Create finetune_mms_optimized.json
Browse files- finetune_mms_optimized.json +43 -0
finetune_mms_optimized.json
ADDED
|
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"project_name": "mms_gujarati_finetuning",
|
| 3 |
+
"push_to_hub": true,
|
| 4 |
+
"hub_model_id": "Cossale/mms-tts-guj-train",
|
| 5 |
+
"report_to": [
|
| 6 |
+
"wandb"
|
| 7 |
+
],
|
| 8 |
+
"overwrite_output_dir": true,
|
| 9 |
+
"output_dir": "./tmp/vits_finetuned_guj",
|
| 10 |
+
"dataset_name": "Cossale/gujarati-interspeech",
|
| 11 |
+
"dataset_config_name": "default",
|
| 12 |
+
"audio_column_name": "audio",
|
| 13 |
+
"text_column_name": "text",
|
| 14 |
+
"train_split_name": "train",
|
| 15 |
+
"speaker_id_column_name": "speaker_id",
|
| 16 |
+
"override_speaker_embeddings": true,
|
| 17 |
+
"filter_on_speaker_id": 1,
|
| 18 |
+
"full_generation_sample_text": "આજે દુબલીમાં ગરમી છે, શું આશ્ચર્ય છે",
|
| 19 |
+
"max_duration_in_seconds": 20,
|
| 20 |
+
"min_duration_in_seconds": 1.0,
|
| 21 |
+
"max_tokens_length": 500,
|
| 22 |
+
"model_name_or_path": "Cossale/mms-tts-guj-train",
|
| 23 |
+
"preprocessing_num_workers": 40,
|
| 24 |
+
"do_train": true,
|
| 25 |
+
"num_train_epochs": 100,
|
| 26 |
+
"gradient_accumulation_steps": 2,
|
| 27 |
+
"gradient_checkpointing": true,
|
| 28 |
+
"per_device_train_batch_size": 96,
|
| 29 |
+
"learning_rate": 3e-5,
|
| 30 |
+
"adam_beta1": 0.8,
|
| 31 |
+
"adam_beta2": 0.99,
|
| 32 |
+
"warmup_ratio": 0.03,
|
| 33 |
+
"group_by_length": true,
|
| 34 |
+
"do_step_schedule_per_epoch": true,
|
| 35 |
+
"weight_disc": 3,
|
| 36 |
+
"weight_fmaps": 1,
|
| 37 |
+
"weight_gen": 1,
|
| 38 |
+
"weight_kl": 1.5,
|
| 39 |
+
"weight_duration": 1,
|
| 40 |
+
"weight_mel": 35,
|
| 41 |
+
"fp16": true,
|
| 42 |
+
"seed": 456
|
| 43 |
+
}
|