Add model configuration
Browse files- config.json +26 -0
config.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen2ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"model_type": "qwen2",
|
| 6 |
+
"torch_dtype": "bfloat16",
|
| 7 |
+
"transformers_version": "4.44.2",
|
| 8 |
+
"vocab_size": 151936,
|
| 9 |
+
"use_cache": true,
|
| 10 |
+
"rope_theta": 1000000.0,
|
| 11 |
+
"max_position_embeddings": 131072,
|
| 12 |
+
"thinking_tokens": 512000,
|
| 13 |
+
"_name_or_path": "zenlm/zen-coder-480b-instruct",
|
| 14 |
+
"_base_model": "Qwen/Qwen2.5-Coder-32B-Instruct",
|
| 15 |
+
"hidden_size": 5120,
|
| 16 |
+
"num_hidden_layers": 64,
|
| 17 |
+
"num_attention_heads": 40,
|
| 18 |
+
"num_key_value_heads": 8,
|
| 19 |
+
"intermediate_size": 27648,
|
| 20 |
+
"num_experts": 16,
|
| 21 |
+
"num_experts_per_tok": 2,
|
| 22 |
+
"expert_interval": 1,
|
| 23 |
+
"_architecture_type": "moe",
|
| 24 |
+
"_total_params": "480B",
|
| 25 |
+
"_active_params": "30B"
|
| 26 |
+
}
|