Upload 2 files
Browse files- config.json +6 -5
config.json
CHANGED
|
@@ -1,4 +1,5 @@
|
|
| 1 |
{
|
|
|
|
| 2 |
"architectures": [
|
| 3 |
"MoonshineForConditionalGeneration"
|
| 4 |
],
|
|
@@ -8,21 +9,21 @@
|
|
| 8 |
"pad_token_id": 2,
|
| 9 |
"decoder_hidden_act": "silu",
|
| 10 |
"decoder_num_attention_heads": 8,
|
| 11 |
-
"decoder_num_hidden_layers":
|
| 12 |
"decoder_num_key_value_heads": 8,
|
| 13 |
"decoder_start_token_id": 1,
|
| 14 |
"encoder_hidden_act": "gelu",
|
| 15 |
"encoder_num_attention_heads": 8,
|
| 16 |
-
"encoder_num_hidden_layers":
|
| 17 |
"encoder_num_key_value_heads": 8,
|
| 18 |
"eos_token_id": 2,
|
| 19 |
-
"hidden_size":
|
| 20 |
"initializer_range": 0.02,
|
| 21 |
-
"intermediate_size":
|
| 22 |
"is_encoder_decoder": true,
|
| 23 |
"max_position_embeddings": 194,
|
| 24 |
"model_type": "moonshine",
|
| 25 |
-
"partial_rotary_factor": 0.
|
| 26 |
"rope_scaling": null,
|
| 27 |
"rope_theta": 10000.0,
|
| 28 |
"torch_dtype": "float32",
|
|
|
|
| 1 |
{
|
| 2 |
+
"_name_or_path": "/home/eustache_lebihan/dev/add-moonshine/moonshine-base",
|
| 3 |
"architectures": [
|
| 4 |
"MoonshineForConditionalGeneration"
|
| 5 |
],
|
|
|
|
| 9 |
"pad_token_id": 2,
|
| 10 |
"decoder_hidden_act": "silu",
|
| 11 |
"decoder_num_attention_heads": 8,
|
| 12 |
+
"decoder_num_hidden_layers": 8,
|
| 13 |
"decoder_num_key_value_heads": 8,
|
| 14 |
"decoder_start_token_id": 1,
|
| 15 |
"encoder_hidden_act": "gelu",
|
| 16 |
"encoder_num_attention_heads": 8,
|
| 17 |
+
"encoder_num_hidden_layers": 8,
|
| 18 |
"encoder_num_key_value_heads": 8,
|
| 19 |
"eos_token_id": 2,
|
| 20 |
+
"hidden_size": 416,
|
| 21 |
"initializer_range": 0.02,
|
| 22 |
+
"intermediate_size": 1664,
|
| 23 |
"is_encoder_decoder": true,
|
| 24 |
"max_position_embeddings": 194,
|
| 25 |
"model_type": "moonshine",
|
| 26 |
+
"partial_rotary_factor": 0.62,
|
| 27 |
"rope_scaling": null,
|
| 28 |
"rope_theta": 10000.0,
|
| 29 |
"torch_dtype": "float32",
|