Hugging Face's logo

medmekk
/
test-mlx-quantized

Text Generation
Transformers
Safetensors
MLX
llama
conversational
text-generation-inference
4-bit precision
Model card Files
xet
 Community
medmekk commited on
Commit
69506bf
·
verified ·
1 Parent(s): 031a782

Upload LlamaForCausalLM

Browse files
Files changed (3) hide show
  1. config.json +5 -1
  2. generation_config.json +5 -2
  3. model.safetensors +1 -1
config.json CHANGED
@@ -6,7 +6,11 @@
6
  "attention_dropout": 0.0,
7
  "bos_token_id": 128000,
8
  "dtype": "bfloat16",
9
- "eos_token_id": 128001,
 
 
 
 
10
  "head_dim": 64,
11
  "hidden_act": "silu",
12
  "hidden_size": 2048,
 
6
  "attention_dropout": 0.0,
7
  "bos_token_id": 128000,
8
  "dtype": "bfloat16",
9
+ "eos_token_id": [
10
+ 128001,
11
+ 128008,
12
+ 128009
13
+ ],
14
  "head_dim": 64,
15
  "hidden_act": "silu",
16
  "hidden_size": 2048,
generation_config.json CHANGED
@@ -1,8 +1,11 @@
1
  {
2
- "_from_model_config": true,
3
  "bos_token_id": 128000,
4
  "do_sample": true,
5
- "eos_token_id": 128001,
 
 
 
 
6
  "temperature": 0.6,
7
  "top_p": 0.9,
8
  "transformers_version": "5.2.0.dev0"
 
1
  {
 
2
  "bos_token_id": 128000,
3
  "do_sample": true,
4
+ "eos_token_id": [
5
+ 128001,
6
+ 128008,
7
+ 128009
8
+ ],
9
  "temperature": 0.6,
10
  "top_p": 0.9,
11
  "transformers_version": "5.2.0.dev0"
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:29677f1896a521195af71a72b811d36b6a62bccec8252b46e0267437037c15ae
3
  size 1042461648
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1a03f1a9fbbebe6c170a4aab4a4b131f5e0791e5259645f237b9ddcf28bd0422
3
  size 1042461648