length // (4 ** scale)
+        z = F.pad(z, (0, 0, 0, 1))
+        if self.hybrid:
+            z = F.pad(z, (2, 2))
+            pad = hl // 2 * 3
+            if not self.hybrid_old:
+                le = hl * int(math.ceil(length / hl)) + 2 * pad
+            else:
+                le = hl * int(math.ceil(length / hl))
+            x = ispectro(z, hl, length=le)
+            if not self.hybrid_old:
+                x = x[..., pad:pad + length]
+            else:
+                x = x[..., :length]
+        else:
+            x = ispectro(z, hl, length)
+        return x
+
+    def _magnitude(self, z):
+        # return the magnitude of the spectrogram, except when cac is True,
+        # in which case we just move the complex dimension to the channel one.
+        if self.cac:
+            B, C, Fr, T = z.shape
+            m = torch.view_as_real(z).permute(0, 1, 4, 2, 3)
+            m = m.reshape(B, C * 2, Fr, T)
+        else:
+            m = z.abs()
+        return m
+
+    def _mask(self, z, m):
+        # Apply masking given the mixture spectrogram `z` and the estimated mask `m`.
+        # If `cac` is True, `m` is actually a full spectrogram and `z` is ignored.
+        niters = self.wiener_iters
+        if self.cac:
+            B, S, C, Fr, T = m.shape
+            out = m.view(B, S, -1, 2, Fr, T).permute(0, 1, 2, 4, 5, 3)
+            out = torch.view_as_complex(out.contiguous())
+            return out
+        if self.training:
+            niters = self.end_iters
+        if niters < 0:
+            z = z[:, None]
+            return z / (1e-8 + z.abs()) * m
+        else:
+            return self._wiener(m, z, niters)
+
+    def _wiener(self, mag_out, mix_stft, niters):
+        # apply wiener filtering from OpenUnmix.
+        init = mix_stft.dtype
+        wiener_win_len = 300
+        residual = self.wiener_residual
+
+        B, S, C, Fq, T = mag_out.shape
+        mag_out = mag_out.permute(0, 4, 3, 2, 1)
+        mix_stft = torch.view_as_real(mix_stft.permute(0, 3, 2, 1))
+
+        outs = []
+        for sample in range(B):
+            pos = 0
+            out = []
+            for pos in range(0, T, wiener_win_len):
+                frame = slice(pos, pos + wiener_win_len)
+                z_out = wiener(
+                    mag_out[sample, frame], mix_stft[sample, frame], niters,
+                    residual=residual)
+                out.append(z_out.transpose(-1, -2))
+            outs.append(torch.cat(out, dim=0))
+        out = torch.view_as_complex(torch.stack(outs, 0))
+        out = out.permute(0, 4, 3, 2, 1).contiguous()
+        if residual:
+            out = out[:, :-1]
+        assert list(out.shape) == [B, S, C, Fq, T]
+        return out.to(init)
+
+    def forward(self, mix):
+        x = mix
+        length = x.shape[-1]
+
+        z = self._spec(mix)
+        mag = self._magnitude(z)
+        x = mag
+
+        B, C, Fq, T = x.shape
+
+        # unlike previous Demucs, we always normalize because it is easier.
+        mean = x.mean(dim=(1, 2, 3), keepdim=True)
+        std = x.std(dim=(1, 2, 3), keepdim=True)
+        x = (x - mean) / (1e-5 + std)
+        # x will be the freq. branch input.
+
+        if self.hybrid:
+            # Prepare the time branch input.
+            xt = mix
+            meant = xt.mean(dim=(1, 2), keepdim=True)
+            stdt = xt.std(dim=(1, 2), keepdim=True)
+            xt = (xt - meant) / (1e-5 + stdt)
+
+        # okay, this is a giant mess I know...
+        saved = []  # skip connections, freq.
+        saved_t = []  # skip connections, time.
+        lengths = []  # saved lengths to properly remove padding, freq branch.
+        lengths_t = []  # saved lengths for time branch.
+        for idx, encode in enumerate(self.encoder):
+            lengths.append(x.shape[-1])
+            inject = None
+            if self.hybrid and idx < len(self.tencoder):
+                # we have not yet merged branches.
+                lengths_t.append(xt.shape[-1])
+                tenc = self.tencoder[idx]
+                xt = tenc(xt)
+                if not tenc.empty:
+                    # save for skip connection
+                    saved_t.append(xt)
+                else:
+                    # tenc contains just the first conv., so that now time and freq.
+                    # branches have the same shape and can be merged.
+                    inject = xt
+            x = encode(x, inject)
+            if idx == 0 and self.freq_emb is not None:
+                # add frequency embedding to allow for non equivariant convolutions
+                # over the frequency axis.
+                frs = torch.arange(x.shape[-2], device=x.device)
+                emb = self.freq_emb(frs).t()[None, :, :, None].expand_as(x)
+                x = x + self.freq_emb_scale * emb
+
+            saved.append(x)
+
+        x = torch.zeros_like(x)
+        if self.hybrid:
+            xt = torch.zeros_like(x)
+        # initialize everything to zero (signal will go through u-net skips).
+
+        for idx, decode in enumerate(self.decoder):
+            skip = saved.pop(-1)
+            x, pre = decode(x, skip, lengths.pop(-1))
+            # `pre` contains the output just before final transposed convolution,
+            # which is used when the freq. and time branch separate.
+
+            if self.hybrid:
+                offset = self.depth - len(self.tdecoder)
+            if self.hybrid and idx >= offset:
+                tdec = self.tdecoder[idx - offset]
+                length_t = lengths_t.pop(-1)
+                if tdec.empty:
+                    assert pre.shape[2] == 1, pre.shape
+                    pre = pre[:, :, 0]
+                    xt, _ = tdec(pre, None, length_t)
+                else:
+                    skip = saved_t.pop(-1)
+                    xt, _ = tdec(xt, skip, length_t)
+
+        # Let's make sure we used all stored skip connections.
+        assert len(saved) == 0
+        assert len(lengths_t) == 0
+        assert len(saved_t) == 0
+
+        S = len(self.sources)
+        x = x.view(B, S, -1, Fq, T)
+        x = x * std[:, None] + mean[:, None]
+
+        zout = self._mask(z, x)
+        x = self._ispec(zout, length)
+
+        if self.hybrid:
+            xt = xt.view(B, S, -1, length)
+            xt = xt * stdt[:, None] + meant[:, None]
+            x = xt + x
+        return x
+
+

