Upload config.json
Browse files- config.json +21 -0
config.json
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": ["LlamaForCausalLM"],
|
| 3 |
+
"model_type": "llama",
|
| 4 |
+
"hidden_size": 4096,
|
| 5 |
+
"num_attention_heads": 32,
|
| 6 |
+
"num_hidden_layers": 32,
|
| 7 |
+
"vocab_size": 32000,
|
| 8 |
+
"max_position_embeddings": 2048,
|
| 9 |
+
"embedding_size": 4096,
|
| 10 |
+
"intermediate_size": 11008,
|
| 11 |
+
"layer_norm_eps": 1e-6,
|
| 12 |
+
"initializer_range": 0.02,
|
| 13 |
+
"pad_token_id": 0,
|
| 14 |
+
"bos_token_id": 1,
|
| 15 |
+
"eos_token_id": 2,
|
| 16 |
+
"num_trainable_params": 3407872,
|
| 17 |
+
"total_params": 8033669120,
|
| 18 |
+
"training_loss": 1.15442467674613,
|
| 19 |
+
"epoch": 0.14262791941522554
|
| 20 |
+
}
|
| 21 |
+
|