Upload sft_checkpoints/sft_v2/config.json with huggingface_hub
Browse files
sft_checkpoints/sft_v2/config.json
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"ViuAIForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"model_type": "viuai",
|
| 6 |
+
"vocab_size": 64003,
|
| 7 |
+
"hidden_size": 1280,
|
| 8 |
+
"intermediate_size": 3456,
|
| 9 |
+
"num_hidden_layers": 24,
|
| 10 |
+
"num_attention_heads": 20,
|
| 11 |
+
"num_key_value_heads": 4,
|
| 12 |
+
"max_position_embeddings": 2048,
|
| 13 |
+
"rms_norm_eps": 1e-05,
|
| 14 |
+
"rope_theta": 10000.0,
|
| 15 |
+
"tie_word_embeddings": true,
|
| 16 |
+
"bos_token_id": 64000,
|
| 17 |
+
"eos_token_id": 64002,
|
| 18 |
+
"pad_token_id": 64002,
|
| 19 |
+
"torch_dtype": "bfloat16"
|
| 20 |
+
}
|