demucs/htdemucs.py

@@ -0,0 +1,648 @@
+# Copyright (c) Meta, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+# First author is Simon Rouard.
+"""
+This code contains the spectrogram and Hybrid version of Demucs.
+"""
+import math
+
+from .filtering import wiener
+import torch
+from torch import nn
+from torch.nn import functional as F
+from fractions import Fraction
+from einops import rearrange
+
+from .transformer import CrossTransformerEncoder
+
+from .demucs import rescale_module
+from .states import capture_init
+from .spec import spectro, ispectro
+from .hdemucs import pad1d, ScaledEmbedding, HEncLayer, MultiWrap, HDecLayer
+
+
+class HTDemucs(nn.Module):
+    """
+    Spectrogram and hybrid Demucs model.
+    The spectrogram model has the same structure as Demucs, except the first few layers are over the
+    frequency axis, until there is only 1 frequency, and then it moves to time convolutions.
+    Frequency layers can still access information across time steps thanks to the DConv residual.
+
+    Hybrid model have a parallel time branch. At some layer, the time branch has the same stride
+    as the frequency branch and then the two are combined. The opposite happens in the decoder.
+
+    Models can either use naive iSTFT from masking, Wiener filtering ([Ulhih et al. 2017]),
+    or complex as channels (CaC) [Choi et al. 2020]. Wiener filtering is based on
+    Open Unmix implementation [Stoter et al. 2019].
+
+    The loss is always on the temporal domain, by backpropagating through the above
+    output methods and iSTFT. This allows to define hybrid models nicely. However, this breaks
+    a bit Wiener filtering, as doing more iteration at test time will change the spectrogram
+    contribution, without changing the one from the waveform, which will lead to worse performance.
+    I tried using the residual option in OpenUnmix Wiener implementation, but it didn't improve.
+    CaC on the other hand provides similar performance for hybrid, and works naturally with
+    hybrid models.
+
+    This model also uses frequency embeddings are used to improve efficiency on convolutions
+    over the freq. axis, following [Isik et al. 2020] (https://arxiv.org/pdf/2008.04470.pdf).
+
+    Unlike classic Demucs, there is no resampling here, and normalization is always applied.
+    """
+
+    @capture_init
+    def __init__(
+        self,
+        sources,
+        # Channels
+        audio_channels=2,
+        channels=48,
+        channels_time=None,
+        growth=2,
+        # STFT
+        nfft=4096,
+        wiener_iters=0,
+        end_iters=0,
+        wiener_residual=False,
+        cac=True,
+        # Main structure
+        depth=4,
+        rewrite=True,
+        # Frequency branch
+        multi_freqs=None,
+        multi_freqs_depth=3,
+        freq_emb=0.2,
+        emb_scale=10,
+        emb_smooth=True,
+        # Convolutions
+        kernel_size=8,
+        time_stride=2,
+        stride=4,
+        context=1,
+        context_enc=0,
+        # Normalization
+        norm_starts=4,
+        norm_groups=4,
+        # DConv residual branch
+        dconv_mode=1,
+        dconv_depth=2,
+        dconv_comp=8,
+        dconv_init=1e-3,
+        # Before the Transformer
+        bottom_channels=0,
+        # Transformer
+        t_layers=5,
+        t_emb="sin",
+        t_hidden_scale=4.0,
+        t_heads=8,
+        t_dropout=0.0,
+        t_max_positions=10000,
+        t_norm_in=True,
+        t_norm_in_group=False,
+        t_group_norm=False,
+        t_norm_first=True,
+        t_norm_out=True,
+        t_max_period=10000.0,
+        t_weight_decay=0.0,
+        t_lr=None,
+        t_layer_scale=True,
+        t_gelu=True,
+        t_weight_pos_embed=1.0,
+        t_sin_random_shift=0,
+        t_cape_mean_normalize=True,
+        t_cape_augment=True,
+        t_cape_glob_loc_scale=[5000.0, 1.0, 1.4],
+        t_sparse_self_attn=False,
+        t_sparse_cross_attn=False,
+        t_mask_type="diag",
+        t_mask_random_seed=42,
+        t_sparse_attn_window=500,
+        t_global_window=100,
+        t_sparsity=0.95,
+        t_auto_sparsity=False,
+        # ------ Particuliar parameters
+        t_cross_first=False,
+        # Weight init
+        rescale=0.1,
+        # Metadata
+        samplerate=44100,
+        segment=10,
+        use_train_segment=True,
+    ):
+        """
+        Args:
+            sources (list[str]): list of source names.
+            audio_channels (int): input/output audio channels.
+            channels (int): initial number of hidden channels.
+            channels_time: if not None, use a different `channels` value for the time branch.
+            growth: increase the number of hidden channels by this factor at each layer.
+            nfft: number of fft bins. Note that changing this require careful computation of
+                various shape parameters and will not work out of the box for hybrid models.
+            wiener_iters: when using Wiener filtering, number of iterations at test time.
+            end_iters: same but at train time. For a hybrid model, must be equal to `wiener_iters`.
+            wiener_residual: add residual source before wiener filtering.
+            cac: uses complex as channels, i.e. complex numbers are 2 channels each
+                in input and output. no further processing is done before ISTFT.
+            depth (int): number of layers in the encoder and in the decoder.
+            rewrite (bool): add 1x1 convolution to each layer.
+            multi_freqs: list of frequency ratios for splitting frequency bands with `MultiWrap`.
+            multi_freqs_depth: how many layers to wrap with `MultiWrap`. Only the outermost
+                layers will be wrapped.
+            freq_emb: add frequency embedding after the first frequency layer if > 0,
+                the actual value controls the weight of the embedding.
+            emb_scale: equivalent to scaling the embedding learning rate
+            emb_smooth: initialize the embedding with a smooth one (with respect to frequencies).
+            kernel_size: kernel_size for encoder and decoder layers.
+            stride: stride for encoder and decoder layers.
+            time_stride: stride for the final time layer, after the merge.
+            context: context for 1x1 conv in the decoder.
+            context_enc: context for 1x1 conv in the encoder.
+            norm_starts: layer at which group norm starts being used.
+                decoder layers are numbered in reverse order.
+            norm_groups: number of groups for group norm.
+            dconv_mode: if 1: dconv in encoder only, 2: decoder only, 3: both.
+            dconv_depth: depth of residual DConv branch.
+            dconv_comp: compression of DConv branch.
+            dconv_attn: adds attention layers in DConv branch starting at this layer.
+            dconv_lstm: adds a LSTM layer in DConv branch starting at this layer.
+            dconv_init: initial scale for the DConv branch LayerScale.
+            bottom_channels: if >0 it adds a linear layer (1x1 Conv) before and after the
+                transformer in order to change the number of channels
+            t_layers: number of layers in each branch (waveform and spec) of the transformer
+            t_emb: "sin", "cape" or "scaled"
+            t_hidden_scale: the hidden scale of the Feedforward parts of the transformer
+                for instance if C = 384 (the number of channels in the transformer) and
+                t_hidden_scale = 4.0 then the intermediate layer of the FFN has dimension
+                384 * 4 = 1536
+            t_heads: number of heads for the transformer
+            t_dropout: dropout in the transformer
+            t_max_positions: max_positions for the "scaled" positional embedding, only
+                useful if t_emb="scaled"
+            t_norm_in: (bool) norm before addinf positional embedding and getting into the
+                transformer layers
+            t_norm_in_group: (bool) if True while t_norm_in=True, the norm is on all the
+                timesteps (GroupNorm with group=1)
+            t_group_norm: (bool) if True, the norms of the Encoder Layers are on all the
+                timesteps (GroupNorm with group=1)
+            t_norm_first: (bool) if True the norm is before the attention and before the FFN
+            t_norm_out: (bool) if True, there is a GroupNorm (group=1) at the end of each layer
+            t_max_period: (float) denominator in the sinusoidal embedding expression
+            t_weight_decay: (float) weight decay for the transformer
+            t_lr: (float) specific learning rate for the transformer
+            t_layer_scale: (bool) Layer Scale for the transformer
+            t_gelu: (bool) activations of the transformer are GeLU if True, ReLU else
+            t_weight_pos_embed: (float) weighting of the positional embedding
+            t_cape_mean_normalize: (bool) if t_emb="cape", normalisation of positional embeddings
+                see: https://arxiv.org/abs/2106.03143
+            t_cape_augment: (bool) if t_emb="cape", must be True during training and False
+                during the inference, see: https://arxiv.org/abs/2106.03143
+            t_cape_glob_loc_scale: (list of 3 floats) if t_emb="cape", CAPE parameters
+                see: https://arxiv.org/abs/2106.03143
+            t_sparse_self_attn: (bool) if True, the self attentions are sparse
+            t_sparse_cross_attn: (bool) if True, the cross-attentions are sparse (don't use it
+                unless you designed really specific masks)
+            t_mask_type: (str) can be "diag", "jmask", "random", "global" or any combination
+                with '_' between: i.e. "diag_jmask_random" (note that this is permutation
+                invariant i.e. "diag_jmask_random" is equivalent to "jmask_random_diag")
+            t_mask_random_seed: (int) if "random" is in t_mask_type, controls the seed
+                that generated the random part of the mask
+            t_sparse_attn_window: (int) if "diag" is in t_mask_type, for a query (i), and
+                a key (j), the mask is True id |i-j|<=t_sparse_attn_window
+            t_global_window: (int) if "global" is in t_mask_type, mask[:t_global_window, :]
+                and mask[:, :t_global_window] will be True
+            t_sparsity: (float) if "random" is in t_mask_type, t_sparsity is the sparsity
+                level of the random part of the mask.
+            t_cross_first: (bool) if True cross attention is the first layer of the
+                transformer (False seems to be better)
+            rescale: weight rescaling trick
+            use_train_segment: (bool) if True, the actual size that is used during the
+                training is used during inference.
+        """
+        super().__init__()
+        self.cac = cac
+        self.wiener_residual = wiener_residual
+        self.audio_channels = audio_channels
+        self.sources = sources
+        self.kernel_size = kernel_size
+        self.context = context
+        self.stride = stride
+        self.depth = depth
+        self.bottom_channels = bottom_channels
+        self.channels = channels
+        self.samplerate = samplerate
+        self.segment = segment
+        self.use_train_segment = use_train_segment
+        self.nfft = nfft
+        self.hop_length = nfft // 4
+        self.wiener_iters = wiener_iters
+        self.end_iters = end_iters
+        self.freq_emb = None
+        assert wiener_iters == end_iters
+
+        self.encoder = nn.ModuleList()
+        self.decoder = nn.ModuleList()
+
+        self.tencoder = nn.ModuleList()
+        self.tdecoder = nn.ModuleList()
+
+        chin = audio_channels
+        chin_z = chin  # number of channels for the freq branch
+        if self.cac:
+            chin_z *= 2
+        chout = channels_time or channels
+        chout_z = channels
+        freqs = nfft // 2
+
+        for index in range(depth):
+            norm = index >= norm_starts
+            freq = freqs > 1
+            stri = stride
+            ker = kernel_size
+            if not freq:
+                assert freqs == 1
+                ker = time_stride * 2
+                stri = time_stride
+
+            pad = True
+            last_freq = False
+            if freq and freqs <= kernel_size:
+                ker = freqs
+                pad = False
+                last_freq = True
+
+            kw = {
+                "kernel_size": ker,
+                "stride": stri,
+                "freq": freq,
+                "pad": pad,
+                "norm": norm,
+                "rewrite": rewrite,
+                "norm_groups": norm_groups,
+                "dconv_kw": {
+                    "depth": dconv_depth,
+                    "compress": dconv_comp,
+                    "init": dconv_init,
+                    "gelu": True,
+                },
+            }
+            kwt = dict(kw)
+            kwt["freq"] = 0
+            kwt["kernel_size"] = kernel_size
+            kwt["stride"] = stride
+            kwt["pad"] = True
+            kw_dec = dict(kw)
+            multi = False
+            if multi_freqs and index < multi_freqs_depth:
+                multi = True
+                kw_dec["context_freq"] = False
+
+            if last_freq:
+                chout_z = max(chout, chout_z)
+                chout = chout_z
+
+            enc = HEncLayer(
+                chin_z, chout_z, dconv=dconv_mode & 1, context=context_enc, **kw
+            )
+            if freq:
+                tenc = HEncLayer(
+                    chin,
+                    chout,
+                    dconv=dconv_mode & 1,
+                    context=context_enc,
+                    empty=last_freq,
+                    **kwt
+                )
+                self.tencoder.append(tenc)
+
+            if multi:
+                enc = MultiWrap(enc, multi_freqs)
+            self.encoder.append(enc)
+            if index == 0:
+                chin = self.audio_channels * len(self.sources)
+                chin_z = chin
+                if self.cac:
+                    chin_z *= 2
+            dec = HDecLayer(
+                chout_z,
+                chin_z,
+                dconv=dconv_mode & 2,
+                last=index == 0,
+                context=context,
+                **kw_dec
+            )
+            if multi:
+                dec = MultiWrap(dec, multi_freqs)
+            if freq:
+                tdec = HDecLayer(
+                    chout,
+                    chin,
+                    dconv=dconv_mode & 2,
+                    empty=last_freq,
+                    last=index == 0,
+                    context=context,
+                    **kwt
+                )
+                self.tdecoder.insert(0, tdec)
+            self.decoder.insert(0, dec)
+
+            chin = chout
+            chin_z = chout_z
+            chout = int(growth * chout)
+            chout_z = int(growth * chout_z)
+            if freq:
+                if freqs <= kernel_size:
+                    freqs = 1
+                else:
+                    freqs //= stride
+            if index == 0 and freq_emb:
+                self.freq_emb = ScaledEmbedding(
+                    freqs, chin_z, smooth=emb_smooth, scale=emb_scale
+                )
+                self.freq_emb_scale = freq_emb
+
+        if rescale:
+            rescale_module(self, reference=rescale)
+
+        transformer_channels = channels * growth ** (depth - 1)
+        if bottom_channels:
+            self.channel_upsampler = nn.Conv1d(transformer_channels, bottom_channels, 1)
+            self.channel_downsampler = nn.Conv1d(
+                bottom_channels, transformer_channels, 1
+            )
+            self.channel_upsampler_t = nn.Conv1d(
+                transformer_channels, bottom_channels, 1
+            )
+            self.channel_downsampler_t = nn.Conv1d(
+                bottom_channels, transformer_channels, 1
+            )
+
+            transformer_channels = bottom_channels
+
+        if t_layers > 0:
+            self.crosstransformer = CrossTransformerEncoder(
+                dim=transformer_channels,
+                emb=t_emb,
+                hidden_scale=t_hidden_scale,
+                num_heads=t_heads,
+                num_layers=t_layers,
+                cross_first=t_cross_first,
+                dropout=t_dropout,
+                max_positions=t_max_positions,
+                norm_in=t_norm_in,
+                norm_in_group=t_norm_in_group,
+                group_norm=t_group_norm,
+                norm_first=t_norm_first,
+                norm_out=t_norm_out,
+                max_period=t_max_period,
+                weight_decay=t_weight_decay,
+                lr=t_lr,
+                layer_scale=t_layer_scale,
+                gelu=t_gelu,
+                sin_random_shift=t_sin_random_shift,
+                weight_pos_embed=t_weight_pos_embed,
+                cape_mean_normalize=t_cape_mean_normalize,
