default model

README.md

@@ -167,9 +167,9 @@ Other pre-trained models can be selected with the `-n` flag.
 The list of pre-trained models is:
 - `mdx`: trained only on MusDB HQ, winning model on track A at the [MDX][mdx] challenge.
 - `mdx_extra`: trained with extra training data (including MusDB test set), ranked 2nd on the track B
-    of the [MDX][mdx] challenge. This is the default one.
-- `mdx_q, mdx_extra_q`: quantized version of the previous models. Smaller download and storage
-    but quality can be slightly worse.
+    of the [MDX][mdx] challenge.
+- `mdx_q`, `mdx_extra_q`: quantized version of the previous models. Smaller download and storage
+    but quality can be slightly worse. `mdx_extra_q` is the default model used.
 - `SIG`: where `SIG` is a single model from the [model zoo](docs/training.md#model-zoo).
 
 The `--shifts=SHIFTS` performs multiple predictions with random shifts (a.k.a the *shift trick*) of the input and average them. This makes prediction `SHIFTS` times
@@ -221,4 +221,4 @@ Demucs is released under the MIT license as found in the [LICENSE](LICENSE) file
 [mdx]: https://www.aicrowd.com/challenges/music-demixing-challenge-ismir-2021
 [kuielab]: https://github.com/kuielab/mdx-net-submission
 [decouple]: https://arxiv.org/abs/2109.05418
-[mdx_submission]: https://github.com/adefossez/mdx21_demucs
+[mdx_submission]: https://github.com/adefossez/mdx21_demucs

conf/variant/finetune.yaml

@@ -15,3 +15,5 @@ augment:
   scale:
     proba: 0
   shift_same: true
+  remix:
+    proba: 0

demucs/pretrained.py

@@ -30,18 +30,17 @@ def demucs_unittest():
 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",
-                       help="Pretrained model name or signature. Default is mdx_extra.")
+    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 get_model(name: tp.Optional[str] = None,
+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.
     """
-    assert name is not None
     if name == 'demucs_unittest':
         return demucs_unittest()
     model_repo: ModelOnlyRepo

demucs/remote/files.txt

@@ -10,3 +10,7 @@ e51eebcc-c1b80bdd.th
 b72baf4e-8778635e.th
 42e558d4-196e0e1b.th
 305bc58f-18378783.th
+14fc6a69-a89dd0ee.th
+464b36d7-e5a9386e.th
+7fd6ef75-a905dd85.th
+83fc094f-4a16d450.th

docs/windows.md

@@ -14,15 +14,14 @@ git clone -b main --single-branch https://github.com/facebookresearch/demucs ./d
 cd ./demucs
 conda env update -f environment-cpu.yml
 conda activate demucs
+python.exe -m pip install -e .
 demucs -d cpu "PATH_TO_AUDIO_FILE_1" ["PATH_TO_AUDIO_FILE_2" ...]
 ```
 
 The `"` around the filename are required if the path contains spaces.
 The separated files will be under `C:\Users\YOUR_USERNAME\demucs\separated\demucs\`. The next time you want to use Demucs, start again the [Anaconda prompt][prompt] and simply run
 ```bash
-cd %HOMEPATH%
 conda activate demucs
-cd demucs
 demucs -d cpu "PATH_TO_AUDIO_FILE_1" ...
 ```
 
@@ -34,6 +33,7 @@ cd %HOMEPATH%
 cd demucs
 git pull
 conda env update -f environment-cpu.yml
+python.exe -m pip install -e .
 ```
 
 ## Potential errors