Push model using huggingface_hub.

config.json

@@ -1,5 +1,5 @@
 {
-  "audio_embedding_dim": 2048,
+  "audio_embedding_dim": 1024,
   "audio_embedding_dropout": 0,
   "audio_max_length": 20.0,
   "audio_min_length": 2.0,
@@ -9,7 +9,7 @@
   "batch_size": 100,
   "clipping_update_period": 1000,
   "codebook_weight": "[5,1,0.5,0.1]",
-  "d_model": 2048,
+  "d_model": 1024,
   "dataset": "gigaspeech",
   "dataset_dir": "/data/scratch/pyp/datasets/gigaspeech_phn_enc_manifest/xl",
   "drop_long": 1,
@@ -21,11 +21,11 @@
   "encodec_sr": 50,
   "eog": 2049,
   "eos": 2051,
-  "exp_dir": "/data/scratch/pyp/exp_pyp/VoiceCraft/gigaspeech/e5",
-  "gradient_accumulation_steps": 8,
+  "exp_dir": "/data/scratch/pyp/exp_pyp/VoiceCraft/gigaspeech/e7",
+  "gradient_accumulation_steps": 10,
   "gradient_clip_val": 1.0,
   "load_model_from": null,
-  "lr": 0.03,
+  "lr": 0.05,
   "manifest_name": "manifest_large16khz_lessambi",
   "mask_len_max": 600,
   "mask_len_min": 1,
@@ -38,7 +38,7 @@
   "n_special": 4,
   "nhead": 16,
   "num_buckets": 6,
-  "num_decoder_layers": 8,
+  "num_decoder_layers": 24,
   "num_epochs": 10,
   "num_steps": 50000,
   "num_workers": 8,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fc33980ce676aadeb20272a449fe06dcb6ddae19286f7b14501f75156c3fa8ea
-size 1746809632
+oid sha256:44330087bbcf64b3669187c247abe3c5e2c0760be4894c83a548446027785f83
+size 1293771496