TheBloke/llama2_7b_chat_uncensored-AWQ
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

Update for Transformers AWQ support

Browse Source
This commit is contained in:
Tom Jobbins
2023-11-09 18:20:20 +00:00
committed by huggingface-web
parent 6ac73a5842
commit 9be86ac7e8
1 changed files with 29 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files

51
config.json
View File

@@ -1,24 +1,31 @@
{ {
"_name_or_path": "TheBloke/Llama-2-7B-fp16", "_name_or_path": "TheBloke/Llama-2-7B-fp16",
"architectures": [ "architectures": [
"LlamaForCausalLM" "LlamaForCausalLM"
], ],
"bos_token_id": 1, "bos_token_id": 1,
"eos_token_id": 2, "eos_token_id": 2,
"hidden_act": "silu", "hidden_act": "silu",
"hidden_size": 4096, "hidden_size": 4096,
"initializer_range": 0.02, "initializer_range": 0.02,
"intermediate_size": 11008, "intermediate_size": 11008,
"max_position_embeddings": 2048, "max_position_embeddings": 2048,
"model_type": "llama", "model_type": "llama",
"num_attention_heads": 32, "num_attention_heads": 32,
"num_hidden_layers": 32, "num_hidden_layers": 32,
"pad_token_id": 0, "pad_token_id": 0,
"rms_norm_eps": 1e-05, "rms_norm_eps": 1e-05,
"tie_word_embeddings": false, "tie_word_embeddings": false,
"torch_dtype": "float32", "torch_dtype": "float32",
"transformers_version": "4.30.2", "transformers_version": "4.30.2",
"use_cache": true, "use_cache": true,
"vocab_size": 32000, "vocab_size": 32000,
"pretraining_tp": 1 "pretraining_tp": 1,
"quantization_config": {
"quant_method": "awq",
"zero_point": true,
"group_size": 128,
"bits": 4,
"version": "gemm"
}
} }