+                cape_augment=t_cape_augment,
+                cape_glob_loc_scale=t_cape_glob_loc_scale,
+                sparse_self_attn=t_sparse_self_attn,
+                sparse_cross_attn=t_sparse_cross_attn,
+                mask_type=t_mask_type,
+                mask_random_seed=t_mask_random_seed,
+                sparse_attn_window=t_sparse_attn_window,
+                global_window=t_global_window,
+                sparsity=t_sparsity,
+                auto_sparsity=t_auto_sparsity,
+            )
+        else:
+            self.crosstransformer = None
+
+    def _spec(self, x):
+        hl = self.hop_length
+        nfft = self.nfft
+        x0 = x  # noqa
+
+        # We re-pad the signal in order to keep the property
+        # that the size of the output is exactly the size of the input
+        # divided by the stride (here hop_length), when divisible.
+        # This is achieved by padding by 1/4th of the kernel size (here nfft).
+        # which is not supported by torch.stft.
+        # Having all convolution operations follow this convention allow to easily
+        # align the time and frequency branches later on.
+        assert hl == nfft // 4
+        le = int(math.ceil(x.shape[-1] / hl))
+        pad = hl // 2 * 3
+        x = pad1d(x, (pad, pad + le * hl - x.shape[-1]), mode="reflect")
+
+        z = spectro(x, nfft, hl)[..., :-1, :]
+        assert z.shape[-1] == le + 4, (z.shape, x.shape, le)
+        z = z[..., 2: 2 + le]
+        return z
+
+    def _ispec(self, z, length=None, scale=0):
+        hl = self.hop_length // (4**scale)
+        z = F.pad(z, (0, 0, 0, 1))
+        z = F.pad(z, (2, 2))
+        pad = hl // 2 * 3
+        le = hl * int(math.ceil(length / hl)) + 2 * pad
+        x = ispectro(z, hl, length=le)
+        x = x[..., pad: pad + length]
+        return x
+
+    def _magnitude(self, z):
+        # return the magnitude of the spectrogram, except when cac is True,
+        # in which case we just move the complex dimension to the channel one.
+        if self.cac:
+            B, C, Fr, T = z.shape
+            m = torch.view_as_real(z).permute(0, 1, 4, 2, 3)
+            m = m.reshape(B, C * 2, Fr, T)
+        else:
+            m = z.abs()
+        return m
+
+    def _mask(self, z, m):
+        # Apply masking given the mixture spectrogram `z` and the estimated mask `m`.
+        # If `cac` is True, `m` is actually a full spectrogram and `z` is ignored.
+        niters = self.wiener_iters
+        if self.cac:
+            B, S, C, Fr, T = m.shape
+            out = m.view(B, S, -1, 2, Fr, T).permute(0, 1, 2, 4, 5, 3)
+            out = torch.view_as_complex(out.contiguous())
+            return out
+        if self.training:
+            niters = self.end_iters
+        if niters < 0:
+            z = z[:, None]
+            return z / (1e-8 + z.abs()) * m
+        else:
+            return self._wiener(m, z, niters)
+
+    def _wiener(self, mag_out, mix_stft, niters):
+        # apply wiener filtering from OpenUnmix.
+        init = mix_stft.dtype
+        wiener_win_len = 300
+        residual = self.wiener_residual
+
+        B, S, C, Fq, T = mag_out.shape
+        mag_out = mag_out.permute(0, 4, 3, 2, 1)
+        mix_stft = torch.view_as_real(mix_stft.permute(0, 3, 2, 1))
+
+        outs = []
+        for sample in range(B):
+            pos = 0
+            out = []
+            for pos in range(0, T, wiener_win_len):
+                frame = slice(pos, pos + wiener_win_len)
+                z_out = wiener(
+                    mag_out[sample, frame],
+                    mix_stft[sample, frame],
+                    niters,
+                    residual=residual,
+                )
+                out.append(z_out.transpose(-1, -2))
+            outs.append(torch.cat(out, dim=0))
+        out = torch.view_as_complex(torch.stack(outs, 0))
+        out = out.permute(0, 4, 3, 2, 1).contiguous()
+        if residual:
+            out = out[:, :-1]
+        assert list(out.shape) == [B, S, C, Fq, T]
+        return out.to(init)
+
+    def valid_length(self, length: int):
+        """
+        Return a length that is appropriate for evaluation.
+        In our case, always return the training length, unless
+        it is smaller than the given length, in which case this
+        raises an error.
+        """
+        if not self.use_train_segment:
+            return length
+        training_length = int(self.segment * self.samplerate)
+        if training_length < length:
+            raise ValueError(
+                    f"Given length {length} is longer than "
+                    f"training length {training_length}")
+        return training_length
+
+    def forward(self, mix):
+        length = mix.shape[-1]
+        length_pre_pad = None
+        if self.use_train_segment:
+            if self.training:
+                self.segment = Fraction(mix.shape[-1], self.samplerate)
+            else:
+                training_length = int(self.segment * self.samplerate)
+                if mix.shape[-1] < training_length:
+                    length_pre_pad = mix.shape[-1]
+                    mix = F.pad(mix, (0, training_length - length_pre_pad))
+        z = self._spec(mix)
+        mag = self._magnitude(z)
+        x = mag
+
+        B, C, Fq, T = x.shape
+
+        # unlike previous Demucs, we always normalize because it is easier.
+        mean = x.mean(dim=(1, 2, 3), keepdim=True)
+        std = x.std(dim=(1, 2, 3), keepdim=True)
+        x = (x - mean) / (1e-5 + std)
+        # x will be the freq. branch input.
+
+        # Prepare the time branch input.
+        xt = mix
+        meant = xt.mean(dim=(1, 2), keepdim=True)
+        stdt = xt.std(dim=(1, 2), keepdim=True)
+        xt = (xt - meant) / (1e-5 + stdt)
+
+        # okay, this is a giant mess I know...
+        saved = []  # skip connections, freq.
+        saved_t = []  # skip connections, time.
+        lengths = []  # saved lengths to properly remove padding, freq branch.
+        lengths_t = []  # saved lengths for time branch.
+        for idx, encode in enumerate(self.encoder):
+            lengths.append(x.shape[-1])
+            inject = None
+            if idx < len(self.tencoder):
+                # we have not yet merged branches.
+                lengths_t.append(xt.shape[-1])
+                tenc = self.tencoder[idx]
+                xt = tenc(xt)
+                if not tenc.empty:
+                    # save for skip connection
+                    saved_t.append(xt)
+                else:
+                    # tenc contains just the first conv., so that now time and freq.
+                    # branches have the same shape and can be merged.
+                    inject = xt
+            x = encode(x, inject)
+            if idx == 0 and self.freq_emb is not None:
+                # add frequency embedding to allow for non equivariant convolutions
+                # over the frequency axis.
+                frs = torch.arange(x.shape[-2], device=x.device)
+                emb = self.freq_emb(frs).t()[None, :, :, None].expand_as(x)
+                x = x + self.freq_emb_scale * emb
+
+            saved.append(x)
+        if self.crosstransformer:
+            if self.bottom_channels:
+                b, c, f, t = x.shape
+                x = rearrange(x, "b c f t-> b c (f t)")
+                x = self.channel_upsampler(x)
+                x = rearrange(x, "b c (f t)-> b c f t", f=f)
+                xt = self.channel_upsampler_t(xt)
+
+            x, xt = self.crosstransformer(x, xt)
+
+            if self.bottom_channels:
+                x = rearrange(x, "b c f t-> b c (f t)")
+                x = self.channel_downsampler(x)
+                x = rearrange(x, "b c (f t)-> b c f t", f=f)
+                xt = self.channel_downsampler_t(xt)
+
+        for idx, decode in enumerate(self.decoder):
+            skip = saved.pop(-1)
+            x, pre = decode(x, skip, lengths.pop(-1))
+            # `pre` contains the output just before final transposed convolution,
+            # which is used when the freq. and time branch separate.
+
+            offset = self.depth - len(self.tdecoder)
+            if idx >= offset:
+                tdec = self.tdecoder[idx - offset]
+                length_t = lengths_t.pop(-1)
+                if tdec.empty:
+                    assert pre.shape[2] == 1, pre.shape
+                    pre = pre[:, :, 0]
+                    xt, _ = tdec(pre, None, length_t)
+                else:
+                    skip = saved_t.pop(-1)
+                    xt, _ = tdec(xt, skip, length_t)
+
+        # Let's make sure we used all stored skip connections.
+        assert len(saved) == 0
+        assert len(lengths_t) == 0
+        assert len(saved_t) == 0
+
+        S = len(self.sources)
+        x = x.view(B, S, -1, Fq, T)
+        x = x * std[:, None] + mean[:, None]
+
+        zout = self._mask(z, x)
+        if self.use_train_segment:
+            if self.training:
+                x = self._ispec(zout, length)
+            else:
+                x = self._ispec(zout, training_length)
+        else:
+            x = self._ispec(zout, length)
+
+        if self.use_train_segment:
+            if self.training:
+                xt = xt.view(B, S, -1, length)
+            else:
+                xt = xt.view(B, S, -1, training_length)
+        else:
+            xt = xt.view(B, S, -1, length)
+        xt = xt * stdt[:, None] + meant[:, None]
+        x = xt + x
+        if length_pre_pad:
+            x = x[..., :length_pre_pad]
+        return x

demucs/model.py

@@ -0,0 +1,218 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+import math
+
+import torch as th
+from torch import nn
+
+from .utils import capture_init, center_trim
+
+
+class BLSTM(nn.Module):
+    def __init__(self, dim, layers=1):
+        super().__init__()
+        self.lstm = nn.LSTM(bidirectional=True, num_layers=layers, hidden_size=dim, input_size=dim)
+        self.linear = nn.Linear(2 * dim, dim)
+
+    def forward(self, x):
+        x = x.permute(2, 0, 1)
+        x = self.lstm(x)[0]
+        x = self.linear(x)
+        x = x.permute(1, 2, 0)
+        return x
+
+
+def rescale_conv(conv, reference):
+    std = conv.weight.std().detach()
+    scale = (std / reference)**0.5
+    conv.weight.data /= scale
+    if conv.bias is not None:
+        conv.bias.data /= scale
+
+
+def rescale_module(module, reference):
+    for sub in module.modules():
+        if isinstance(sub, (nn.Conv1d, nn.ConvTranspose1d)):
+            rescale_conv(sub, reference)
+
+
+def upsample(x, stride):
+    """
+    Linear upsampling, the output will be `stride` times longer.
+    """
+    batch, channels, time = x.size()
+    weight = th.arange(stride, device=x.device, dtype=th.float) / stride
+    x = x.view(batch, channels, time, 1)
+    out = x[..., :-1, :] * (1 - weight) + x[..., 1:, :] * weight
+    return out.reshape(batch, channels, -1)
+
+
+def downsample(x, stride):
+    """
+    Downsample x by decimation.
+    """
+    return x[:, :, ::stride]
+
+
+class Demucs(nn.Module):
+    @capture_init
+    def __init__(self,
+                 sources=4,
+                 audio_channels=2,
+                 channels=64,
+                 depth=6,
+                 rewrite=True,
+                 glu=True,
+                 upsample=False,
+                 rescale=0.1,
+                 kernel_size=8,
+                 stride=4,
+                 growth=2.,
+                 lstm_layers=2,
+                 context=3,
+                 samplerate=44100):
+        """
+        Args:
+            sources (int): number of sources to separate
+            audio_channels (int): stereo or mono
+            channels (int): first convolution channels
+            depth (int): number of encoder/decoder layers
+            rewrite (bool): add 1x1 convolution to each encoder layer
+                and a convolution to each decoder layer.
+                For the decoder layer, `context` gives the kernel size.
+            glu (bool): use glu instead of ReLU
+            upsample (bool): use linear upsampling with convolutions
+                Wave-U-Net style, instead of transposed convolutions
+            rescale (int): rescale initial weights of convolutions
+                to get their standard deviation closer to `rescale`
+            kernel_size (int): kernel size for convolutions
+            stride (int): stride for convolutions
+            growth (float): multiply (resp divide) number of channels by that
+                for each layer of the encoder (resp decoder)
+            lstm_layers (int): number of lstm layers, 0 = no lstm
+            context (int): kernel size of the convolution in the
+                decoder before the transposed convolution. If > 1,
+                will provide some context from neighboring time
+                steps.
+        """
+
+        super().__init__()
+        self.audio_channels = audio_channels
+        self.sources = sources
+        self.kernel_size = kernel_size
+        self.context = context
+        self.stride = stride
+        self.depth = depth
+        self.upsample = upsample
+        self.channels = channels
+        self.samplerate = samplerate
+
+        self.encoder = nn.ModuleList()
+        self.decoder = nn.ModuleList()
+
+        self.final = None
+        if upsample:
+            self.final = nn.Conv1d(channels + audio_channels, sources * audio_channels, 1)
+            stride = 1
+
+        if glu:
+            activation = nn.GLU(dim=1)
+            ch_scale = 2
+        else:
+            activation = nn.ReLU()
+            ch_scale = 1
+        in_channels = audio_channels
+        for index in range(depth):
+            encode = []
+            encode += [nn.Conv1d(in_channels, channels, kernel_size, stride), nn.ReLU()]
+            if rewrite:
+                encode += [nn.Conv1d(channels, ch_scale * channels, 1), activation]
+            self.encoder.append(nn.Sequential(*encode))
+
+            decode = []
+            if index > 0:
+                out_channels = in_channels
+            else:
+                if upsample:
+                    out_channels = channels
+                else:
+                    out_channels = sources * audio_channels
+            if rewrite:
+                decode += [nn.Conv1d(channels, ch_scale * channels, context), activation]
+            if upsample:
+                decode += [
+                    nn.Conv1d(channels, out_channels, kernel_size, stride=1),
+                ]
+            else:
+                decode += [nn.ConvTranspose1d(channels, out_channels, kernel_size, stride)]
+            if index > 0:
+                decode.append(nn.ReLU())
+            self.decoder.insert(0, nn.Sequential(*decode))
+            in_channels = channels
+            channels = int(growth * channels)
+
+        channels = in_channels
+
+        if lstm_layers:
+            self.lstm = BLSTM(channels, lstm_layers)
+        else:
+            self.lstm = None
+
+        if rescale:
+            rescale_module(self, reference=rescale)
+
+    def valid_length(self, length):
+        """
+        Return the nearest valid length to use with the model so that
+        there is no time steps left over in a convolutions, e.g. for all
+        layers, size of the input - kernel_size % stride = 0.
+
+        If the mixture has a valid length, the estimated sources
+        will have exactly the same length when context = 1. If context > 1,
+        the two signals can be center trimmed to match.
+
+        For training, extracts should have a valid length.For evaluation
+        on full tracks we recommend passing `pad = True` to :method:`forward`.
+        """
+        for _ in range(self.depth):
+            if self.upsample:
+                length = math.ceil(length / self.stride) + self.kernel_size - 1
+            else:
+                length = math.ceil((length - self.kernel_size) / self.stride) + 1
+            length = max(1, length)
+            length += self.context - 1
+        for _ in range(self.depth):
+            if self.upsample:
+                length = length * self.stride + self.kernel_size - 1
+            else:
+                length = (length - 1) * self.stride + self.kernel_size
+
+        return int(length)
+
+    def forward(self, mix):
+        x = mix
+        saved = [x]
+        for encode in self.encoder:
+            x = encode(x)
+            saved.append(x)
+            if self.upsample:
+                x = downsample(x, self.stride)
+        if self.lstm:
+            x = self.lstm(x)
+        for decode in self.decoder:
+            if self.upsample:
+                x = upsample(x, stride=self.stride)
+            skip = center_trim(saved.pop(-1), x)
+            x = x + skip
+            x = decode(x)
+        if self.final:
+            skip = center_trim(saved.pop(-1), x)
+            x = th.cat([x, skip], dim=1)
+            x = self.final(x)
+
+        x = x.view(x.size(0), self.sources, self.audio_channels, x.size(-1))
+        return x

