Upload config
Browse files- config.json +25 -0
config.json
ADDED
|
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "kimi-k2",
|
| 3 |
+
"vocab_size": 50304,
|
| 4 |
+
"n_layer": 12,
|
| 5 |
+
"n_head": 12,
|
| 6 |
+
"n_embd": 768,
|
| 7 |
+
"dropout": 0.0,
|
| 8 |
+
"bias": true,
|
| 9 |
+
"use_moe": true,
|
| 10 |
+
"num_experts": 8,
|
| 11 |
+
"expert_capacity": 32,
|
| 12 |
+
"top_k_experts": 2,
|
| 13 |
+
"use_latent_attention": true,
|
| 14 |
+
"latent_dim": 64,
|
| 15 |
+
"block_size": 256,
|
| 16 |
+
"tokenizer": "tiktoken/gpt2",
|
| 17 |
+
"architectures": [
|
| 18 |
+
"KimiK2ForCausalLM"
|
| 19 |
+
],
|
| 20 |
+
"torch_dtype": "float32",
|
| 21 |
+
"auto_map": {
|
| 22 |
+
"AutoConfig": "modeling_kimik2.KimiK2Config",
|
| 23 |
+
"AutoModelForCausalLM": "modeling_kimik2.KimiK2ForCausalLM"
|
| 24 |
+
}
|
| 25 |
+
}
|