Commit
•
accd4d1
1
Parent(s):
8fb3642
Upload config.json
Browse files- config.json +45 -0
config.json
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "bigcode/starcoder2-3b",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"Starcoder2ForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"attention_dropout": 0.1,
|
| 7 |
+
"bos_token_id": 0,
|
| 8 |
+
"embedding_dropout": 0.1,
|
| 9 |
+
"eos_token_id": 0,
|
| 10 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 11 |
+
"hidden_size": 3072,
|
| 12 |
+
"initializer_range": 0.018042,
|
| 13 |
+
"intermediate_size": 12288,
|
| 14 |
+
"max_position_embeddings": 16384,
|
| 15 |
+
"mlp_type": "default",
|
| 16 |
+
"model_type": "starcoder2",
|
| 17 |
+
"norm_epsilon": 1e-05,
|
| 18 |
+
"norm_type": "layer_norm",
|
| 19 |
+
"num_attention_heads": 24,
|
| 20 |
+
"num_hidden_layers": 30,
|
| 21 |
+
"num_key_value_heads": 2,
|
| 22 |
+
"quantization_config": {
|
| 23 |
+
"_load_in_4bit": true,
|
| 24 |
+
"_load_in_8bit": false,
|
| 25 |
+
"bnb_4bit_compute_dtype": "bfloat16",
|
| 26 |
+
"bnb_4bit_quant_storage": "uint8",
|
| 27 |
+
"bnb_4bit_quant_type": "nf4",
|
| 28 |
+
"bnb_4bit_use_double_quant": false,
|
| 29 |
+
"llm_int8_enable_fp32_cpu_offload": false,
|
| 30 |
+
"llm_int8_has_fp16_weight": false,
|
| 31 |
+
"llm_int8_skip_modules": null,
|
| 32 |
+
"llm_int8_threshold": 6.0,
|
| 33 |
+
"load_in_4bit": true,
|
| 34 |
+
"load_in_8bit": false,
|
| 35 |
+
"quant_method": "bitsandbytes"
|
| 36 |
+
},
|
| 37 |
+
"residual_dropout": 0.1,
|
| 38 |
+
"rope_theta": 999999.4420358813,
|
| 39 |
+
"sliding_window": 4096,
|
| 40 |
+
"torch_dtype": "float32",
|
| 41 |
+
"transformers_version": "4.40.0.dev0",
|
| 42 |
+
"use_bias": true,
|
| 43 |
+
"use_cache": true,
|
| 44 |
+
"vocab_size": 49152
|
| 45 |
+
}
|