demucs/model_v2.py

@@ -0,0 +1,218 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+import math
+
+import julius
+from torch import nn
+from .tasnet_v2 import ConvTasNet
+
+from .utils import capture_init, center_trim
+
+
+class BLSTM(nn.Module):
+    def __init__(self, dim, layers=1):
+        super().__init__()
+        self.lstm = nn.LSTM(bidirectional=True, num_layers=layers, hidden_size=dim, input_size=dim)
+        self.linear = nn.Linear(2 * dim, dim)
+
+    def forward(self, x):
+        x = x.permute(2, 0, 1)
+        x = self.lstm(x)[0]
+        x = self.linear(x)
+        x = x.permute(1, 2, 0)
+        return x
+
+
+def rescale_conv(conv, reference):
+    std = conv.weight.std().detach()
+    scale = (std / reference)**0.5
+    conv.weight.data /= scale
+    if conv.bias is not None:
+        conv.bias.data /= scale
+
+
+def rescale_module(module, reference):
+    for sub in module.modules():
+        if isinstance(sub, (nn.Conv1d, nn.ConvTranspose1d)):
+            rescale_conv(sub, reference)
+
+def auto_load_demucs_model_v2(sources, demucs_model_name):
+    
+    if '48' in demucs_model_name:
+        channels=48
+    elif 'unittest' in demucs_model_name:
+        channels=4
+    else:
+        channels=64
+    
+    if 'tasnet' in demucs_model_name:
+        init_demucs_model = ConvTasNet(sources, X=10)
+    else:
+        init_demucs_model = Demucs(sources, channels=channels)
+        
+    return init_demucs_model
+
+class Demucs(nn.Module):
+    @capture_init
+    def __init__(self,
+                 sources,
+                 audio_channels=2,
+                 channels=64,
+                 depth=6,
+                 rewrite=True,
+                 glu=True,
+                 rescale=0.1,
+                 resample=True,
+                 kernel_size=8,
+                 stride=4,
+                 growth=2.,
+                 lstm_layers=2,
+                 context=3,
+                 normalize=False,
+                 samplerate=44100,
+                 segment_length=4 * 10 * 44100):
+        """
+        Args:
+            sources (list[str]): list of source names
+            audio_channels (int): stereo or mono
+            channels (int): first convolution channels
+            depth (int): number of encoder/decoder layers
+            rewrite (bool): add 1x1 convolution to each encoder layer
+                and a convolution to each decoder layer.
+                For the decoder layer, `context` gives the kernel size.
+            glu (bool): use glu instead of ReLU
+            resample_input (bool): upsample x2 the input and downsample /2 the output.
+            rescale (int): rescale initial weights of convolutions
+                to get their standard deviation closer to `rescale`
+            kernel_size (int): kernel size for convolutions
+            stride (int): stride for convolutions
+            growth (float): multiply (resp divide) number of channels by that
+                for each layer of the encoder (resp decoder)
+            lstm_layers (int): number of lstm layers, 0 = no lstm
+            context (int): kernel size of the convolution in the
+                decoder before the transposed convolution. If > 1,
+                will provide some context from neighboring time
+                steps.
+            samplerate (int): stored as meta information for easing
+                future evaluations of the model.
+            segment_length (int): stored as meta information for easing
+                future evaluations of the model. Length of the segments on which
+                the model was trained.
+        """
+
+        super().__init__()
+        self.audio_channels = audio_channels
+        self.sources = sources
+        self.kernel_size = kernel_size
+        self.context = context
+        self.stride = stride
+        self.depth = depth
+        self.resample = resample
+        self.channels = channels
+        self.normalize = normalize
+        self.samplerate = samplerate
+        self.segment_length = segment_length
+
+        self.encoder = nn.ModuleList()
+        self.decoder = nn.ModuleList()
+
+        if glu:
+            activation = nn.GLU(dim=1)
+            ch_scale = 2
+        else:
+            activation = nn.ReLU()
+            ch_scale = 1
+        in_channels = audio_channels
+        for index in range(depth):
+            encode = []
+            encode += [nn.Conv1d(in_channels, channels, kernel_size, stride), nn.ReLU()]
+            if rewrite:
+                encode += [nn.Conv1d(channels, ch_scale * channels, 1), activation]
+            self.encoder.append(nn.Sequential(*encode))
+
+            decode = []
+            if index > 0:
+                out_channels = in_channels
+            else:
+                out_channels = len(self.sources) * audio_channels
+            if rewrite:
+                decode += [nn.Conv1d(channels, ch_scale * channels, context), activation]
+            decode += [nn.ConvTranspose1d(channels, out_channels, kernel_size, stride)]
+            if index > 0:
+                decode.append(nn.ReLU())
+            self.decoder.insert(0, nn.Sequential(*decode))
+            in_channels = channels
+            channels = int(growth * channels)
+
+        channels = in_channels
+
+        if lstm_layers:
+            self.lstm = BLSTM(channels, lstm_layers)
+        else:
+            self.lstm = None
+
+        if rescale:
+            rescale_module(self, reference=rescale)
+
+    def valid_length(self, length):
+        """
+        Return the nearest valid length to use with the model so that
+        there is no time steps left over in a convolutions, e.g. for all
+        layers, size of the input - kernel_size % stride = 0.
+
+        If the mixture has a valid length, the estimated sources
+        will have exactly the same length when context = 1. If context > 1,
+        the two signals can be center trimmed to match.
+
+        For training, extracts should have a valid length.For evaluation
+        on full tracks we recommend passing `pad = True` to :method:`forward`.
+        """
+        if self.resample:
+            length *= 2
+        for _ in range(self.depth):
+            length = math.ceil((length - self.kernel_size) / self.stride) + 1
+            length = max(1, length)
+            length += self.context - 1
+        for _ in range(self.depth):
+            length = (length - 1) * self.stride + self.kernel_size
+
+        if self.resample:
+            length = math.ceil(length / 2)
+        return int(length)
+
+    def forward(self, mix):
+        x = mix
+
+        if self.normalize:
+            mono = mix.mean(dim=1, keepdim=True)
+            mean = mono.mean(dim=-1, keepdim=True)
+            std = mono.std(dim=-1, keepdim=True)
+        else:
+            mean = 0
+            std = 1
+
+        x = (x - mean) / (1e-5 + std)
+
+        if self.resample:
+            x = julius.resample_frac(x, 1, 2)
+
+        saved = []
+        for encode in self.encoder:
+            x = encode(x)
+            saved.append(x)
+        if self.lstm:
+            x = self.lstm(x)
+        for decode in self.decoder:
+            skip = center_trim(saved.pop(-1), x)
+            x = x + skip
+            x = decode(x)
+
+        if self.resample:
+            x = julius.resample_frac(x, 2, 1)
+        x = x * std + mean
+        x = x.view(x.size(0), len(self.sources), self.audio_channels, x.size(-1))
+        return x

demucs/pretrained.py

@@ -0,0 +1,180 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+"""Loading pretrained models.
+"""
+
+import logging
+from pathlib import Path
+import typing as tp
+
+from dora.log import fatal
+
+import logging
+
+from diffq import DiffQuantizer
+import torch.hub
+
+from .model import Demucs
+from .tasnet_v2 import ConvTasNet
+from .utils import set_state
+
+from .hdemucs import HDemucs
+from .repo import RemoteRepo, LocalRepo, ModelOnlyRepo, BagOnlyRepo, AnyModelRepo, ModelLoadingError  # noqa
+
+logger = logging.getLogger(__name__)
+ROOT_URL = "https://dl.fbaipublicfiles.com/demucs/mdx_final/"
+REMOTE_ROOT = Path(__file__).parent / 'remote'
+
+SOURCES = ["drums", "bass", "other", "vocals"]
+
+
+def demucs_unittest():
+    model = HDemucs(channels=4, sources=SOURCES)
+    return model
+
+
+def add_model_flags(parser):
+    group = parser.add_mutually_exclusive_group(required=False)
+    group.add_argument("-s", "--sig", help="Locally trained XP signature.")
+    group.add_argument("-n", "--name", default="mdx_extra_q",
+                       help="Pretrained model name or signature. Default is mdx_extra_q.")
+    parser.add_argument("--repo", type=Path,
+                        help="Folder containing all pre-trained models for use with -n.")
+
+
+def _parse_remote_files(remote_file_list) -> tp.Dict[str, str]:
+    root: str = ''
+    models: tp.Dict[str, str] = {}
+    for line in remote_file_list.read_text().split('\n'):
+        line = line.strip()
+        if line.startswith('#'):
+            continue
+        elif line.startswith('root:'):
+            root = line.split(':', 1)[1].strip()
+        else:
+            sig = line.split('-', 1)[0]
+            assert sig not in models
+            models[sig] = ROOT_URL + root + line
+    return models
+
+def get_model(name: str,
+              repo: tp.Optional[Path] = None):
+    """`name` must be a bag of models name or a pretrained signature
+    from the remote AWS model repo or the specified local repo if `repo` is not None.
+    """
+    if name == 'demucs_unittest':
+        return demucs_unittest()
+    model_repo: ModelOnlyRepo
+    if repo is None:
+        models = _parse_remote_files(REMOTE_ROOT / 'files.txt')
+        model_repo = RemoteRepo(models)
+        bag_repo = BagOnlyRepo(REMOTE_ROOT, model_repo)
+    else:
+        if not repo.is_dir():
+            fatal(f"{repo} must exist and be a directory.")
+        model_repo = LocalRepo(repo)
+        bag_repo = BagOnlyRepo(repo, model_repo)
+    any_repo = AnyModelRepo(model_repo, bag_repo)
+    model = any_repo.get_model(name)
+    model.eval()
+    return model
+
+def get_model_from_args(args):
+    """
+    Load local model package or pre-trained model.
+    """
+    return get_model(name=args.name, repo=args.repo)
+
+logger = logging.getLogger(__name__)
+ROOT = "https://dl.fbaipublicfiles.com/demucs/v3.0/"
+
+PRETRAINED_MODELS = {
+    'demucs': 'e07c671f',
+    'demucs48_hq': '28a1282c',
+    'demucs_extra': '3646af93',
+    'demucs_quantized': '07afea75',
+    'tasnet': 'beb46fac',
+    'tasnet_extra': 'df3777b2',
+    'demucs_unittest': '09ebc15f',
+}
+
+SOURCES = ["drums", "bass", "other", "vocals"]
+
+
+def get_url(name):
+    sig = PRETRAINED_MODELS[name]
+    return ROOT + name + "-" + sig[:8] + ".th"
+
+def is_pretrained(name):
+    return name in PRETRAINED_MODELS
+
+
+def load_pretrained(name):
+    if name == "demucs":
+        return demucs(pretrained=True)
+    elif name == "demucs48_hq":
+        return demucs(pretrained=True, hq=True, channels=48)
+    elif name == "demucs_extra":
+        return demucs(pretrained=True, extra=True)
+    elif name == "demucs_quantized":
+        return demucs(pretrained=True, quantized=True)
+    elif name == "demucs_unittest":
+        return demucs_unittest(pretrained=True)
+    elif name == "tasnet":
+        return tasnet(pretrained=True)
+    elif name == "tasnet_extra":
+        return tasnet(pretrained=True, extra=True)
+    else:
+        raise ValueError(f"Invalid pretrained name {name}")
+
+
+def _load_state(name, model, quantizer=None):
+    url = get_url(name)
+    state = torch.hub.load_state_dict_from_url(url, map_location='cpu', check_hash=True)
+    set_state(model, quantizer, state)
+    if quantizer:
+        quantizer.detach()
+
+
+def demucs_unittest(pretrained=True):
+    model = Demucs(channels=4, sources=SOURCES)
+    if pretrained:
+        _load_state('demucs_unittest', model)
+    return model
+
+
+def demucs(pretrained=True, extra=False, quantized=False, hq=False, channels=64):
+    if not pretrained and (extra or quantized or hq):
+        raise ValueError("if extra or quantized is True, pretrained must be True.")
+    model = Demucs(sources=SOURCES, channels=channels)
+    if pretrained:
+        name = 'demucs'
+        if channels != 64:
+            name += str(channels)
+        quantizer = None
+        if sum([extra, quantized, hq]) > 1:
+            raise ValueError("Only one of extra, quantized, hq, can be True.")
+        if quantized:
+            quantizer = DiffQuantizer(model, group_size=8, min_size=1)
+            name += '_quantized'
+        if extra:
+            name += '_extra'
+        if hq:
+            name += '_hq'
+        _load_state(name, model, quantizer)
+    return model
+
+
+def tasnet(pretrained=True, extra=False):
+    if not pretrained and extra:
+        raise ValueError("if extra is True, pretrained must be True.")
+    model = ConvTasNet(X=10, sources=SOURCES)
+    if pretrained:
+        name = 'tasnet'
+        if extra:
+            name = 'tasnet_extra'
+        _load_state(name, model)
+    return model

