Upload folder using huggingface_hub
Browse files- README.md +17 -0
- config.json +57 -0
- lfm/config.json +57 -0
- lfm/model.safetensors +3 -0
- model.safetensors +3 -0
- pytorch_model.bin +3 -0
- training_state.pt +3 -0
README.md
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# LFM2-Audio Arabic TTS
|
| 2 |
+
|
| 3 |
+
## Loading
|
| 4 |
+
```python
|
| 5 |
+
from liquid_audio import LFM2AudioModel, LFM2AudioProcessor
|
| 6 |
+
import torch
|
| 7 |
+
|
| 8 |
+
# Load base model and processor
|
| 9 |
+
processor = LFM2AudioProcessor.from_pretrained("LiquidAI/LFM2-Audio-1.5B")
|
| 10 |
+
model = LFM2AudioModel.from_pretrained("LiquidAI/LFM2-Audio-1.5B").eval().cuda()
|
| 11 |
+
|
| 12 |
+
# Load fine-tuned weights
|
| 13 |
+
state_dict = torch.load("pytorch_model.bin")
|
| 14 |
+
model.load_state_dict(state_dict)
|
| 15 |
+
```
|
| 16 |
+
|
| 17 |
+
Training: Stage 1 (Depthformer only), 10 epochs, Audio loss: 0.058
|
config.json
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Lfm2Model"
|
| 4 |
+
],
|
| 5 |
+
"block_auto_adjust_ff_dim": true,
|
| 6 |
+
"block_dim": 2048,
|
| 7 |
+
"block_ff_dim": 12288,
|
| 8 |
+
"block_ffn_dim_multiplier": 1,
|
| 9 |
+
"block_mlp_init_scale": 1,
|
| 10 |
+
"block_multiple_of": 256,
|
| 11 |
+
"block_norm_eps": 1e-05,
|
| 12 |
+
"block_out_init_scale": 1,
|
| 13 |
+
"block_use_swiglu": true,
|
| 14 |
+
"block_use_xavier_init": true,
|
| 15 |
+
"bos_token_id": 1,
|
| 16 |
+
"conv_L_cache": 3,
|
| 17 |
+
"conv_bias": false,
|
| 18 |
+
"conv_dim": 2048,
|
| 19 |
+
"conv_dim_out": 2048,
|
| 20 |
+
"conv_use_xavier_init": true,
|
| 21 |
+
"dtype": "bfloat16",
|
| 22 |
+
"eos_token_id": 7,
|
| 23 |
+
"hidden_size": 2048,
|
| 24 |
+
"initializer_range": 0.02,
|
| 25 |
+
"intermediate_size": 12288,
|
| 26 |
+
"layer_types": [
|
| 27 |
+
"conv",
|
| 28 |
+
"conv",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"conv",
|
| 31 |
+
"conv",
|
| 32 |
+
"full_attention",
|
| 33 |
+
"conv",
|
| 34 |
+
"conv",
|
| 35 |
+
"full_attention",
|
| 36 |
+
"conv",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"conv",
|
| 39 |
+
"full_attention",
|
| 40 |
+
"conv",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"conv"
|
| 43 |
+
],
|
| 44 |
+
"max_position_embeddings": 128000,
|
| 45 |
+
"model_type": "lfm2",
|
| 46 |
+
"norm_eps": 1e-05,
|
| 47 |
+
"num_attention_heads": 32,
|
| 48 |
+
"num_heads": 32,
|
| 49 |
+
"num_hidden_layers": 16,
|
| 50 |
+
"num_key_value_heads": 8,
|
| 51 |
+
"pad_token_id": 0,
|
| 52 |
+
"rope_theta": 1000000,
|
| 53 |
+
"transformers_version": "4.57.3",
|
| 54 |
+
"use_cache": true,
|
| 55 |
+
"use_pos_enc": true,
|
| 56 |
+
"vocab_size": 65536
|
| 57 |
+
}
|
lfm/config.json
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Lfm2Model"
|
| 4 |
+
],
|
| 5 |
+
"block_auto_adjust_ff_dim": true,
|
| 6 |
+
"block_dim": 2048,
|
| 7 |
+
"block_ff_dim": 12288,
|
| 8 |
+
"block_ffn_dim_multiplier": 1,
|
| 9 |
+
"block_mlp_init_scale": 1,
|
| 10 |
+
"block_multiple_of": 256,
|
| 11 |
+
"block_norm_eps": 1e-05,
|
| 12 |
+
"block_out_init_scale": 1,
|
| 13 |
+
"block_use_swiglu": true,
|
| 14 |
+
"block_use_xavier_init": true,
|
| 15 |
+
"bos_token_id": 1,
|
| 16 |
+
"conv_L_cache": 3,
|
| 17 |
+
"conv_bias": false,
|
| 18 |
+
"conv_dim": 2048,
|
| 19 |
+
"conv_dim_out": 2048,
|
| 20 |
+
"conv_use_xavier_init": true,
|
| 21 |
+
"dtype": "bfloat16",
|
| 22 |
+
"eos_token_id": 7,
|
| 23 |
+
"hidden_size": 2048,
|
| 24 |
+
"initializer_range": 0.02,
|
| 25 |
+
"intermediate_size": 12288,
|
| 26 |
+
"layer_types": [
|
| 27 |
+
"conv",
|
| 28 |
+
"conv",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"conv",
|
| 31 |
+
"conv",
|
| 32 |
+
"full_attention",
|
| 33 |
+
"conv",
|
| 34 |
+
"conv",
|
| 35 |
+
"full_attention",
|
| 36 |
+
"conv",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"conv",
|
| 39 |
+
"full_attention",
|
| 40 |
+
"conv",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"conv"
|
| 43 |
+
],
|
| 44 |
+
"max_position_embeddings": 128000,
|
| 45 |
+
"model_type": "lfm2",
|
| 46 |
+
"norm_eps": 1e-05,
|
| 47 |
+
"num_attention_heads": 32,
|
| 48 |
+
"num_heads": 32,
|
| 49 |
+
"num_hidden_layers": 16,
|
| 50 |
+
"num_key_value_heads": 8,
|
| 51 |
+
"pad_token_id": 0,
|
| 52 |
+
"rope_theta": 1000000,
|
| 53 |
+
"transformers_version": "4.57.3",
|
| 54 |
+
"use_cache": true,
|
| 55 |
+
"use_pos_enc": true,
|
| 56 |
+
"vocab_size": 65536
|
| 57 |
+
}
|
lfm/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:84e0ba68a67d4d1ad2e4ec8eea771758b00992017e543cfe7d5cfcb60b648fe1
|
| 3 |
+
size 2340697048
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:84e0ba68a67d4d1ad2e4ec8eea771758b00992017e543cfe7d5cfcb60b648fe1
|
| 3 |
+
size 2340697048
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:51d2fbbe61007e1bdec8960180e0da756a820bca22cea7cd2ad305dd1dfc78bd
|
| 3 |
+
size 2907408335
|
training_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d65906fe5294c57051eb3fc3649a8103fdc801a054994a41240cf9b4f376d76c
|
| 3 |
+
size 404961569
|