dendimaki commited on
Commit
6f9b093
1 Parent(s): 2d7fe55

Upload config

Browse files
Files changed (1) hide show
  1. config.json +2 -18
config.json CHANGED
@@ -1,7 +1,7 @@
1
  {
2
  "_name_or_path": "./mistralai/Mistral-7B-v0.1",
3
  "architectures": [
4
- "MistralForCausalLM"
5
  ],
6
  "attention_dropout": 0.0,
7
  "bos_token_id": 1,
@@ -77,28 +77,12 @@
77
  "num_attention_heads": 32,
78
  "num_hidden_layers": 32,
79
  "num_key_value_heads": 8,
80
- "pad_token_id": 2,
81
- "quantization_config": {
82
- "_load_in_4bit": true,
83
- "_load_in_8bit": false,
84
- "bnb_4bit_compute_dtype": "bfloat16",
85
- "bnb_4bit_quant_storage": "uint8",
86
- "bnb_4bit_quant_type": "nf4",
87
- "bnb_4bit_use_double_quant": true,
88
- "llm_int8_enable_fp32_cpu_offload": false,
89
- "llm_int8_has_fp16_weight": false,
90
- "llm_int8_skip_modules": null,
91
- "llm_int8_threshold": 6.0,
92
- "load_in_4bit": true,
93
- "load_in_8bit": false,
94
- "quant_method": "bitsandbytes"
95
- },
96
  "rms_norm_eps": 1e-05,
97
  "rope_theta": 10000.0,
98
  "sliding_window": 4096,
99
  "tie_word_embeddings": false,
100
  "torch_dtype": "bfloat16",
101
  "transformers_version": "4.42.0.dev0",
102
- "use_cache": false,
103
  "vocab_size": 32000
104
  }
 
1
  {
2
  "_name_or_path": "./mistralai/Mistral-7B-v0.1",
3
  "architectures": [
4
+ "MistralForSequenceClassification"
5
  ],
6
  "attention_dropout": 0.0,
7
  "bos_token_id": 1,
 
77
  "num_attention_heads": 32,
78
  "num_hidden_layers": 32,
79
  "num_key_value_heads": 8,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
80
  "rms_norm_eps": 1e-05,
81
  "rope_theta": 10000.0,
82
  "sliding_window": 4096,
83
  "tie_word_embeddings": false,
84
  "torch_dtype": "bfloat16",
85
  "transformers_version": "4.42.0.dev0",
86
+ "use_cache": true,
87
  "vocab_size": 32000
88
  }