demucs/repo.py

@@ -0,0 +1,148 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+"""Represents a model repository, including pre-trained models and bags of models.
+A repo can either be the main remote repository stored in AWS, or a local repository
+with your own models.
+"""
+
+from hashlib import sha256
+from pathlib import Path
+import typing as tp
+
+import torch
+import yaml
+
+from .apply import BagOfModels, Model
+from .states import load_model
+
+
+AnyModel = tp.Union[Model, BagOfModels]
+
+
+class ModelLoadingError(RuntimeError):
+    pass
+
+
+def check_checksum(path: Path, checksum: str):
+    sha = sha256()
+    with open(path, 'rb') as file:
+        while True:
+            buf = file.read(2**20)
+            if not buf:
+                break
+            sha.update(buf)
+    actual_checksum = sha.hexdigest()[:len(checksum)]
+    if actual_checksum != checksum:
+        raise ModelLoadingError(f'Invalid checksum for file {path}, '
+                                f'expected {checksum} but got {actual_checksum}')
+
+class ModelOnlyRepo:
+    """Base class for all model only repos.
+    """
+    def has_model(self, sig: str) -> bool:
+        raise NotImplementedError()
+
+    def get_model(self, sig: str) -> Model:
+        raise NotImplementedError()
+
+
+class RemoteRepo(ModelOnlyRepo):
+    def __init__(self, models: tp.Dict[str, str]):
+        self._models = models
+
+    def has_model(self, sig: str) -> bool:
+        return sig in self._models
+
+    def get_model(self, sig: str) -> Model:
+        try:
+            url = self._models[sig]
+        except KeyError:
+            raise ModelLoadingError(f'Could not find a pre-trained model with signature {sig}.')
+        pkg = torch.hub.load_state_dict_from_url(url, map_location='cpu', check_hash=True)
+        return load_model(pkg)
+
+
+class LocalRepo(ModelOnlyRepo):
+    def __init__(self, root: Path):
+        self.root = root
+        self.scan()
+
+    def scan(self):
+        self._models = {}
+        self._checksums = {}
+        for file in self.root.iterdir():
+            if file.suffix == '.th':
+                if '-' in file.stem:
+                    xp_sig, checksum = file.stem.split('-')
+                    self._checksums[xp_sig] = checksum
+                else:
+                    xp_sig = file.stem
+                if xp_sig in self._models:
+                    print('Whats xp? ', xp_sig)
+                    raise ModelLoadingError(
+                        f'Duplicate pre-trained model exist for signature {xp_sig}. '
+                        'Please delete all but one.')
+                self._models[xp_sig] = file
+
+    def has_model(self, sig: str) -> bool:
+        return sig in self._models
+
+    def get_model(self, sig: str) -> Model:
+        try:
+            file = self._models[sig]
+        except KeyError:
+            raise ModelLoadingError(f'Could not find pre-trained model with signature {sig}.')
+        if sig in self._checksums:
+            check_checksum(file, self._checksums[sig])
+        return load_model(file)
+
+
+class BagOnlyRepo:
+    """Handles only YAML files containing bag of models, leaving the actual
+    model loading to some Repo.
+    """
+    def __init__(self, root: Path, model_repo: ModelOnlyRepo):
+        self.root = root
+        self.model_repo = model_repo
+        self.scan()
+
+    def scan(self):
+        self._bags = {}
+        for file in self.root.iterdir():
+            if file.suffix == '.yaml':
+                self._bags[file.stem] = file
+
+    def has_model(self, name: str) -> bool:
+        return name in self._bags
+
+    def get_model(self, name: str) -> BagOfModels:
+        try:
+            yaml_file = self._bags[name]
+        except KeyError:
+            raise ModelLoadingError(f'{name} is neither a single pre-trained model or '
+                                    'a bag of models.')
+        bag = yaml.safe_load(open(yaml_file))
+        signatures = bag['models']
+        models = [self.model_repo.get_model(sig) for sig in signatures]
+        weights = bag.get('weights')
+        segment = bag.get('segment')
+        return BagOfModels(models, weights, segment)
+
+
+class AnyModelRepo:
+    def __init__(self, model_repo: ModelOnlyRepo, bag_repo: BagOnlyRepo):
+        self.model_repo = model_repo
+        self.bag_repo = bag_repo
+
+    def has_model(self, name_or_sig: str) -> bool:
+        return self.model_repo.has_model(name_or_sig) or self.bag_repo.has_model(name_or_sig)
+
+    def get_model(self, name_or_sig: str) -> AnyModel:
+        print('name_or_sig: ', name_or_sig)
+        if self.model_repo.has_model(name_or_sig):
+            return self.model_repo.get_model(name_or_sig)
+        else:
+            return self.bag_repo.get_model(name_or_sig)

demucs/spec.py

@@ -0,0 +1,41 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+"""Conveniance wrapper to perform STFT and iSTFT"""
+
+import torch as th
+
+
+def spectro(x, n_fft=512, hop_length=None, pad=0):
+    *other, length = x.shape
+    x = x.reshape(-1, length)
+    z = th.stft(x,
+                n_fft * (1 + pad),
+                hop_length or n_fft // 4,
+                window=th.hann_window(n_fft).to(x),
+                win_length=n_fft,
+                normalized=True,
+                center=True,
+                return_complex=True,
+                pad_mode='reflect')
+    _, freqs, frame = z.shape
+    return z.view(*other, freqs, frame)
+
+
+def ispectro(z, hop_length=None, length=None, pad=0):
+    *other, freqs, frames = z.shape
+    n_fft = 2 * freqs - 2
+    z = z.view(-1, freqs, frames)
+    win_length = n_fft // (1 + pad)
+    x = th.istft(z,
+                 n_fft,
+                 hop_length,
+                 window=th.hann_window(win_length).to(z.real),
+                 win_length=win_length,
+                 normalized=True,
+                 length=length,
+                 center=True)
+    _, length = x.shape
+    return x.view(*other, length)

demucs/states.py

@@ -0,0 +1,148 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+"""
+Utilities to save and load models.
+"""
+from contextlib import contextmanager
+
+import functools
+import hashlib
+import inspect
+import io
+from pathlib import Path
+import warnings
+
+from omegaconf import OmegaConf
+from diffq import DiffQuantizer, UniformQuantizer, restore_quantized_state
+import torch
+
+
+def get_quantizer(model, args, optimizer=None):
+    """Return the quantizer given the XP quantization args."""
+    quantizer = None
+    if args.diffq:
+        quantizer = DiffQuantizer(
+            model, min_size=args.min_size, group_size=args.group_size)
+        if optimizer is not None:
+            quantizer.setup_optimizer(optimizer)
+    elif args.qat:
+        quantizer = UniformQuantizer(
+                model, bits=args.qat, min_size=args.min_size)
+    return quantizer
+
+
+def load_model(path_or_package, strict=False):
+    """Load a model from the given serialized model, either given as a dict (already loaded)
+    or a path to a file on disk."""
+    if isinstance(path_or_package, dict):
+        package = path_or_package
+    elif isinstance(path_or_package, (str, Path)):
+        with warnings.catch_warnings():
+            warnings.simplefilter("ignore")
+            path = path_or_package
+            package = torch.load(path, 'cpu')
+    else:
+        raise ValueError(f"Invalid type for {path_or_package}.")
+
+    klass = package["klass"]
+    args = package["args"]
+    kwargs = package["kwargs"]
+
+    if strict:
+        model = klass(*args, **kwargs)
+    else:
+        sig = inspect.signature(klass)
+        for key in list(kwargs):
+            if key not in sig.parameters:
+                warnings.warn("Dropping inexistant parameter " + key)
+                del kwargs[key]
+        model = klass(*args, **kwargs)
+
+    state = package["state"]
+
+    set_state(model, state)
+    return model
+
+
+def get_state(model, quantizer, half=False):
+    """Get the state from a model, potentially with quantization applied.
+    If `half` is True, model are stored as half precision, which shouldn't impact performance
+    but half the state size."""
+    if quantizer is None:
+        dtype = torch.half if half else None
+        state = {k: p.data.to(device='cpu', dtype=dtype) for k, p in model.state_dict().items()}
+    else:
+        state = quantizer.get_quantized_state()
+        state['__quantized'] = True
+    return state
+
+
+def set_state(model, state, quantizer=None):
+    """Set the state on a given model."""
+    if state.get('__quantized'):
+        if quantizer is not None:
+            quantizer.restore_quantized_state(model, state['quantized'])
+        else:
+            restore_quantized_state(model, state)
+    else:
+        model.load_state_dict(state)
+    return state
+
+
+def save_with_checksum(content, path):
+    """Save the given value on disk, along with a sha256 hash.
+    Should be used with the output of either `serialize_model` or `get_state`."""
+    buf = io.BytesIO()
+    torch.save(content, buf)
+    sig = hashlib.sha256(buf.getvalue()).hexdigest()[:8]
+
+    path = path.parent / (path.stem + "-" + sig + path.suffix)
+    path.write_bytes(buf.getvalue())
+
+
+def serialize_model(model, training_args, quantizer=None, half=True):
+    args, kwargs = model._init_args_kwargs
+    klass = model.__class__
+
+    state = get_state(model, quantizer, half)
+    return {
+        'klass': klass,
+        'args': args,
+        'kwargs': kwargs,
+        'state': state,
+        'training_args': OmegaConf.to_container(training_args, resolve=True),
+    }
+
+
+def copy_state(state):
+    return {k: v.cpu().clone() for k, v in state.items()}
+
+
+@contextmanager
+def swap_state(model, state):
+    """
+    Context manager that swaps the state of a model, e.g:
+
+        # model is in old state
+        with swap_state(model, new_state):
+            # model in new state
+        # model back to old state
+    """
+    old_state = copy_state(model.state_dict())
+    model.load_state_dict(state, strict=False)
+    try:
+        yield
+    finally:
+        model.load_state_dict(old_state)
+
+
+def capture_init(init):
+    @functools.wraps(init)
+    def __init__(self, *args, **kwargs):
+        self._init_args_kwargs = (args, kwargs)
+        init(self, *args, **kwargs)
+
+    return __init__

demucs/tasnet.py

@@ -0,0 +1,447 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+#
+# Created on 2018/12
+# Author: Kaituo XU
+# Modified on 2019/11 by Alexandre Defossez, added support for multiple output channels
+# Here is the original license:
+# The MIT License (MIT)
+#
+# Copyright (c) 2018 Kaituo XU
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to deal
+# in the Software without restriction, including without limitation the rights
+# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+# copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in all
+# copies or substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+# SOFTWARE.
+
+import math
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from .utils import capture_init
+
+EPS = 1e-8
+
+
+def overlap_and_add(signal, frame_step):
+    outer_dimensions = signal.size()[:-2]
+    frames, frame_length = signal.size()[-2:]
+
+    subframe_length = math.gcd(frame_length, frame_step)  # gcd=Greatest Common Divisor
+    subframe_step = frame_step // subframe_length
+    subframes_per_frame = frame_length // subframe_length
+    output_size = frame_step * (frames - 1) + frame_length
+    output_subframes = output_size // subframe_length
+
+    subframe_signal = signal.view(*outer_dimensions, -1, subframe_length)
+
+    frame = torch.arange(0, output_subframes,
+                         device=signal.device).unfold(0, subframes_per_frame, subframe_step)
+    frame = frame.long()  # signal may in GPU or CPU
+    frame = frame.contiguous().view(-1)
+
+    result = signal.new_zeros(*outer_dimensions, output_subframes, subframe_length)
+    result.index_add_(-2, frame, subframe_signal)
+    result = result.view(*outer_dimensions, -1)
+    return result
+
+
+class ConvTasNet(nn.Module):
+    @capture_init
+    def __init__(self,
+                 N=256,
+                 L=20,
+                 B=256,
+                 H=512,
+                 P=3,
+                 X=8,
+                 R=4,
+                 C=4,
+                 audio_channels=1,
+                 samplerate=44100,
+                 norm_type="gLN",
+                 causal=False,
+                 mask_nonlinear='relu'):
+        """
+        Args:
+            N: Number of filters in autoencoder
+            L: Length of the filters (in samples)
+            B: Number of channels in bottleneck 1 × 1-conv block
+            H: Number of channels in convolutional blocks
+            P: Kernel size in convolutional blocks
+            X: Number of convolutional blocks in each repeat
+            R: Number of repeats
+            C: Number of speakers
+            norm_type: BN, gLN, cLN
+            causal: causal or non-causal
+            mask_nonlinear: use which non-linear function to generate mask
+        """
+        super(ConvTasNet, self).__init__()
+        # Hyper-parameter
+        self.N, self.L, self.B, self.H, self.P, self.X, self.R, self.C = N, L, B, H, P, X, R, C
+        self.norm_type = norm_type
+        self.causal = causal
+        self.mask_nonlinear = mask_nonlinear
+        self.audio_channels = audio_channels
+        self.samplerate = samplerate
+        # Components
+        self.encoder = Encoder(L, N, audio_channels)
+        self.separator = TemporalConvNet(N, B, H, P, X, R, C, norm_type, causal, mask_nonlinear)
+        self.decoder = Decoder(N, L, audio_channels)
+        # init
+        for p in self.parameters():
+            if p.dim() > 1:
+                nn.init.xavier_normal_(p)
+
+    def valid_length(self, length):
+        return length
+
+    def forward(self, mixture):
+        """
+        Args:
+            mixture: [M, T], M is batch size, T is #samples
+        Returns:
+            est_source: [M, C, T]
+        """
+        mixture_w = self.encoder(mixture)
+        est_mask = self.separator(mixture_w)
+        est_source = self.decoder(mixture_w, est_mask)
+
+        # T changed after conv1d in encoder, fix it here
+        T_origin = mixture.size(-1)
+        T_conv = est_source.size(-1)
+        est_source = F.pad(est_source, (0, T_origin - T_conv))
+        return est_source
+
+
+class Encoder(nn.Module):
+    """Estimation of the nonnegative mixture weight by a 1-D conv layer.
+    """
+    def __init__(self, L, N, audio_channels):
+        super(Encoder, self).__init__()
+        # Hyper-parameter
+        self.L, self.N = L, N
+        # Components
+        # 50% overlap
+        self.conv1d_U = nn.Conv1d(audio_channels, N, kernel_size=L, stride=L // 2, bias=False)
+
+    def forward(self, mixture):
+        """
+        Args:
+            mixture: [M, T], M is batch size, T is #samples
+        Returns:
+            mixture_w: [M, N, K], where K = (T-L)/(L/2)+1 = 2T/L-1
+        """
+        mixture_w = F.relu(self.conv1d_U(mixture))  # [M, N, K]
+        return mixture_w
+
+
+class Decoder(nn.Module):
+    def __init__(self, N, L, audio_channels):
+        super(Decoder, self).__init__()
+        # Hyper-parameter
+        self.N, self.L = N, L
+        self.audio_channels = audio_channels
+        # Components
+        self.basis_signals = nn.Linear(N, audio_channels * L, bias=False)
+
+    def forward(self, mixture_w, est_mask):
+        """
+        Args:
+            mixture_w: [M, N, K]
+            est_mask: [M, C, N, K]
+        Returns:
+            est_source: [M, C, T]
+        """
+        # D = W * M
+        source_w = torch.unsqueeze(mixture_w, 1) * est_mask  # [M, C, N, K]
+        source_w = torch.transpose(source_w, 2, 3)  # [M, C, K, N]
+        # S = DV
+        est_source = self.basis_signals(source_w)  # [M, C, K, ac * L]
+        m, c, k, _ = est_source.size()
+        est_source = est_source.view(m, c, k, self.audio_channels, -1).transpose(2, 3).contiguous()
+        est_source = overlap_and_add(est_source, self.L // 2)  # M x C x ac x T
+        return est_source
+
+
+class TemporalConvNet(nn.Module):
+    def __init__(self, N, B, H, P, X, R, C, norm_type="gLN", causal=False, mask_nonlinear='relu'):
+        """
+        Args:
+            N: Number of filters in autoencoder
+            B: Number of channels in bottleneck 1 × 1-conv block
+            H: Number of channels in convolutional blocks
+            P: Kernel size in convolutional blocks
+            X: Number of convolutional blocks in each repeat
+            R: Number of repeats
+            C: Number of speakers
+            norm_type: BN, gLN, cLN
+            causal: causal or non-causal
+            mask_nonlinear: use which non-linear function to generate mask
+        """
+        super(TemporalConvNet, self).__init__()
+        # Hyper-parameter
+        self.C = C
+        self.mask_nonlinear = mask_nonlinear
+        # Components
+        # [M, N, K] -> [M, N, K]
+        layer_norm = ChannelwiseLayerNorm(N)
+        # [M, N, K] -> [M, B, K]
+        bottleneck_conv1x1 = nn.Conv1d(N, B, 1, bias=False)
+        # [M, B, K] -> [M, B, K]
+        repeats = []
+        for r in range(R):
+            blocks = []
+            for x in range(X):
+                dilation = 2**x
+                padding = (P - 1) * dilation if causal else (P - 1) * dilation // 2
+                blocks += [
+                    TemporalBlock(B,
+                                  H,
+                                  P,
+                                  stride=1,
+                                  padding=padding,
+                                  dilation=dilation,
+                                  norm_type=norm_type,
+                                  causal=causal)
+                ]
+            repeats += [nn.Sequential(*blocks)]
+        temporal_conv_net = nn.Sequential(*repeats)
+        # [M, B, K] -> [M, C*N, K]
+        mask_conv1x1 = nn.Conv1d(B, C * N, 1, bias=False)
+        # Put together
+        self.network = nn.Sequential(layer_norm, bottleneck_conv1x1, temporal_conv_net,
+                                     mask_conv1x1)
+
+    def forward(self, mixture_w):
+        """
+        Keep this API same with TasNet
+        Args:
+            mixture_w: [M, N, K], M is batch size
+        returns:
+            est_mask: [M, C, N, K]
+        """
+        M, N, K = mixture_w.size()
+        score = self.network(mixture_w)  # [M, N, K] -> [M, C*N, K]
+        score = score.view(M, self.C, N, K)  # [M, C*N, K] -> [M, C, N, K]
+        if self.mask_nonlinear == 'softmax':
+            est_mask = F.softmax(score, dim=1)
+        elif self.mask_nonlinear == 'relu':
+            est_mask = F.relu(score)
+        else:
+            raise ValueError("Unsupported mask non-linear function")
+        return est_mask
+
+
+class TemporalBlock(nn.Module):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 kernel_size,
+                 stride,
+                 padding,
+                 dilation,
+                 norm_type="gLN",
+                 causal=False):
+        super(TemporalBlock, self).__init__()
+        # [M, B, K] -> [M, H, K]
+        conv1x1 = nn.Conv1d(in_channels, out_channels, 1, bias=False)
+        prelu = nn.PReLU()
+        norm = chose_norm(norm_type, out_channels)
+        # [M, H, K] -> [M, B, K]
+        dsconv = DepthwiseSeparableConv(out_channels, in_channels, kernel_size, stride, padding,
+                                        dilation, norm_type, causal)
+        # Put together
+        self.net = nn.Sequential(conv1x1, prelu, norm, dsconv)
+
+    def forward(self, x):
+        """
+        Args:
+            x: [M, B, K]
+        Returns:
+            [M, B, K]
+        """
+        residual = x
+        out = self.net(x)
+        # TODO: when P = 3 here works fine, but when P = 2 maybe need to pad?
+        return out + residual  # look like w/o F.relu is better than w/ F.relu
+        # return F.relu(out + residual)
+
+
+class DepthwiseSeparableConv(nn.Module):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 kernel_size,
+                 stride,
+                 padding,
+                 dilation,
+                 norm_type="gLN",
+                 causal=False):
+        super(DepthwiseSeparableConv, self).__init__()
+        # Use `groups` option to implement depthwise convolution
+        # [M, H, K] -> [M, H, K]
+        depthwise_conv = nn.Conv1d(in_channels,
+                                   in_channels,
+                                   kernel_size,
+                                   stride=stride,
+                                   padding=padding,
+                                   dilation=dilation,
+                                   groups=in_channels,
+                                   bias=False)
+        if causal:
+            chomp = Chomp1d(padding)
+        prelu = nn.PReLU()
+        norm = chose_norm(norm_type, in_channels)
+        # [M, H, K] -> [M, B, K]
+        pointwise_conv = nn.Conv1d(in_channels, out_channels, 1, bias=False)
+        # Put together
+        if causal:
+            self.net = nn.Sequential(depthwise_conv, chomp, prelu, norm, pointwise_conv)
+        else:
+            self.net = nn.Sequential(depthwise_conv, prelu, norm, pointwise_conv)
+
+    def forward(self, x):
+        """
+        Args:
+            x: [M, H, K]
+        Returns:
+            result: [M, B, K]
+        """
+        return self.net(x)
+
+
+class Chomp1d(nn.Module):
+    """To ensure the output length is the same as the input.
+    """
+    def __init__(self, chomp_size):
+        super(Chomp1d, self).__init__()
+        self.chomp_size = chomp_size
+
+    def forward(self, x):
+        """
+        Args:
+            x: [M, H, Kpad]
+        Returns:
+            [M, H, K]
+        """
+        return x[:, :, :-self.chomp_size].contiguous()
+
+
+def chose_norm(norm_type, channel_size):
+    """The input of normlization will be (M, C, K), where M is batch size,
+       C is channel size and K is sequence length.
+    """
+    if norm_type == "gLN":
+        return GlobalLayerNorm(channel_size)
+    elif norm_type == "cLN":
+        return ChannelwiseLayerNorm(channel_size)
+    elif norm_type == "id":
+        return nn.Identity()
+    else:  # norm_type == "BN":
+        # Given input (M, C, K), nn.BatchNorm1d(C) will accumulate statics
+        # along M and K, so this BN usage is right.
+        return nn.BatchNorm1d(channel_size)
+
+
+# TODO: Use nn.LayerNorm to impl cLN to speed up
+class ChannelwiseLayerNorm(nn.Module):
+    """Channel-wise Layer Normalization (cLN)"""
+    def __init__(self, channel_size):
+        super(ChannelwiseLayerNorm, self).__init__()
+        self.gamma = nn.Parameter(torch.Tensor(1, channel_size, 1))  # [1, N, 1]
+        self.beta = nn.Parameter(torch.Tensor(1, channel_size, 1))  # [1, N, 1]
+        self.reset_parameters()
+
+    def reset_parameters(self):
+        self.gamma.data.fill_(1)
+        self.beta.data.zero_()
+
+    def forward(self, y):
+        """
+        Args:
+            y: [M, N, K], M is batch size, N is channel size, K is length
+        Returns:
+            cLN_y: [M, N, K]
+        """
+        mean = torch.mean(y, dim=1, keepdim=True)  # [M, 1, K]
+        var = torch.var(y, dim=1, keepdim=True, unbiased=False)  # [M, 1, K]
+        cLN_y = self.gamma * (y - mean) / torch.pow(var + EPS, 0.5) + self.beta
+        return cLN_y
+
+
+class GlobalLayerNorm(nn.Module):
+    """Global Layer Normalization (gLN)"""
+    def __init__(self, channel_size):
+        super(GlobalLayerNorm, self).__init__()
+        self.gamma = nn.Parameter(torch.Tensor(1, channel_size, 1))  # [1, N, 1]
+        self.beta = nn.Parameter(torch.Tensor(1, channel_size, 1))  # [1, N, 1]
+        self.reset_parameters()
+
+    def reset_parameters(self):
+        self.gamma.data.fill_(1)
+        self.beta.data.zero_()
+
+    def forward(self, y):
+        """
+        Args:
+            y: [M, N, K], M is batch size, N is channel size, K is length
+        Returns:
+            gLN_y: [M, N, K]
+        """
+        # TODO: in torch 1.0, torch.mean() support dim list
+        mean = y.mean(dim=1, keepdim=True).mean(dim=2, keepdim=True)  # [M, 1, 1]
+        var = (torch.pow(y - mean, 2)).mean(dim=1, keepdim=True).mean(dim=2, keepdim=True)
+        gLN_y = self.gamma * (y - mean) / torch.pow(var + EPS, 0.5) + self.beta
+        return gLN_y
+
+
+if __name__ == "__main__":
+    torch.manual_seed(123)
+    M, N, L, T = 2, 3, 4, 12
+    K = 2 * T // L - 1
+    B, H, P, X, R, C, norm_type, causal = 2, 3, 3, 3, 2, 2, "gLN", False
+    mixture = torch.randint(3, (M, T))
+    # test Encoder
+    encoder = Encoder(L, N)
+    encoder.conv1d_U.weight.data = torch.randint(2, encoder.conv1d_U.weight.size())
+    mixture_w = encoder(mixture)
+    print('mixture', mixture)
+    print('U', encoder.conv1d_U.weight)
+    print('mixture_w', mixture_w)
+    print('mixture_w size', mixture_w.size())
+
+    # test TemporalConvNet
+    separator = TemporalConvNet(N, B, H, P, X, R, C, norm_type=norm_type, causal=causal)
+    est_mask = separator(mixture_w)
+    print('est_mask', est_mask)
+
+    # test Decoder
+    decoder = Decoder(N, L)
+    est_mask = torch.randint(2, (B, K, C, N))
+    est_source = decoder(mixture_w, est_mask)
+    print('est_source', est_source)
+
+    # test Conv-TasNet
+    conv_tasnet = ConvTasNet(N, L, B, H, P, X, R, C, norm_type=norm_type)
+    est_source = conv_tasnet(mixture)
+    print('est_source', est_source)
+    print('est_source size', est_source.size())

