Upload 5 files
Browse files- .gitattributes +1 -0
- checkpoints/bestval.pth +3 -0
- events.out.tfevents.1731345469.ozen.905748.0 +3 -0
- features.json +11 -0
- mo.bestval +3 -0
- tiny_topk.json +36 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
mo.bestval filter=lfs diff=lfs merge=lfs -text
|
checkpoints/bestval.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bfbd788acef162e40a8f978ad44368fefb55c5d2aded39a901cbbe7d2ac4ef46
|
| 3 |
+
size 56707430
|
events.out.tfevents.1731345469.ozen.905748.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4f4d877935f31e6cce0c092306a9a607a08c1468c456c5f4430a17f0c19cfbb3
|
| 3 |
+
size 8301804
|
features.json
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"whisper_model": "tiny",
|
| 3 |
+
"sae_model": "pretrained/topkautoencoder_baseline/checkpoints/bestval.pth",
|
| 4 |
+
"layer_name": "encoder.blocks.2",
|
| 5 |
+
"batch_size": 160,
|
| 6 |
+
"data_path": "audio_data/librispeech/LibriSpeech/test-other",
|
| 7 |
+
"device": "cuda",
|
| 8 |
+
"out_folder": "activations/pretrained_topk_baseline",
|
| 9 |
+
"dl_max_workers": 0,
|
| 10 |
+
"collect_max": null
|
| 11 |
+
}
|
mo.bestval
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:08965768f4e83fff7c160cc239e9bf7ec4ba5a5fbfbac0037e9298337b65b8fb
|
| 3 |
+
size 18903336
|
tiny_topk.json
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"whisper_config": {
|
| 3 |
+
"model": "tiny",
|
| 4 |
+
"layer_name": "encoder.blocks.2"
|
| 5 |
+
},
|
| 6 |
+
"autoencoder_variant": "topk",
|
| 7 |
+
"autoencoder_config": {
|
| 8 |
+
"expansion_factor": 16,
|
| 9 |
+
"normalize_decoder": true,
|
| 10 |
+
"k": 32,
|
| 11 |
+
"multi_topk": false,
|
| 12 |
+
"auxk_alpha": 0.03125,
|
| 13 |
+
"dead_feature_threshold": 1000000.0
|
| 14 |
+
},
|
| 15 |
+
"seed": 0,
|
| 16 |
+
"train_folder": "activations/tiny_block_2_train",
|
| 17 |
+
"val_folder": "activations/tiny_block_2_dev",
|
| 18 |
+
"device": "cuda",
|
| 19 |
+
"run_dir": "runs/topkautoencoder_baseline",
|
| 20 |
+
"lr": 1e-4,
|
| 21 |
+
"weight_decay": 0.0,
|
| 22 |
+
"steps": 100000,
|
| 23 |
+
"clip_thresh": 1.0,
|
| 24 |
+
"batch_size": 50,
|
| 25 |
+
"dl_max_workers": 0,
|
| 26 |
+
"log_tb_every": 10,
|
| 27 |
+
"save_every": 1000,
|
| 28 |
+
"val_every": 100,
|
| 29 |
+
"optimizer": "adam",
|
| 30 |
+
"scheduler": "linear",
|
| 31 |
+
"scheduler_params": {
|
| 32 |
+
"num_warmup_steps": 1000
|
| 33 |
+
},
|
| 34 |
+
"start_checkpoint": null,
|
| 35 |
+
"from_disk": true
|
| 36 |
+
}
|