fix naming
Browse files- config_s2a.json +69 -0
- s2a.ckpt +3 -0
config_s2a.json
ADDED
|
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"saving_path": "/home/ubuntu/experiments/a2s_giga2",
|
| 3 |
+
"resume_checkpoint": null,
|
| 4 |
+
"vocoder_type": "SPEECHTOKENIZER",
|
| 5 |
+
"vocoder_config_path": null,
|
| 6 |
+
"vocoder_ckpt_path": null,
|
| 7 |
+
"metapath": [
|
| 8 |
+
"/home/ubuntu/data/poly/giga-training-data/train.json"
|
| 9 |
+
],
|
| 10 |
+
"val_metapath": [
|
| 11 |
+
"/home/ubuntu/data/poly/giga-training-data/dev.json"
|
| 12 |
+
],
|
| 13 |
+
"pretrained_path": null,
|
| 14 |
+
"speaker_embedding_dir": null,
|
| 15 |
+
"sampledir": "/home/ubuntu/experiments/a2s_giga2",
|
| 16 |
+
"lr": 0.0005,
|
| 17 |
+
"batch_size": 400.0,
|
| 18 |
+
"train_bucket_size": 8192,
|
| 19 |
+
"training_step": 800000,
|
| 20 |
+
"optim_flat_percent": 0.0,
|
| 21 |
+
"warmup_step": 10000,
|
| 22 |
+
"adam_beta1": 0.9,
|
| 23 |
+
"adam_beta2": 0.98,
|
| 24 |
+
"ffd_size": 1024,
|
| 25 |
+
"hidden_size": 768,
|
| 26 |
+
"enc_nlayers": 3,
|
| 27 |
+
"dec_nlayers": 6,
|
| 28 |
+
"nheads": 8,
|
| 29 |
+
"dropout": 0.1,
|
| 30 |
+
"depthwise_conv_kernel_size": 5,
|
| 31 |
+
"aligner_softmax_temp": 1.0,
|
| 32 |
+
"layer_norm_eps": 1e-05,
|
| 33 |
+
"use_sem_tokens": true,
|
| 34 |
+
"use_spkr_emb": true,
|
| 35 |
+
"use_text_emb": false,
|
| 36 |
+
"fairseq": false,
|
| 37 |
+
"only_inference": false,
|
| 38 |
+
"speaker_embed_dropout": 0.05,
|
| 39 |
+
"label_smoothing": 0.0,
|
| 40 |
+
"val_check_interval": 1,
|
| 41 |
+
"max_dataset_samples": -1,
|
| 42 |
+
"check_val_every_n_epoch": 1,
|
| 43 |
+
"precision": "bf16",
|
| 44 |
+
"nworkers": 12,
|
| 45 |
+
"distributed": true,
|
| 46 |
+
"accelerator": "gpu",
|
| 47 |
+
"version": null,
|
| 48 |
+
"accumulate_grad_batches": 1,
|
| 49 |
+
"sagemaker": false,
|
| 50 |
+
"use_repetition_token": false,
|
| 51 |
+
"use_repetition_gating": false,
|
| 52 |
+
"repetition_penalty": 1.0,
|
| 53 |
+
"sampling_temperature": 1.0,
|
| 54 |
+
"top_k": -1,
|
| 55 |
+
"min_top_k": 3,
|
| 56 |
+
"top_p": 0.8,
|
| 57 |
+
"sample_num": 4,
|
| 58 |
+
"length_penalty_max_length": 150,
|
| 59 |
+
"length_penalty_max_prob": 0.95,
|
| 60 |
+
"max_input_length": 2048,
|
| 61 |
+
"max_output_length": 2000,
|
| 62 |
+
"phone_context_window": 3,
|
| 63 |
+
"sample_rate": 16000,
|
| 64 |
+
"n_codes": 1024,
|
| 65 |
+
"n_cluster_groups": 7,
|
| 66 |
+
"first_n_lvls": 7,
|
| 67 |
+
"use_pretrained_ckpt_cfg": false,
|
| 68 |
+
"n_semantic_codes": 1024
|
| 69 |
+
}
|
s2a.ckpt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f083b82b7a2e902cf318163562f6b3faa2d2b4ce72d20b92e3de8b8dfc125383
|
| 3 |
+
size 671800067
|