demucs/tasnet_v2.py

@@ -0,0 +1,452 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+#
+# Created on 2018/12
+# Author: Kaituo XU
+# Modified on 2019/11 by Alexandre Defossez, added support for multiple output channels
+# Here is the original license:
+# The MIT License (MIT)
+#
+# Copyright (c) 2018 Kaituo XU
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to deal
+# in the Software without restriction, including without limitation the rights
+# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+# copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in all
+# copies or substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+# SOFTWARE.
+
+import math
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from .utils import capture_init
+
+EPS = 1e-8
+
+
+def overlap_and_add(signal, frame_step):
+    outer_dimensions = signal.size()[:-2]
+    frames, frame_length = signal.size()[-2:]
+
+    subframe_length = math.gcd(frame_length, frame_step)  # gcd=Greatest Common Divisor
+    subframe_step = frame_step // subframe_length
+    subframes_per_frame = frame_length // subframe_length
+    output_size = frame_step * (frames - 1) + frame_length
+    output_subframes = output_size // subframe_length
+
+    subframe_signal = signal.view(*outer_dimensions, -1, subframe_length)
+
+    frame = torch.arange(0, output_subframes,
+                         device=signal.device).unfold(0, subframes_per_frame, subframe_step)
+    frame = frame.long()  # signal may in GPU or CPU
+    frame = frame.contiguous().view(-1)
+
+    result = signal.new_zeros(*outer_dimensions, output_subframes, subframe_length)
+    result.index_add_(-2, frame, subframe_signal)
+    result = result.view(*outer_dimensions, -1)
+    return result
+
+
+class ConvTasNet(nn.Module):
+    @capture_init
+    def __init__(self,
+                 sources,
+                 N=256,
+                 L=20,
+                 B=256,
+                 H=512,
+                 P=3,
+                 X=8,
+                 R=4,
+                 audio_channels=2,
+                 norm_type="gLN",
+                 causal=False,
+                 mask_nonlinear='relu',
+                 samplerate=44100,
+                 segment_length=44100 * 2 * 4):
+        """
+        Args:
+            sources: list of sources
+            N: Number of filters in autoencoder
+            L: Length of the filters (in samples)
+            B: Number of channels in bottleneck 1 × 1-conv block
+            H: Number of channels in convolutional blocks
+            P: Kernel size in convolutional blocks
+            X: Number of convolutional blocks in each repeat
+            R: Number of repeats
+            norm_type: BN, gLN, cLN
+            causal: causal or non-causal
+            mask_nonlinear: use which non-linear function to generate mask
+        """
+        super(ConvTasNet, self).__init__()
+        # Hyper-parameter
+        self.sources = sources
+        self.C = len(sources)
+        self.N, self.L, self.B, self.H, self.P, self.X, self.R = N, L, B, H, P, X, R
+        self.norm_type = norm_type
+        self.causal = causal
+        self.mask_nonlinear = mask_nonlinear
+        self.audio_channels = audio_channels
+        self.samplerate = samplerate
+        self.segment_length = segment_length
+        # Components
+        self.encoder = Encoder(L, N, audio_channels)
+        self.separator = TemporalConvNet(
+            N, B, H, P, X, R, self.C, norm_type, causal, mask_nonlinear)
+        self.decoder = Decoder(N, L, audio_channels)
+        # init
+        for p in self.parameters():
+            if p.dim() > 1:
+                nn.init.xavier_normal_(p)
+
+    def valid_length(self, length):
+        return length
+
+    def forward(self, mixture):
+        """
+        Args:
+            mixture: [M, T], M is batch size, T is #samples
+        Returns:
+            est_source: [M, C, T]
+        """
+        mixture_w = self.encoder(mixture)
+        est_mask = self.separator(mixture_w)
+        est_source = self.decoder(mixture_w, est_mask)
+
+        # T changed after conv1d in encoder, fix it here
+        T_origin = mixture.size(-1)
+        T_conv = est_source.size(-1)
+        est_source = F.pad(est_source, (0, T_origin - T_conv))
+        return est_source
+
+
+class Encoder(nn.Module):
+    """Estimation of the nonnegative mixture weight by a 1-D conv layer.
+    """
+    def __init__(self, L, N, audio_channels):
+        super(Encoder, self).__init__()
+        # Hyper-parameter
+        self.L, self.N = L, N
+        # Components
+        # 50% overlap
+        self.conv1d_U = nn.Conv1d(audio_channels, N, kernel_size=L, stride=L // 2, bias=False)
+
+    def forward(self, mixture):
+        """
+        Args:
+            mixture: [M, T], M is batch size, T is #samples
+        Returns:
+            mixture_w: [M, N, K], where K = (T-L)/(L/2)+1 = 2T/L-1
+        """
+        mixture_w = F.relu(self.conv1d_U(mixture))  # [M, N, K]
+        return mixture_w
+
+
+class Decoder(nn.Module):
+    def __init__(self, N, L, audio_channels):
+        super(Decoder, self).__init__()
+        # Hyper-parameter
+        self.N, self.L = N, L
+        self.audio_channels = audio_channels
+        # Components
+        self.basis_signals = nn.Linear(N, audio_channels * L, bias=False)
+
+    def forward(self, mixture_w, est_mask):
+        """
+        Args:
+            mixture_w: [M, N, K]
+            est_mask: [M, C, N, K]
+        Returns:
+            est_source: [M, C, T]
+        """
+        # D = W * M
+        source_w = torch.unsqueeze(mixture_w, 1) * est_mask  # [M, C, N, K]
+        source_w = torch.transpose(source_w, 2, 3)  # [M, C, K, N]
+        # S = DV
+        est_source = self.basis_signals(source_w)  # [M, C, K, ac * L]
+        m, c, k, _ = est_source.size()
+        est_source = est_source.view(m, c, k, self.audio_channels, -1).transpose(2, 3).contiguous()
+        est_source = overlap_and_add(est_source, self.L // 2)  # M x C x ac x T
+        return est_source
+
+
+class TemporalConvNet(nn.Module):
+    def __init__(self, N, B, H, P, X, R, C, norm_type="gLN", causal=False, mask_nonlinear='relu'):
+        """
+        Args:
+            N: Number of filters in autoencoder
+            B: Number of channels in bottleneck 1 × 1-conv block
+            H: Number of channels in convolutional blocks
+            P: Kernel size in convolutional blocks
+            X: Number of convolutional blocks in each repeat
+            R: Number of repeats
+            C: Number of speakers
+            norm_type: BN, gLN, cLN
+            causal: causal or non-causal
+            mask_nonlinear: use which non-linear function to generate mask
+        """
+        super(TemporalConvNet, self).__init__()
+        # Hyper-parameter
+        self.C = C
+        self.mask_nonlinear = mask_nonlinear
+        # Components
+        # [M, N, K] -> [M, N, K]
+        layer_norm = ChannelwiseLayerNorm(N)
+        # [M, N, K] -> [M, B, K]
+        bottleneck_conv1x1 = nn.Conv1d(N, B, 1, bias=False)
+        # [M, B, K] -> [M, B, K]
+        repeats = []
+        for r in range(R):
+            blocks = []
+            for x in range(X):
+                dilation = 2**x
+                padding = (P - 1) * dilation if causal else (P - 1) * dilation // 2
+                blocks += [
+                    TemporalBlock(B,
+                                  H,
+                                  P,
+                                  stride=1,
+                                  padding=padding,
+                                  dilation=dilation,
+                                  norm_type=norm_type,
+                                  causal=causal)
+                ]
+            repeats += [nn.Sequential(*blocks)]
+        temporal_conv_net = nn.Sequential(*repeats)
+        # [M, B, K] -> [M, C*N, K]
+        mask_conv1x1 = nn.Conv1d(B, C * N, 1, bias=False)
+        # Put together
+        self.network = nn.Sequential(layer_norm, bottleneck_conv1x1, temporal_conv_net,
+                                     mask_conv1x1)
+
+    def forward(self, mixture_w):
+        """
+        Keep this API same with TasNet
+        Args:
+            mixture_w: [M, N, K], M is batch size
+        returns:
+            est_mask: [M, C, N, K]
+        """
+        M, N, K = mixture_w.size()
+        score = self.network(mixture_w)  # [M, N, K] -> [M, C*N, K]
+        score = score.view(M, self.C, N, K)  # [M, C*N, K] -> [M, C, N, K]
+        if self.mask_nonlinear == 'softmax':
+            est_mask = F.softmax(score, dim=1)
+        elif self.mask_nonlinear == 'relu':
+            est_mask = F.relu(score)
+        else:
+            raise ValueError("Unsupported mask non-linear function")
+        return est_mask
+
+
+class TemporalBlock(nn.Module):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 kernel_size,
+                 stride,
+                 padding,
+                 dilation,
+                 norm_type="gLN",
+                 causal=False):
+        super(TemporalBlock, self).__init__()
+        # [M, B, K] -> [M, H, K]
+        conv1x1 = nn.Conv1d(in_channels, out_channels, 1, bias=False)
+        prelu = nn.PReLU()
+        norm = chose_norm(norm_type, out_channels)
+        # [M, H, K] -> [M, B, K]
+        dsconv = DepthwiseSeparableConv(out_channels, in_channels, kernel_size, stride, padding,
+                                        dilation, norm_type, causal)
+        # Put together
+        self.net = nn.Sequential(conv1x1, prelu, norm, dsconv)
+
+    def forward(self, x):
+        """
+        Args:
+            x: [M, B, K]
+        Returns:
+            [M, B, K]
+        """
+        residual = x
+        out = self.net(x)
+        # TODO: when P = 3 here works fine, but when P = 2 maybe need to pad?
+        return out + residual  # look like w/o F.relu is better than w/ F.relu
+        # return F.relu(out + residual)
+
+
+class DepthwiseSeparableConv(nn.Module):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 kernel_size,
+                 stride,
+                 padding,
+                 dilation,
+                 norm_type="gLN",
+                 causal=False):
+        super(DepthwiseSeparableConv, self).__init__()
+        # Use `groups` option to implement depthwise convolution
+        # [M, H, K] -> [M, H, K]
+        depthwise_conv = nn.Conv1d(in_channels,
+                                   in_channels,
+                                   kernel_size,
+                                   stride=stride,
+                                   padding=padding,
+                                   dilation=dilation,
+                                   groups=in_channels,
+                                   bias=False)
+        if causal:
+            chomp = Chomp1d(padding)
+        prelu = nn.PReLU()
+        norm = chose_norm(norm_type, in_channels)
+        # [M, H, K] -> [M, B, K]
+        pointwise_conv = nn.Conv1d(in_channels, out_channels, 1, bias=False)
+        # Put together
+        if causal:
+            self.net = nn.Sequential(depthwise_conv, chomp, prelu, norm, pointwise_conv)
+        else:
+            self.net = nn.Sequential(depthwise_conv, prelu, norm, pointwise_conv)
+
+    def forward(self, x):
+        """
+        Args:
+            x: [M, H, K]
+        Returns:
+            result: [M, B, K]
+        """
+        return self.net(x)
+
+
+class Chomp1d(nn.Module):
+    """To ensure the output length is the same as the input.
+    """
+    def __init__(self, chomp_size):
+        super(Chomp1d, self).__init__()
+        self.chomp_size = chomp_size
+
+    def forward(self, x):
+        """
+        Args:
+            x: [M, H, Kpad]
+        Returns:
+            [M, H, K]
+        """
+        return x[:, :, :-self.chomp_size].contiguous()
+
+
+def chose_norm(norm_type, channel_size):
+    """The input of normlization will be (M, C, K), where M is batch size,
+       C is channel size and K is sequence length.
+    """
+    if norm_type == "gLN":
+        return GlobalLayerNorm(channel_size)
+    elif norm_type == "cLN":
+        return ChannelwiseLayerNorm(channel_size)
+    elif norm_type == "id":
+        return nn.Identity()
+    else:  # norm_type == "BN":
+        # Given input (M, C, K), nn.BatchNorm1d(C) will accumulate statics
+        # along M and K, so this BN usage is right.
+        return nn.BatchNorm1d(channel_size)
+
+
+# TODO: Use nn.LayerNorm to impl cLN to speed up
+class ChannelwiseLayerNorm(nn.Module):
+    """Channel-wise Layer Normalization (cLN)"""
+    def __init__(self, channel_size):
+        super(ChannelwiseLayerNorm, self).__init__()
+        self.gamma = nn.Parameter(torch.Tensor(1, channel_size, 1))  # [1, N, 1]
+        self.beta = nn.Parameter(torch.Tensor(1, channel_size, 1))  # [1, N, 1]
+        self.reset_parameters()
+
+    def reset_parameters(self):
+        self.gamma.data.fill_(1)
+        self.beta.data.zero_()
+
+    def forward(self, y):
+        """
+        Args:
+            y: [M, N, K], M is batch size, N is channel size, K is length
+        Returns:
+            cLN_y: [M, N, K]
+        """
+        mean = torch.mean(y, dim=1, keepdim=True)  # [M, 1, K]
+        var = torch.var(y, dim=1, keepdim=True, unbiased=False)  # [M, 1, K]
+        cLN_y = self.gamma * (y - mean) / torch.pow(var + EPS, 0.5) + self.beta
+        return cLN_y
+
+
+class GlobalLayerNorm(nn.Module):
+    """Global Layer Normalization (gLN)"""
+    def __init__(self, channel_size):
+        super(GlobalLayerNorm, self).__init__()
+        self.gamma = nn.Parameter(torch.Tensor(1, channel_size, 1))  # [1, N, 1]
+        self.beta = nn.Parameter(torch.Tensor(1, channel_size, 1))  # [1, N, 1]
+        self.reset_parameters()
+
+    def reset_parameters(self):
+        self.gamma.data.fill_(1)
+        self.beta.data.zero_()
+
+    def forward(self, y):
+        """
+        Args:
+            y: [M, N, K], M is batch size, N is channel size, K is length
+        Returns:
+            gLN_y: [M, N, K]
+        """
+        # TODO: in torch 1.0, torch.mean() support dim list
+        mean = y.mean(dim=1, keepdim=True).mean(dim=2, keepdim=True)  # [M, 1, 1]
+        var = (torch.pow(y - mean, 2)).mean(dim=1, keepdim=True).mean(dim=2, keepdim=True)
+        gLN_y = self.gamma * (y - mean) / torch.pow(var + EPS, 0.5) + self.beta
+        return gLN_y
+
+
+if __name__ == "__main__":
+    torch.manual_seed(123)
+    M, N, L, T = 2, 3, 4, 12
+    K = 2 * T // L - 1
+    B, H, P, X, R, C, norm_type, causal = 2, 3, 3, 3, 2, 2, "gLN", False
+    mixture = torch.randint(3, (M, T))
+    # test Encoder
+    encoder = Encoder(L, N)
+    encoder.conv1d_U.weight.data = torch.randint(2, encoder.conv1d_U.weight.size())
+    mixture_w = encoder(mixture)
+    print('mixture', mixture)
+    print('U', encoder.conv1d_U.weight)
+    print('mixture_w', mixture_w)
+    print('mixture_w size', mixture_w.size())
+
+    # test TemporalConvNet
+    separator = TemporalConvNet(N, B, H, P, X, R, C, norm_type=norm_type, causal=causal)
+    est_mask = separator(mixture_w)
+    print('est_mask', est_mask)
+
+    # test Decoder
+    decoder = Decoder(N, L)
+    est_mask = torch.randint(2, (B, K, C, N))
+    est_source = decoder(mixture_w, est_mask)
+    print('est_source', est_source)
+
+    # test Conv-TasNet
+    conv_tasnet = ConvTasNet(N, L, B, H, P, X, R, C, norm_type=norm_type)
+    est_source = conv_tasnet(mixture)
+    print('est_source', est_source)
+    print('est_source size', est_source.size())

