Add iOS ONNX base model pack
Browse files- 500m-v3/config.json +64 -0
- 500m-v3/context_kv.onnx +3 -0
- 500m-v3/dacvae_decode.onnx +3 -0
- 500m-v3/dacvae_encode.onnx +3 -0
- 500m-v3/dit_step.onnx +3 -0
- 500m-v3/duration.onnx +3 -0
- 500m-v3/speaker_encoder.onnx +3 -0
- 500m-v3/text_encoder.onnx +3 -0
- 500m-v3/tokenizer/tokenizer.json +0 -0
500m-v3/config.json
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": {
|
| 3 |
+
"latent_dim": 32,
|
| 4 |
+
"latent_patch_size": 1,
|
| 5 |
+
"model_dim": 1280,
|
| 6 |
+
"num_layers": 12,
|
| 7 |
+
"num_heads": 20,
|
| 8 |
+
"mlp_ratio": 2.875,
|
| 9 |
+
"text_mlp_ratio": 2.6,
|
| 10 |
+
"speaker_mlp_ratio": 2.6,
|
| 11 |
+
"dropout": 0.0,
|
| 12 |
+
"text_vocab_size": 99574,
|
| 13 |
+
"text_tokenizer_repo": "llm-jp/llm-jp-3-150m",
|
| 14 |
+
"text_add_bos": true,
|
| 15 |
+
"text_dim": 512,
|
| 16 |
+
"text_layers": 10,
|
| 17 |
+
"text_heads": 8,
|
| 18 |
+
"use_caption_condition": false,
|
| 19 |
+
"use_speaker_condition": null,
|
| 20 |
+
"caption_vocab_size": null,
|
| 21 |
+
"caption_tokenizer_repo": null,
|
| 22 |
+
"caption_add_bos": null,
|
| 23 |
+
"caption_dim": null,
|
| 24 |
+
"caption_layers": null,
|
| 25 |
+
"caption_heads": null,
|
| 26 |
+
"caption_mlp_ratio": null,
|
| 27 |
+
"speaker_dim": 768,
|
| 28 |
+
"speaker_layers": 8,
|
| 29 |
+
"speaker_heads": 12,
|
| 30 |
+
"speaker_patch_size": 1,
|
| 31 |
+
"timestep_embed_dim": 512,
|
| 32 |
+
"adaln_rank": 192,
|
| 33 |
+
"norm_eps": 1e-05,
|
| 34 |
+
"use_duration_predictor": true,
|
| 35 |
+
"duration_aux_dim": 14,
|
| 36 |
+
"duration_hidden_dim": 1024,
|
| 37 |
+
"duration_layers": 3,
|
| 38 |
+
"duration_dropout": 0.1,
|
| 39 |
+
"duration_attention_heads": 8,
|
| 40 |
+
"duration_architecture": "token_sum_adarn_zero_no_aux",
|
| 41 |
+
"duration_token_init_frames": 9.0,
|
| 42 |
+
"duration_speaker_fusion": "adarn_zero",
|
| 43 |
+
"duration_caption_fusion": "adarn_zero",
|
| 44 |
+
"duration_caption_pooling": "masked_mean"
|
| 45 |
+
},
|
| 46 |
+
"inference": {
|
| 47 |
+
"max_text_len": 256
|
| 48 |
+
},
|
| 49 |
+
"export": {
|
| 50 |
+
"seq_len": 128,
|
| 51 |
+
"text_len": 256,
|
| 52 |
+
"caption_len": 64,
|
| 53 |
+
"ref_len": 200,
|
| 54 |
+
"sample_rate": 48000,
|
| 55 |
+
"hop_length": 1920,
|
| 56 |
+
"latent_dim": 32,
|
| 57 |
+
"num_layers": 12,
|
| 58 |
+
"has_caption": false,
|
| 59 |
+
"has_speaker_kv_scale": true,
|
| 60 |
+
"lora_merged": false,
|
| 61 |
+
"dacvae_snake_coreml_safe": true,
|
| 62 |
+
"context_kv_cache": true
|
| 63 |
+
}
|
| 64 |
+
}
|
500m-v3/context_kv.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:58727ae357463d004c660e4c4c3f7e9e8607244fde0027ca6ce1aa885bd7026d
|
| 3 |
+
size 157419444
|
500m-v3/dacvae_decode.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8a48e87e347601cb5eb6ba7dfbdc3aafa586e588448980c5678670fd74b25715
|
| 3 |
+
size 261377318
|
500m-v3/dacvae_encode.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:43f5d220fd3126a39803d3c779953449a566edab555ed76f0ea16dc0c2a013a9
|
| 3 |
+
size 109422044
|
500m-v3/dit_step.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:79cd707d198afec190d67947dd8042e7b52be65ca7c0fbed26904060d264856a
|
| 3 |
+
size 1243778849
|
500m-v3/duration.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a9ba94bdc981c0caf1753fab68e5ce64106096addeb5c5bb165de11aebfaffcb
|
| 3 |
+
size 68260372
|
500m-v3/speaker_encoder.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2a076114f8d104e850133752043ae6808d792dfc634657094dc40c35cba885eb
|
| 3 |
+
size 242582289
|
500m-v3/text_encoder.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5ef033d48e31e12c8eb5fdad79d2ff866ccac3366aecf91036fb1d4e3cddad64
|
| 3 |
+
size 339121171
|
500m-v3/tokenizer/tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|