iqrakiran commited on
Commit
683aa85
1 Parent(s): 21600d5

Upload config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +50 -0
config.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "ybelkada/falcon-7b-sharded-bf16",
3
+ "activation": "gelu",
4
+ "alibi": false,
5
+ "apply_residual_connection_post_layernorm": false,
6
+ "architectures": [
7
+ "FalconForCausalLM"
8
+ ],
9
+ "attention_dropout": 0.0,
10
+ "bias": false,
11
+ "bos_token_id": 11,
12
+ "eos_token_id": 11,
13
+ "ffn_hidden_size": 18176,
14
+ "hidden_dropout": 0.0,
15
+ "hidden_size": 4544,
16
+ "initializer_range": 0.02,
17
+ "layer_norm_epsilon": 1e-05,
18
+ "max_position_embeddings": 2048,
19
+ "model_type": "falcon",
20
+ "multi_query": true,
21
+ "n_head": 71,
22
+ "n_layer": 32,
23
+ "new_decoder_architecture": false,
24
+ "num_attention_heads": 71,
25
+ "num_hidden_layers": 32,
26
+ "num_kv_heads": 71,
27
+ "num_ln_in_parallel_attn": null,
28
+ "parallel_attn": true,
29
+ "quantization_config": {
30
+ "_load_in_4bit": true,
31
+ "_load_in_8bit": false,
32
+ "bnb_4bit_compute_dtype": "float16",
33
+ "bnb_4bit_quant_storage": "uint8",
34
+ "bnb_4bit_quant_type": "nf4",
35
+ "bnb_4bit_use_double_quant": false,
36
+ "llm_int8_enable_fp32_cpu_offload": false,
37
+ "llm_int8_has_fp16_weight": false,
38
+ "llm_int8_skip_modules": null,
39
+ "llm_int8_threshold": 6.0,
40
+ "load_in_4bit": true,
41
+ "load_in_8bit": false,
42
+ "quant_method": "bitsandbytes"
43
+ },
44
+ "rope_scaling": null,
45
+ "rope_theta": 10000.0,
46
+ "torch_dtype": "float32",
47
+ "transformers_version": "4.44.2",
48
+ "use_cache": true,
49
+ "vocab_size": 65024
50
+ }