demucs/transformer.py

@@ -0,0 +1,839 @@
+# Copyright (c) 2019-present, Meta, Inc.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+# First author is Simon Rouard.
+
+import random
+import typing as tp
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import numpy as np
+import math
+from einops import rearrange
+
+
+def create_sin_embedding(
+    length: int, dim: int, shift: int = 0, device="cpu", max_period=10000
+):
+    # We aim for TBC format
+    assert dim % 2 == 0
+    pos = shift + torch.arange(length, device=device).view(-1, 1, 1)
+    half_dim = dim // 2
+    adim = torch.arange(dim // 2, device=device).view(1, 1, -1)
+    phase = pos / (max_period ** (adim / (half_dim - 1)))
+    return torch.cat(
+        [
+            torch.cos(phase),
+            torch.sin(phase),
+        ],
+        dim=-1,
+    )
+
+
+def create_2d_sin_embedding(d_model, height, width, device="cpu", max_period=10000):
+    """
+    :param d_model: dimension of the model
+    :param height: height of the positions
+    :param width: width of the positions
+    :return: d_model*height*width position matrix
+    """
+    if d_model % 4 != 0:
+        raise ValueError(
+            "Cannot use sin/cos positional encoding with "
+            "odd dimension (got dim={:d})".format(d_model)
+        )
+    pe = torch.zeros(d_model, height, width)
+    # Each dimension use half of d_model
+    d_model = int(d_model / 2)
+    div_term = torch.exp(
+        torch.arange(0.0, d_model, 2) * -(math.log(max_period) / d_model)
+    )
+    pos_w = torch.arange(0.0, width).unsqueeze(1)
+    pos_h = torch.arange(0.0, height).unsqueeze(1)
+    pe[0:d_model:2, :, :] = (
+        torch.sin(pos_w * div_term).transpose(0, 1).unsqueeze(1).repeat(1, height, 1)
+    )
+    pe[1:d_model:2, :, :] = (
+        torch.cos(pos_w * div_term).transpose(0, 1).unsqueeze(1).repeat(1, height, 1)
+    )
+    pe[d_model::2, :, :] = (
+        torch.sin(pos_h * div_term).transpose(0, 1).unsqueeze(2).repeat(1, 1, width)
+    )
+    pe[d_model + 1:: 2, :, :] = (
+        torch.cos(pos_h * div_term).transpose(0, 1).unsqueeze(2).repeat(1, 1, width)
+    )
+
+    return pe[None, :].to(device)
+
+
+def create_sin_embedding_cape(
+    length: int,
+    dim: int,
+    batch_size: int,
+    mean_normalize: bool,
+    augment: bool,  # True during training
+    max_global_shift: float = 0.0,  # delta max
+    max_local_shift: float = 0.0,  # epsilon max
+    max_scale: float = 1.0,
+    device: str = "cpu",
+    max_period: float = 10000.0,
+):
+    # We aim for TBC format
+    assert dim % 2 == 0
+    pos = 1.0 * torch.arange(length).view(-1, 1, 1)  # (length, 1, 1)
+    pos = pos.repeat(1, batch_size, 1)  # (length, batch_size, 1)
+    if mean_normalize:
+        pos -= torch.nanmean(pos, dim=0, keepdim=True)
+
+    if augment:
+        delta = np.random.uniform(
+            -max_global_shift, +max_global_shift, size=[1, batch_size, 1]
+        )
+        delta_local = np.random.uniform(
+            -max_local_shift, +max_local_shift, size=[length, batch_size, 1]
+        )
+        log_lambdas = np.random.uniform(
+            -np.log(max_scale), +np.log(max_scale), size=[1, batch_size, 1]
+        )
+        pos = (pos + delta + delta_local) * np.exp(log_lambdas)
+
+    pos = pos.to(device)
+
+    half_dim = dim // 2
+    adim = torch.arange(dim // 2, device=device).view(1, 1, -1)
+    phase = pos / (max_period ** (adim / (half_dim - 1)))
+    return torch.cat(
+        [
+            torch.cos(phase),
+            torch.sin(phase),
+        ],
+        dim=-1,
+    ).float()
+
+
+def get_causal_mask(length):
+    pos = torch.arange(length)
+    return pos > pos[:, None]
+
+
+def get_elementary_mask(
+    T1,
+    T2,
+    mask_type,
+    sparse_attn_window,
+    global_window,
+    mask_random_seed,
+    sparsity,
+    device,
+):
+    """
+    When the input of the Decoder has length T1 and the output T2
+    The mask matrix has shape (T2, T1)
+    """
+    assert mask_type in ["diag", "jmask", "random", "global"]
+
+    if mask_type == "global":
+        mask = torch.zeros(T2, T1, dtype=torch.bool)
+        mask[:, :global_window] = True
+        line_window = int(global_window * T2 / T1)
+        mask[:line_window, :] = True
+
+    if mask_type == "diag":
+
+        mask = torch.zeros(T2, T1, dtype=torch.bool)
+        rows = torch.arange(T2)[:, None]
+        cols = (
+            (T1 / T2 * rows + torch.arange(-sparse_attn_window, sparse_attn_window + 1))
+            .long()
+            .clamp(0, T1 - 1)
+        )
+        mask.scatter_(1, cols, torch.ones(1, dtype=torch.bool).expand_as(cols))
+
+    elif mask_type == "jmask":
+        mask = torch.zeros(T2 + 2, T1 + 2, dtype=torch.bool)
+        rows = torch.arange(T2 + 2)[:, None]
+        t = torch.arange(0, int((2 * T1) ** 0.5 + 1))
+        t = (t * (t + 1) / 2).int()
+        t = torch.cat([-t.flip(0)[:-1], t])
+        cols = (T1 / T2 * rows + t).long().clamp(0, T1 + 1)
+        mask.scatter_(1, cols, torch.ones(1, dtype=torch.bool).expand_as(cols))
+        mask = mask[1:-1, 1:-1]
+
+    elif mask_type == "random":
+        gene = torch.Generator(device=device)
+        gene.manual_seed(mask_random_seed)
+        mask = (
+            torch.rand(T1 * T2, generator=gene, device=device).reshape(T2, T1)
+            > sparsity
+        )
+
+    mask = mask.to(device)
+    return mask
+
+
+def get_mask(
+    T1,
+    T2,
+    mask_type,
+    sparse_attn_window,
+    global_window,
+    mask_random_seed,
+    sparsity,
+    device,
+):
+    """
+    Return a SparseCSRTensor mask that is a combination of elementary masks
+    mask_type can be a combination of multiple masks: for instance "diag_jmask_random"
+    """
+    from xformers.sparse import SparseCSRTensor
+    # create a list
+    mask_types = mask_type.split("_")
+
+    all_masks = [
+        get_elementary_mask(
+            T1,
+            T2,
+            mask,
+            sparse_attn_window,
+            global_window,
+            mask_random_seed,
+            sparsity,
+            device,
+        )
+        for mask in mask_types
+    ]
+
+    final_mask = torch.stack(all_masks).sum(axis=0) > 0
+
+    return SparseCSRTensor.from_dense(final_mask[None])
+
+
+class ScaledEmbedding(nn.Module):
+    def __init__(
+        self,
+        num_embeddings: int,
+        embedding_dim: int,
+        scale: float = 1.0,
+        boost: float = 3.0,
+    ):
+        super().__init__()
+        self.embedding = nn.Embedding(num_embeddings, embedding_dim)
+        self.embedding.weight.data *= scale / boost
+        self.boost = boost
+
+    @property
+    def weight(self):
+        return self.embedding.weight * self.boost
+
+    def forward(self, x):
+        return self.embedding(x) * self.boost
+
+
+class LayerScale(nn.Module):
+    """Layer scale from [Touvron et al 2021] (https://arxiv.org/pdf/2103.17239.pdf).
+    This rescales diagonaly residual outputs close to 0 initially, then learnt.
+    """
+
+    def __init__(self, channels: int, init: float = 0, channel_last=False):
+        """
+        channel_last = False corresponds to (B, C, T) tensors
+        channel_last = True corresponds to (T, B, C) tensors
+        """
+        super().__init__()
+        self.channel_last = channel_last
+        self.scale = nn.Parameter(torch.zeros(channels, requires_grad=True))
+        self.scale.data[:] = init
+
+    def forward(self, x):
+        if self.channel_last:
+            return self.scale * x
+        else:
+            return self.scale[:, None] * x
+
+
+class MyGroupNorm(nn.GroupNorm):
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+
+    def forward(self, x):
+        """
+        x: (B, T, C)
+        if num_groups=1: Normalisation on all T and C together for each B
+        """
+        x = x.transpose(1, 2)
+        return super().forward(x).transpose(1, 2)
+
+
+class MyTransformerEncoderLayer(nn.TransformerEncoderLayer):
+    def __init__(
+        self,
+        d_model,
+        nhead,
+        dim_feedforward=2048,
+        dropout=0.1,
+        activation=F.relu,
+        group_norm=0,
+        norm_first=False,
+        norm_out=False,
+        layer_norm_eps=1e-5,
+        layer_scale=False,
+        init_values=1e-4,
+        device=None,
+        dtype=None,
+        sparse=False,
+        mask_type="diag",
+        mask_random_seed=42,
+        sparse_attn_window=500,
+        global_window=50,
+        auto_sparsity=False,
+        sparsity=0.95,
+        batch_first=False,
+    ):
+        factory_kwargs = {"device": device, "dtype": dtype}
+        super().__init__(
+            d_model=d_model,
+            nhead=nhead,
+            dim_feedforward=dim_feedforward,
+            dropout=dropout,
+            activation=activation,
+            layer_norm_eps=layer_norm_eps,
+            batch_first=batch_first,
+            norm_first=norm_first,
+            device=device,
+            dtype=dtype,
+        )
+        self.sparse = sparse
+        self.auto_sparsity = auto_sparsity
+        if sparse:
+            if not auto_sparsity:
+                self.mask_type = mask_type
+                self.sparse_attn_window = sparse_attn_window
+                self.global_window = global_window
+            self.sparsity = sparsity
+        if group_norm:
+            self.norm1 = MyGroupNorm(int(group_norm), d_model, eps=layer_norm_eps, **factory_kwargs)
+            self.norm2 = MyGroupNorm(int(group_norm), d_model, eps=layer_norm_eps, **factory_kwargs)
+
+        self.norm_out = None
+        if self.norm_first & norm_out:
+            self.norm_out = MyGroupNorm(num_groups=int(norm_out), num_channels=d_model)
+        self.gamma_1 = (
+            LayerScale(d_model, init_values, True) if layer_scale else nn.Identity()
+        )
+        self.gamma_2 = (
+            LayerScale(d_model, init_values, True) if layer_scale else nn.Identity()
+        )
+
+        if sparse:
+            self.self_attn = MultiheadAttention(
+                d_model, nhead, dropout=dropout, batch_first=batch_first,
+                auto_sparsity=sparsity if auto_sparsity else 0,
+            )
+            self.__setattr__("src_mask", torch.zeros(1, 1))
+            self.mask_random_seed = mask_random_seed
+
+    def forward(self, src, src_mask=None, src_key_padding_mask=None):
+        """
+        if batch_first = False, src shape is (T, B, C)
+        the case where batch_first=True is not covered
+        """
+        device = src.device
+        x = src
+        T, B, C = x.shape
+        if self.sparse and not self.auto_sparsity:
+            assert src_mask is None
+            src_mask = self.src_mask
+            if src_mask.shape[-1] != T:
+                src_mask = get_mask(
+                    T,
+                    T,
+                    self.mask_type,
+                    self.sparse_attn_window,
+                    self.global_window,
+                    self.mask_random_seed,
+                    self.sparsity,
+                    device,
+                )
+                self.__setattr__("src_mask", src_mask)
+
+        if self.norm_first:
+            x = x + self.gamma_1(
+                self._sa_block(self.norm1(x), src_mask, src_key_padding_mask)
+            )
+            x = x + self.gamma_2(self._ff_block(self.norm2(x)))
+
+            if self.norm_out:
+                x = self.norm_out(x)
+        else:
+            x = self.norm1(
+                x + self.gamma_1(self._sa_block(x, src_mask, src_key_padding_mask))
+            )
+            x = self.norm2(x + self.gamma_2(self._ff_block(x)))
+
+        return x
+
+
+class CrossTransformerEncoderLayer(nn.Module):
+    def __init__(
+        self,
+        d_model: int,
+        nhead: int,
+        dim_feedforward: int = 2048,
+        dropout: float = 0.1,
+        activation=F.relu,
+        layer_norm_eps: float = 1e-5,
+        layer_scale: bool = False,
+        init_values: float = 1e-4,
+        norm_first: bool = False,
+        group_norm: bool = False,
+        norm_out: bool = False,
+        sparse=False,
+        mask_type="diag",
+        mask_random_seed=42,
+        sparse_attn_window=500,
+        global_window=50,
+        sparsity=0.95,
+        auto_sparsity=None,
+        device=None,
+        dtype=None,
+        batch_first=False,
+    ):
+        factory_kwargs = {"device": device, "dtype": dtype}
+        super().__init__()
+
+        self.sparse = sparse
+        self.auto_sparsity = auto_sparsity
+        if sparse:
+            if not auto_sparsity:
+                self.mask_type = mask_type
+                self.sparse_attn_window = sparse_attn_window
+                self.global_window = global_window
+            self.sparsity = sparsity
+
+        self.cross_attn: nn.Module
+        self.cross_attn = nn.MultiheadAttention(
+            d_model, nhead, dropout=dropout, batch_first=batch_first)
+        # Implementation of Feedforward model
+        self.linear1 = nn.Linear(d_model, dim_feedforward, **factory_kwargs)
+        self.dropout = nn.Dropout(dropout)
+        self.linear2 = nn.Linear(dim_feedforward, d_model, **factory_kwargs)
+
+        self.norm_first = norm_first
+        self.norm1: nn.Module
+        self.norm2: nn.Module
+        self.norm3: nn.Module
+        if group_norm:
+            self.norm1 = MyGroupNorm(int(group_norm), d_model, eps=layer_norm_eps, **factory_kwargs)
+            self.norm2 = MyGroupNorm(int(group_norm), d_model, eps=layer_norm_eps, **factory_kwargs)
+            self.norm3 = MyGroupNorm(int(group_norm), d_model, eps=layer_norm_eps, **factory_kwargs)
+        else:
+            self.norm1 = nn.LayerNorm(d_model, eps=layer_norm_eps, **factory_kwargs)
+            self.norm2 = nn.LayerNorm(d_model, eps=layer_norm_eps, **factory_kwargs)
+            self.norm3 = nn.LayerNorm(d_model, eps=layer_norm_eps, **factory_kwargs)
+
+        self.norm_out = None
+        if self.norm_first & norm_out:
+            self.norm_out = MyGroupNorm(num_groups=int(norm_out), num_channels=d_model)
+
+        self.gamma_1 = (
+            LayerScale(d_model, init_values, True) if layer_scale else nn.Identity()
+        )
+        self.gamma_2 = (
+            LayerScale(d_model, init_values, True) if layer_scale else nn.Identity()
+        )
+
+        self.dropout1 = nn.Dropout(dropout)
+        self.dropout2 = nn.Dropout(dropout)
+
+        # Legacy string support for activation function.
+        if isinstance(activation, str):
+            self.activation = self._get_activation_fn(activation)
+        else:
+            self.activation = activation
+
+        if sparse:
+            self.cross_attn = MultiheadAttention(
+                d_model, nhead, dropout=dropout, batch_first=batch_first,
+                auto_sparsity=sparsity if auto_sparsity else 0)
+            if not auto_sparsity:
+                self.__setattr__("mask", torch.zeros(1, 1))
+                self.mask_random_seed = mask_random_seed
+
+    def forward(self, q, k, mask=None):
+        """
+        Args:
+            q: tensor of shape (T, B, C)
+            k: tensor of shape (S, B, C)
+            mask: tensor of shape (T, S)
+
+        """
+        device = q.device
+        T, B, C = q.shape
+        S, B, C = k.shape
+        if self.sparse and not self.auto_sparsity:
+            assert mask is None
+            mask = self.mask
+            if mask.shape[-1] != S or mask.shape[-2] != T:
+                mask = get_mask(
+                    S,
+                    T,
+                    self.mask_type,
+                    self.sparse_attn_window,
+                    self.global_window,
+                    self.mask_random_seed,
+                    self.sparsity,
+                    device,
+                )
+                self.__setattr__("mask", mask)
+
+        if self.norm_first:
+            x = q + self.gamma_1(self._ca_block(self.norm1(q), self.norm2(k), mask))
+            x = x + self.gamma_2(self._ff_block(self.norm3(x)))
+            if self.norm_out:
+                x = self.norm_out(x)
+        else:
+            x = self.norm1(q + self.gamma_1(self._ca_block(q, k, mask)))
+            x = self.norm2(x + self.gamma_2(self._ff_block(x)))
+
+        return x
+
+    # self-attention block
+    def _ca_block(self, q, k, attn_mask=None):
+        x = self.cross_attn(q, k, k, attn_mask=attn_mask, need_weights=False)[0]
+        return self.dropout1(x)
+
+    # feed forward block
+    def _ff_block(self, x):
+        x = self.linear2(self.dropout(self.activation(self.linear1(x))))
+        return self.dropout2(x)
+
+    def _get_activation_fn(self, activation):
+        if activation == "relu":
+            return F.relu
+        elif activation == "gelu":
+            return F.gelu
+
+        raise RuntimeError("activation should be relu/gelu, not {}".format(activation))
+
+
+# ----------------- MULTI-BLOCKS MODELS: -----------------------
+
+
+class CrossTransformerEncoder(nn.Module):
+    def __init__(
+        self,
+        dim: int,
+        emb: str = "sin",
+        hidden_scale: float = 4.0,
+        num_heads: int = 8,
+        num_layers: int = 6,
+        cross_first: bool = False,
+        dropout: float = 0.0,
+        max_positions: int = 1000,
+        norm_in: bool = True,
+        norm_in_group: bool = False,
+        group_norm: int = False,
+        norm_first: bool = False,
+        norm_out: bool = False,
+        max_period: float = 10000.0,
+        weight_decay: float = 0.0,
+        lr: tp.Optional[float] = None,
+        layer_scale: bool = False,
+        gelu: bool = True,
+        sin_random_shift: int = 0,
+        weight_pos_embed: float = 1.0,
+        cape_mean_normalize: bool = True,
+        cape_augment: bool = True,
+        cape_glob_loc_scale: list = [5000.0, 1.0, 1.4],
+        sparse_self_attn: bool = False,
+        sparse_cross_attn: bool = False,
+        mask_type: str = "diag",
+        mask_random_seed: int = 42,
+        sparse_attn_window: int = 500,
+        global_window: int = 50,
+        auto_sparsity: bool = False,
+        sparsity: float = 0.95,
+    ):
+        super().__init__()
+        """
+        """
+        assert dim % num_heads == 0
+
+        hidden_dim = int(dim * hidden_scale)
+
+        self.num_layers = num_layers
+        # classic parity = 1 means that if idx%2 == 1 there is a
+        # classical encoder else there is a cross encoder
+        self.classic_parity = 1 if cross_first else 0
+        self.emb = emb
+        self.max_period = max_period
+        self.weight_decay = weight_decay
+        self.weight_pos_embed = weight_pos_embed
+        self.sin_random_shift = sin_random_shift
+        if emb == "cape":
+            self.cape_mean_normalize = cape_mean_normalize
+            self.cape_augment = cape_augment
+            self.cape_glob_loc_scale = cape_glob_loc_scale
+        if emb == "scaled":
+            self.position_embeddings = ScaledEmbedding(max_positions, dim, scale=0.2)
+
+        self.lr = lr
+
+        activation: tp.Any = F.gelu if gelu else F.relu
+
+        self.norm_in: nn.Module
+        self.norm_in_t: nn.Module
+        if norm_in:
+            self.norm_in = nn.LayerNorm(dim)
+            self.norm_in_t = nn.LayerNorm(dim)
+        elif norm_in_group:
+            self.norm_in = MyGroupNorm(int(norm_in_group), dim)
+            self.norm_in_t = MyGroupNorm(int(norm_in_group), dim)
+        else:
+            self.norm_in = nn.Identity()
+            self.norm_in_t = nn.Identity()
+
+        # spectrogram layers
+        self.layers = nn.ModuleList()
+        # temporal layers
+        self.layers_t = nn.ModuleList()
+
+        kwargs_common = {
+            "d_model": dim,
+            "nhead": num_heads,
+            "dim_feedforward": hidden_dim,
+            "dropout": dropout,
+            "activation": activation,
+            "group_norm": group_norm,
+            "norm_first": norm_first,
+            "norm_out": norm_out,
+            "layer_scale": layer_scale,
+            "mask_type": mask_type,
+            "mask_random_seed": mask_random_seed,
+            "sparse_attn_window": sparse_attn_window,
+            "global_window": global_window,
+            "sparsity": sparsity,
+            "auto_sparsity": auto_sparsity,
+            "batch_first": True,
+        }
+
+        kwargs_classic_encoder = dict(kwargs_common)
+        kwargs_classic_encoder.update({
+            "sparse": sparse_self_attn,
+        })
+        kwargs_cross_encoder = dict(kwargs_common)
+        kwargs_cross_encoder.update({
+            "sparse": sparse_cross_attn,
+        })
+
+        for idx in range(num_layers):
+            if idx % 2 == self.classic_parity:
+
+                self.layers.append(MyTransformerEncoderLayer(**kwargs_classic_encoder))
+                self.layers_t.append(
+                    MyTransformerEncoderLayer(**kwargs_classic_encoder)
+                )
+
+            else:
+                self.layers.append(CrossTransformerEncoderLayer(**kwargs_cross_encoder))
+
+                self.layers_t.append(
+                    CrossTransformerEncoderLayer(**kwargs_cross_encoder)
+                )
+
+    def forward(self, x, xt):
+        B, C, Fr, T1 = x.shape
+        pos_emb_2d = create_2d_sin_embedding(
+            C, Fr, T1, x.device, self.max_period
+        )  # (1, C, Fr, T1)
+        pos_emb_2d = rearrange(pos_emb_2d, "b c fr t1 -> b (t1 fr) c")
+        x = rearrange(x, "b c fr t1 -> b (t1 fr) c")
+        x = self.norm_in(x)
+        x = x + self.weight_pos_embed * pos_emb_2d
+
+        B, C, T2 = xt.shape
+        xt = rearrange(xt, "b c t2 -> b t2 c")  # now T2, B, C
+        pos_emb = self._get_pos_embedding(T2, B, C, x.device)
+        pos_emb = rearrange(pos_emb, "t2 b c -> b t2 c")
+        xt = self.norm_in_t(xt)
+        xt = xt + self.weight_pos_embed * pos_emb
+
+        for idx in range(self.num_layers):
+            if idx % 2 == self.classic_parity:
+                x = self.layers[idx](x)
+                xt = self.layers_t[idx](xt)
+            else:
+                old_x = x
+                x = self.layers[idx](x, xt)
+                xt = self.layers_t[idx](xt, old_x)
+
+        x = rearrange(x, "b (t1 fr) c -> b c fr t1", t1=T1)
+        xt = rearrange(xt, "b t2 c -> b c t2")
+        return x, xt
+
+    def _get_pos_embedding(self, T, B, C, device):
+        if self.emb == "sin":
+            shift = random.randrange(self.sin_random_shift + 1)
+            pos_emb = create_sin_embedding(
+                T, C, shift=shift, device=device, max_period=self.max_period
+            )
+        elif self.emb == "cape":
+            if self.training:
+                pos_emb = create_sin_embedding_cape(
+                    T,
+                    C,
+                    B,
+                    device=device,
+                    max_period=self.max_period,
+                    mean_normalize=self.cape_mean_normalize,
+                    augment=self.cape_augment,
+                    max_global_shift=self.cape_glob_loc_scale[0],
+                    max_local_shift=self.cape_glob_loc_scale[1],
+                    max_scale=self.cape_glob_loc_scale[2],
+                )
+            else:
+                pos_emb = create_sin_embedding_cape(
+                    T,
+                    C,
+                    B,
+                    device=device,
+                    max_period=self.max_period,
+                    mean_normalize=self.cape_mean_normalize,
+                    augment=False,
+                )
+
+        elif self.emb == "scaled":
+            pos = torch.arange(T, device=device)
+            pos_emb = self.position_embeddings(pos)[:, None]
+
+        return pos_emb
+
+    def make_optim_group(self):
+        group = {"params": list(self.parameters()), "weight_decay": self.weight_decay}
+        if self.lr is not None:
+            group["lr"] = self.lr
+        return group
+
+
+# Attention Modules
+
+
+class MultiheadAttention(nn.Module):
+    def __init__(
+        self,
+        embed_dim,
+        num_heads,
+        dropout=0.0,
+        bias=True,
+        add_bias_kv=False,
+        add_zero_attn=False,
+        kdim=None,
+        vdim=None,
+        batch_first=False,
+        auto_sparsity=None,
+    ):
+        super().__init__()
+        assert auto_sparsity is not None, "sanity check"
+        self.num_heads = num_heads
+        self.q = torch.nn.Linear(embed_dim, embed_dim, bias=bias)
+        self.k = torch.nn.Linear(embed_dim, embed_dim, bias=bias)
+        self.v = torch.nn.Linear(embed_dim, embed_dim, bias=bias)
+        self.attn_drop = torch.nn.Dropout(dropout)
+        self.proj = torch.nn.Linear(embed_dim, embed_dim, bias)
+        self.proj_drop = torch.nn.Dropout(dropout)
+        self.batch_first = batch_first
+        self.auto_sparsity = auto_sparsity
+
+    def forward(
+        self,
+        query,
+        key,
+        value,
+        key_padding_mask=None,
+        need_weights=True,
+        attn_mask=None,
+        average_attn_weights=True,
+    ):
+
+        if not self.batch_first:  # N, B, C
+            query = query.permute(1, 0, 2)  # B, N_q, C
+            key = key.permute(1, 0, 2)  # B, N_k, C
+            value = value.permute(1, 0, 2)  # B, N_k, C
+        B, N_q, C = query.shape
+        B, N_k, C = key.shape
+
+        q = (
+            self.q(query)
+            .reshape(B, N_q, self.num_heads, C // self.num_heads)
+            .permute(0, 2, 1, 3)
+        )
+        q = q.flatten(0, 1)
+        k = (
+            self.k(key)
+            .reshape(B, N_k, self.num_heads, C // self.num_heads)
+            .permute(0, 2, 1, 3)
+        )
+        k = k.flatten(0, 1)
+        v = (
+            self.v(value)
+            .reshape(B, N_k, self.num_heads, C // self.num_heads)
+            .permute(0, 2, 1, 3)
+        )
+        v = v.flatten(0, 1)
+
+        if self.auto_sparsity:
+            assert attn_mask is None
+            x = dynamic_sparse_attention(q, k, v, sparsity=self.auto_sparsity)
+        else:
+            x = scaled_dot_product_attention(q, k, v, attn_mask, dropout=self.attn_drop)
+        x = x.reshape(B, self.num_heads, N_q, C // self.num_heads)
+
+        x = x.transpose(1, 2).reshape(B, N_q, C)
+        x = self.proj(x)
+        x = self.proj_drop(x)
+        if not self.batch_first:
+            x = x.permute(1, 0, 2)
+        return x, None
+
+
+def scaled_query_key_softmax(q, k, att_mask):
+    from xformers.ops import masked_matmul
+    q = q / (k.size(-1)) ** 0.5
+    att = masked_matmul(q, k.transpose(-2, -1), att_mask)
+    att = torch.nn.functional.softmax(att, -1)
+    return att
+
+
+def scaled_dot_product_attention(q, k, v, att_mask, dropout):
+    att = scaled_query_key_softmax(q, k, att_mask=att_mask)
+    att = dropout(att)
+    y = att @ v
+    return y
+
+
+def _compute_buckets(x, R):
+    qq = torch.einsum('btf,bfhi->bhti', x, R)
+    qq = torch.cat([qq, -qq], dim=-1)
+    buckets = qq.argmax(dim=-1)
+
+    return buckets.permute(0, 2, 1).byte().contiguous()
+
+
+def dynamic_sparse_attention(query, key, value, sparsity, infer_sparsity=True, attn_bias=None):
+    # assert False, "The code for the custom sparse kernel is not ready for release yet."
+    from xformers.ops import find_locations, sparse_memory_efficient_attention
+    n_hashes = 32
+    proj_size = 4
+    query, key, value = [x.contiguous() for x in [query, key, value]]
+    with torch.no_grad():
+        R = torch.randn(1, query.shape[-1], n_hashes, proj_size // 2, device=query.device)
+        bucket_query = _compute_buckets(query, R)
+        bucket_key = _compute_buckets(key, R)
+        row_offsets, column_indices = find_locations(
+            bucket_query, bucket_key, sparsity, infer_sparsity)
+    return sparse_memory_efficient_attention(
+        query, key, value, row_offsets, column_indices, attn_bias)

