cleaning up

.gitignore

@@ -177,4 +177,6 @@ samples-*/**
 
 gradio-outputs/
 models/
-samples*/
+samples*/
+models-all/
+models.zip

README.md

@@ -10,7 +10,6 @@ install [audiotools](https://github.com/descriptinc/audiotools.git)
 ```bash
 git clone https://github.com/descriptinc/audiotools.git
 cd audiotools
-git checkout f35914b5b3c6f1bf589cd09481478d741538828e
 pip install -e .
 ```
 
@@ -51,7 +50,7 @@ source ./env/env.sh
 You can launch a gradio UI to play with vampnet. 
 
 ```bash
-python demo.py --args.load conf/interface/spotdl.yml --Interface.device cuda
+python demo.py --args.load conf/interface.yml --Interface.device cuda
 ```
 
 # Training / Fine-tuning 

conf/{interface/maestro.yml → interface.yml}

@@ -1,10 +1,10 @@
-Interface.coarse_ckpt: ./models/maestro/coarse.pth
-Interface.coarse2fine_ckpt: ./models/maestro/c2f.pth
-Interface.codec_ckpt: ./models/maestro/codec.pth
-Interface.coarse_chunk_size_s: 12
+Interface.coarse_ckpt: ./models/vampnet/coarse.pth
+Interface.coarse2fine_ckpt: ./models/vampnet/c2f.pth
+Interface.codec_ckpt: ./models/vampnet/codec.pth
+Interface.coarse_chunk_size_s: 10
 Interface.coarse2fine_chunk_size_s: 3
 Interface.wavebeat_ckpt: ./models/wavebeat.pth
 
-
 AudioLoader.sources:
-  - /data/maestro-reorg/test
+  - /media/CHONK/null
+

conf/interface/jazzpop.yml

@@ -1,10 +0,0 @@
-Interface.coarse_ckpt: ./models/jazzpop/coarse.pth
-Interface.coarse2fine_ckpt: ./models/jazzpop/c2f.pth
-Interface.codec_ckpt: ./models/jazzpop/codec.pth
-Interface.coarse_chunk_size_s: 5
-Interface.coarse2fine_chunk_size_s: 3
-Interface.wavebeat_ckpt: ./models/wavebeat.pth
-
-AudioLoader.sources:
-  - /data/spotdl-jazzpop/audio/val
-  - /data/spotdl-jazzpop/audio/test

conf/interface/spotdl.yml

@@ -1,12 +0,0 @@
-Interface.coarse_ckpt: ./models/spotdl/coarse.pth
-Interface.coarse2fine_ckpt: ./models/spotdl/c2f.pth
-Interface.codec_ckpt: ./models/spotdl/codec.pth
-Interface.coarse_chunk_size_s: 10
-Interface.coarse2fine_chunk_size_s: 3
-# Interface.wavebeat_ckpt: ./models/wavebeat.pth
-
-
-AudioLoader.sources:
-  # - /media/CHONK/hugo/spotdl/subsets/jazz-blues/
-  - /media/CHONK/null
-

conf/interface/xeno-canto.yml

@@ -1,11 +0,0 @@
-Interface.coarse_ckpt: ./runs/xeno-canto-2/coarse/best/vampnet/weights.pth
-Interface.coarse2fine_ckpt: ./runs/xeno-canto-2/c2f/best/vampnet/weights.pth
-Interface.codec_ckpt: ./models/spotdl/codec.pth
-Interface.coarse_chunk_size_s: 10
-Interface.coarse2fine_chunk_size_s: 3
-# Interface.wavebeat_ckpt: ./models/wavebeat.pth
-
-
-AudioLoader.sources:
-  - /media/CHONK/hugo/xeno-canto-2
-  - /media/CHONK/hugo/xeno-canto-2

demo.py

@@ -33,56 +33,14 @@ dataset = at.data.datasets.AudioDataset(
 
 
 checkpoints = {
-    "spotdl": {
-        "coarse": "./models/spotdl/coarse.pth",
-        "c2f": "./models/spotdl/c2f.pth",
-        "codec": "./models/spotdl/codec.pth",
+    "vampnet": {
+        "coarse": "./models/vampnet/coarse.pth",
+        "c2f": "./models/vampnet/c2f.pth",
+        "codec": "./models/vampnet/codec.pth",
         "full_ckpt": True
     }, 
-    "berta": {
-        "coarse": "./models/finetuned/berta-goldman-speech/coarse.pth",
-        "c2f": "./models/finetuned/berta-goldman-speech/c2f.pth",
-        "codec": "./model/spotdl/codec.pth",
-        "full_ckpt": True
-    }, 
-    "xeno-canto-2": {
-        "coarse": "./models/finetuned/xeno-canto-2/coarse.pth",
-        "c2f": "./models/finetuned/xeno-canto-2/c2f.pth",
-        "codec": "./models/spotdl/codec.pth",
-        "full_ckpt": True
-    }, 
-    "panchos": {
-        "coarse": "./models/finetuned/panchos/coarse.pth",
-        "c2f": "./models/finetuned/panchos/c2f.pth",
-        "codec": "./models/spotdl/codec.pth",
-        "full_ckpt": False
-    },
-    "tv-choir": {
-        "coarse": "./models/finetuned/tv-choir/coarse.pth",
-        "c2f": "./models/finetuned/tv-choir/c2f.pth",
-        "codec": "./models/spotdl/codec.pth",
-        "full_ckpt": False
-    }, 
-    "titi": {
-        "coarse": "./models/finetuned/titi/coarse.pth",
-        "c2f": "./models/finetuned/titi/c2f.pth",
-        "codec": "./models/spotdl/codec.pth",
-        "full_ckpt": False
-    }, 
-    "titi-clean": {
-        "coarse": "./models/finetuned/titi-clean/coarse.pth",
-        "c2f": "./models/finetuned/titi-clean/c2f.pth",
-        "codec": "./models/spotdl/codec.pth",
-        "full_ckpt": False
-    }, 
-    "breaks-steps": {
-        "coarse": "./models/finetuned/breaks-steps/coarse.pth",
-        "c2f": None, #"./models/finetuned/breaks-steps/c2f.pth",
-        "codec": "./models/spotdl/codec.pth",
-        "full_ckpt": False
-    },
 }
-interface.checkpoint_key = "spotdl"
+interface.checkpoint_key = "vampnet"
 
 
 OUT_DIR = Path("gradio-outputs")