demucs/utils.py

@@ -0,0 +1,502 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+from collections import defaultdict
+from contextlib import contextmanager
+import math
+import os
+import tempfile
+import typing as tp
+
+import errno
+import functools
+import hashlib
+import inspect
+import io
+import os
+import random
+import socket
+import tempfile
+import warnings
+import zlib
+import tkinter as tk
+
+from diffq import UniformQuantizer, DiffQuantizer
+import torch as th
+import tqdm
+from torch import distributed
+from torch.nn import functional as F
+
+import torch
+
+def unfold(a, kernel_size, stride):
+    """Given input of size [*OT, T], output Tensor of size [*OT, F, K]
+    with K the kernel size, by extracting frames with the given stride.
+
+    This will pad the input so that `F = ceil(T / K)`.
+
+    see https://github.com/pytorch/pytorch/issues/60466
+    """
+    *shape, length = a.shape
+    n_frames = math.ceil(length / stride)
+    tgt_length = (n_frames - 1) * stride + kernel_size
+    a = F.pad(a, (0, tgt_length - length))
+    strides = list(a.stride())
+    assert strides[-1] == 1, 'data should be contiguous'
+    strides = strides[:-1] + [stride, 1]
+    return a.as_strided([*shape, n_frames, kernel_size], strides)
+
+
+def center_trim(tensor: torch.Tensor, reference: tp.Union[torch.Tensor, int]):
+    """
+    Center trim `tensor` with respect to `reference`, along the last dimension.
+    `reference` can also be a number, representing the length to trim to.
+    If the size difference != 0 mod 2, the extra sample is removed on the right side.
+    """
+    ref_size: int
+    if isinstance(reference, torch.Tensor):
+        ref_size = reference.size(-1)
+    else:
+        ref_size = reference
+    delta = tensor.size(-1) - ref_size
+    if delta < 0:
+        raise ValueError("tensor must be larger than reference. " f"Delta is {delta}.")
+    if delta:
+        tensor = tensor[..., delta // 2:-(delta - delta // 2)]
+    return tensor
+
+
+def pull_metric(history: tp.List[dict], name: str):
+    out = []
+    for metrics in history:
+        metric = metrics
+        for part in name.split("."):
+            metric = metric[part]
+        out.append(metric)
+    return out
+
+
+def EMA(beta: float = 1):
+    """
+    Exponential Moving Average callback.
+    Returns a single function that can be called to repeatidly update the EMA
+    with a dict of metrics. The callback will return
+    the new averaged dict of metrics.
+
+    Note that for `beta=1`, this is just plain averaging.
+    """
+    fix: tp.Dict[str, float] = defaultdict(float)
+    total: tp.Dict[str, float] = defaultdict(float)
+
+    def _update(metrics: dict, weight: float = 1) -> dict:
+        nonlocal total, fix
+        for key, value in metrics.items():
+            total[key] = total[key] * beta + weight * float(value)
+            fix[key] = fix[key] * beta + weight
+        return {key: tot / fix[key] for key, tot in total.items()}
+    return _update
+
+
+def sizeof_fmt(num: float, suffix: str = 'B'):
+    """
+    Given `num` bytes, return human readable size.
+    Taken from https://stackoverflow.com/a/1094933
+    """
+    for unit in ['', 'Ki', 'Mi', 'Gi', 'Ti', 'Pi', 'Ei', 'Zi']:
+        if abs(num) < 1024.0:
+            return "%3.1f%s%s" % (num, unit, suffix)
+        num /= 1024.0
+    return "%.1f%s%s" % (num, 'Yi', suffix)
+
+
+@contextmanager
+def temp_filenames(count: int, delete=True):
+    names = []
+    try:
+        for _ in range(count):
+            names.append(tempfile.NamedTemporaryFile(delete=False).name)
+        yield names
+    finally:
+        if delete:
+            for name in names:
+                os.unlink(name)
+
+def average_metric(metric, count=1.):
+    """
+    Average `metric` which should be a float across all hosts. `count` should be
+    the weight for this particular host (i.e. number of examples).
+    """
+    metric = th.tensor([count, count * metric], dtype=th.float32, device='cuda')
+    distributed.all_reduce(metric, op=distributed.ReduceOp.SUM)
+    return metric[1].item() / metric[0].item()
+
+
+def free_port(host='', low=20000, high=40000):
+    """
+    Return a port number that is most likely free.
+    This could suffer from a race condition although
+    it should be quite rare.
+    """
+    sock = socket.socket()
+    while True:
+        port = random.randint(low, high)
+        try:
+            sock.bind((host, port))
+        except OSError as error:
+            if error.errno == errno.EADDRINUSE:
+                continue
+            raise
+        return port
+
+
+def sizeof_fmt(num, suffix='B'):
+    """
+    Given `num` bytes, return human readable size.
+    Taken from https://stackoverflow.com/a/1094933
+    """
+    for unit in ['', 'Ki', 'Mi', 'Gi', 'Ti', 'Pi', 'Ei', 'Zi']:
+        if abs(num) < 1024.0:
+            return "%3.1f%s%s" % (num, unit, suffix)
+        num /= 1024.0
+    return "%.1f%s%s" % (num, 'Yi', suffix)
+
+
+def human_seconds(seconds, display='.2f'):
+    """
+    Given `seconds` seconds, return human readable duration.
+    """
+    value = seconds * 1e6
+    ratios = [1e3, 1e3, 60, 60, 24]
+    names = ['us', 'ms', 's', 'min', 'hrs', 'days']
+    last = names.pop(0)
+    for name, ratio in zip(names, ratios):
+        if value / ratio < 0.3:
+            break
+        value /= ratio
+        last = name
+    return f"{format(value, display)} {last}"
+
+
+class TensorChunk:
+    def __init__(self, tensor, offset=0, length=None):
+        total_length = tensor.shape[-1]
+        assert offset >= 0
+        assert offset < total_length
+
+        if length is None:
+            length = total_length - offset
+        else:
+            length = min(total_length - offset, length)
+
+        self.tensor = tensor
+        self.offset = offset
+        self.length = length
+        self.device = tensor.device
+
+    @property
+    def shape(self):
+        shape = list(self.tensor.shape)
+        shape[-1] = self.length
+        return shape
+
+    def padded(self, target_length):
+        delta = target_length - self.length
+        total_length = self.tensor.shape[-1]
+        assert delta >= 0
+
+        start = self.offset - delta // 2
+        end = start + target_length
+
+        correct_start = max(0, start)
+        correct_end = min(total_length, end)
+
+        pad_left = correct_start - start
+        pad_right = end - correct_end
+
+        out = F.pad(self.tensor[..., correct_start:correct_end], (pad_left, pad_right))
+        assert out.shape[-1] == target_length
+        return out
+
+
+def tensor_chunk(tensor_or_chunk):
+    if isinstance(tensor_or_chunk, TensorChunk):
+        return tensor_or_chunk
+    else:
+        assert isinstance(tensor_or_chunk, th.Tensor)
+        return TensorChunk(tensor_or_chunk)
+
+
+def apply_model_v1(model, mix, shifts=None, split=False, progress=False, set_progress_bar=None):
+    """
+    Apply model to a given mixture.
+
+    Args:
+        shifts (int): if > 0, will shift in time `mix` by a random amount between 0 and 0.5 sec
+            and apply the oppositve shift to the output. This is repeated `shifts` time and
+            all predictions are averaged. This effectively makes the model time equivariant
+            and improves SDR by up to 0.2 points.
+        split (bool): if True, the input will be broken down in 8 seconds extracts
+            and predictions will be performed individually on each and concatenated.
+            Useful for model with large memory footprint like Tasnet.
+        progress (bool): if True, show a progress bar (requires split=True)
+    """
+
+    channels, length = mix.size()
+    device = mix.device
+    progress_value = 0
+    
+    if split:
+        out = th.zeros(4, channels, length, device=device)
+        shift = model.samplerate * 10
+        offsets = range(0, length, shift)
+        scale = 10
+        if progress:
+            offsets = tqdm.tqdm(offsets, unit_scale=scale, ncols=120, unit='seconds')
+        for offset in offsets:
+            chunk = mix[..., offset:offset + shift]
+            if set_progress_bar:
+                progress_value += 1
+                set_progress_bar(0.1, (0.8/len(offsets)*progress_value))
+                chunk_out = apply_model_v1(model, chunk, shifts=shifts, set_progress_bar=set_progress_bar)
+            else:
+                chunk_out = apply_model_v1(model, chunk, shifts=shifts)
+            out[..., offset:offset + shift] = chunk_out
+            offset += shift
+        return out
+    elif shifts:
+        max_shift = int(model.samplerate / 2)
+        mix = F.pad(mix, (max_shift, max_shift))
+        offsets = list(range(max_shift))
+        random.shuffle(offsets)
+        out = 0
+        for offset in offsets[:shifts]:
+            shifted = mix[..., offset:offset + length + max_shift]
+            if set_progress_bar:
+                shifted_out = apply_model_v1(model, shifted, set_progress_bar=set_progress_bar)
+            else:
+                shifted_out = apply_model_v1(model, shifted)
+            out += shifted_out[..., max_shift - offset:max_shift - offset + length]
+        out /= shifts
+        return out
+    else:
+        valid_length = model.valid_length(length)
+        delta = valid_length - length
+        padded = F.pad(mix, (delta // 2, delta - delta // 2))
+        with th.no_grad():
+            out = model(padded.unsqueeze(0))[0]
+        return center_trim(out, mix)
+
+def apply_model_v2(model, mix, shifts=None, split=False,
+                overlap=0.25, transition_power=1., progress=False, set_progress_bar=None): 
+    """
+    Apply model to a given mixture.
+
+    Args:
+        shifts (int): if > 0, will shift in time `mix` by a random amount between 0 and 0.5 sec
+            and apply the oppositve shift to the output. This is repeated `shifts` time and
+            all predictions are averaged. This effectively makes the model time equivariant
+            and improves SDR by up to 0.2 points.
+        split (bool): if True, the input will be broken down in 8 seconds extracts
+            and predictions will be performed individually on each and concatenated.
+            Useful for model with large memory footprint like Tasnet.
+        progress (bool): if True, show a progress bar (requires split=True)
+    """
+    
+    assert transition_power >= 1, "transition_power < 1 leads to weird behavior."
+    device = mix.device
+    channels, length = mix.shape
+    progress_value = 0
+    
+    if split:
+        out = th.zeros(len(model.sources), channels, length, device=device)
+        sum_weight = th.zeros(length, device=device)
+        segment = model.segment_length
+        stride = int((1 - overlap) * segment)
+        offsets = range(0, length, stride)
+        scale = stride / model.samplerate
+        if progress:
+            offsets = tqdm.tqdm(offsets, unit_scale=scale, ncols=120, unit='seconds')
+        # We start from a triangle shaped weight, with maximal weight in the middle
+        # of the segment. Then we normalize and take to the power `transition_power`.
+        # Large values of transition power will lead to sharper transitions.
+        weight = th.cat([th.arange(1, segment // 2 + 1),
+                         th.arange(segment - segment // 2, 0, -1)]).to(device)
+        assert len(weight) == segment
+        # If the overlap < 50%, this will translate to linear transition when
+        # transition_power is 1.
+        weight = (weight / weight.max())**transition_power
+        for offset in offsets:
+            chunk = TensorChunk(mix, offset, segment)
+            if set_progress_bar:
+                progress_value += 1
+                set_progress_bar(0.1, (0.8/len(offsets)*progress_value))
+                chunk_out = apply_model_v2(model, chunk, shifts=shifts, set_progress_bar=set_progress_bar)
+            else:
+                chunk_out = apply_model_v2(model, chunk, shifts=shifts)
+            chunk_length = chunk_out.shape[-1]
+            out[..., offset:offset + segment] += weight[:chunk_length] * chunk_out
+            sum_weight[offset:offset + segment] += weight[:chunk_length]
+            offset += segment
+        assert sum_weight.min() > 0
+        out /= sum_weight
+        return out
+    elif shifts:
+        max_shift = int(0.5 * model.samplerate)
+        mix = tensor_chunk(mix)
+        padded_mix = mix.padded(length + 2 * max_shift)
+        out = 0
+        for _ in range(shifts):
+            offset = random.randint(0, max_shift)
+            shifted = TensorChunk(padded_mix, offset, length + max_shift - offset)
+            
+            if set_progress_bar:
+                progress_value += 1
+                shifted_out = apply_model_v2(model, shifted, set_progress_bar=set_progress_bar)
+            else:
+                shifted_out = apply_model_v2(model, shifted)
+            out += shifted_out[..., max_shift - offset:]
+        out /= shifts
+        return out
+    else:
+        valid_length = model.valid_length(length)
+        mix = tensor_chunk(mix)
+        padded_mix = mix.padded(valid_length)
+        with th.no_grad():
+            out = model(padded_mix.unsqueeze(0))[0]
+        return center_trim(out, length)
+
+
+@contextmanager
+def temp_filenames(count, delete=True):
+    names = []
+    try:
+        for _ in range(count):
+            names.append(tempfile.NamedTemporaryFile(delete=False).name)
+        yield names
+    finally:
+        if delete:
+            for name in names:
+                os.unlink(name)
+
+
+def get_quantizer(model, args, optimizer=None):
+    quantizer = None
+    if args.diffq:
+        quantizer = DiffQuantizer(
+            model, min_size=args.q_min_size, group_size=8)
+        if optimizer is not None:
+            quantizer.setup_optimizer(optimizer)
+    elif args.qat:
+        quantizer = UniformQuantizer(
+                model, bits=args.qat, min_size=args.q_min_size)
+    return quantizer
+
+
+def load_model(path, strict=False):
+    with warnings.catch_warnings():
+        warnings.simplefilter("ignore")
+        load_from = path
+        package = th.load(load_from, 'cpu')
+
+    klass = package["klass"]
+    args = package["args"]
+    kwargs = package["kwargs"]
+
+    if strict:
+        model = klass(*args, **kwargs)
+    else:
+        sig = inspect.signature(klass)
+        for key in list(kwargs):
+            if key not in sig.parameters:
+                warnings.warn("Dropping inexistant parameter " + key)
+                del kwargs[key]
+        model = klass(*args, **kwargs)
+
+    state = package["state"]
+    training_args = package["training_args"]
+    quantizer = get_quantizer(model, training_args)
+
+    set_state(model, quantizer, state)
+    return model
+
+
+def get_state(model, quantizer):
+    if quantizer is None:
+        state = {k: p.data.to('cpu') for k, p in model.state_dict().items()}
+    else:
+        state = quantizer.get_quantized_state()
+        buf = io.BytesIO()
+        th.save(state, buf)
+        state = {'compressed': zlib.compress(buf.getvalue())}
+    return state
+
+
+def set_state(model, quantizer, state):
+    if quantizer is None:
+        model.load_state_dict(state)
+    else:
+        buf = io.BytesIO(zlib.decompress(state["compressed"]))
+        state = th.load(buf, "cpu")
+        quantizer.restore_quantized_state(state)
+
+    return state
+
+
+def save_state(state, path):
+    buf = io.BytesIO()
+    th.save(state, buf)
+    sig = hashlib.sha256(buf.getvalue()).hexdigest()[:8]
+
+    path = path.parent / (path.stem + "-" + sig + path.suffix)
+    path.write_bytes(buf.getvalue())
+
+
+def save_model(model, quantizer, training_args, path):
+    args, kwargs = model._init_args_kwargs
+    klass = model.__class__
+
+    state = get_state(model, quantizer)
+
+    save_to = path
+    package = {
+        'klass': klass,
+        'args': args,
+        'kwargs': kwargs,
+        'state': state,
+        'training_args': training_args,
+    }
+    th.save(package, save_to)
+
+
+def capture_init(init):
+    @functools.wraps(init)
+    def __init__(self, *args, **kwargs):
+        self._init_args_kwargs = (args, kwargs)
+        init(self, *args, **kwargs)
+
+    return __init__
+
+class DummyPoolExecutor:
+    class DummyResult:
+        def __init__(self, func, *args, **kwargs):
+            self.func = func
+            self.args = args
+            self.kwargs = kwargs
+
+        def result(self):
+            return self.func(*self.args, **self.kwargs)
+
+    def __init__(self, workers=0):
+        pass
+
+    def submit(self, func, *args, **kwargs):
+        return DummyPoolExecutor.DummyResult(func, *args, **kwargs)
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, exc_type, exc_value, exc_tb):
+        return