Hugging Face's logo

GreenBitAI
/
Qwen-2.5-3B-Instruct-layer-mix-bpw-4.0-mlx

MLX
Safetensors
qwen2
Model card Files Community
1

Upload folder using huggingface_hub

#1
by NicoNico - opened
base: refs/heads/main
←
from: refs/pr/1
Discussion Files changed
+459919
-0
Files changed (11) hide show
  1. README.md +24 -0
  2. added_tokens.json +24 -0
  3. config.json +85 -0
  4. merges.txt +0 -0
  5. model.safetensors +3 -0
  6. model.safetensors.index.json +0 -0
  7. quant_strategy.json +3424 -0
  8. special_tokens_map.json +31 -0
  9. tokenizer.json +0 -0
  10. tokenizer_config.json +207 -0
  11. vocab.json +0 -0
README.md ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: GreenBitAI/Qwen-2.5-3B-Instruct-layer-mix-bpw-4.0
3
+ license: apache-2.0
4
+ tags:
5
+ - mlx
6
+ ---
7
+
8
+ # GreenBitAI/Qwen-2.5-3B-Instruct-layer-mix-bpw-4.0-mlx
9
+
10
+ This quantized low-bit model [GreenBitAI/Qwen-2.5-3B-Instruct-layer-mix-bpw-4.0-mlx](https://huggingface.co/GreenBitAI/Qwen-2.5-3B-Instruct-layer-mix-bpw-4.0-mlx) was converted to MLX format from [`GreenBitAI/Qwen-2.5-3B-Instruct-layer-mix-bpw-4.0`](https://huggingface.co/GreenBitAI/Qwen-2.5-3B-Instruct-layer-mix-bpw-4.0) using gbx-lm version **0.3.4**.
11
+ Refer to the [original model card](https://huggingface.co/GreenBitAI/Qwen-2.5-3B-Instruct-layer-mix-bpw-4.0) for more details on the model.
12
+
13
+ ## Use with mlx
14
+
15
+ ```bash
16
+ pip install gbx-lm
17
+ ```
18
+
19
+ ```python
20
+ from gbx_lm import load, generate
21
+
22
+ model, tokenizer = load("GreenBitAI/Qwen-2.5-3B-Instruct-layer-mix-bpw-4.0-mlx")
23
+ response = generate(model, tokenizer, prompt="hello", verbose=True)
24
+ ```
added_tokens.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</tool_call>": 151658,
3
+ "<tool_call>": 151657,
4
+ "<|box_end|>": 151649,
5
+ "<|box_start|>": 151648,
6
+ "<|endoftext|>": 151643,
7
+ "<|file_sep|>": 151664,
8
+ "<|fim_middle|>": 151660,
9
+ "<|fim_pad|>": 151662,
10
+ "<|fim_prefix|>": 151659,
11
+ "<|fim_suffix|>": 151661,
12
+ "<|im_end|>": 151645,
13
+ "<|im_start|>": 151644,
14
+ "<|image_pad|>": 151655,
15
+ "<|object_ref_end|>": 151647,
16
+ "<|object_ref_start|>": 151646,
17
+ "<|quad_end|>": 151651,
18
+ "<|quad_start|>": 151650,
19
+ "<|repo_name|>": 151663,
20
+ "<|video_pad|>": 151656,
21
+ "<|vision_end|>": 151653,
22
+ "<|vision_pad|>": 151654,
23
+ "<|vision_start|>": 151652
24
+ }
config.json ADDED
@@ -0,0 +1,85 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_cross_attention": false,
3
+ "architectures": [
4
+ "Qwen2ForCausalLM"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bad_words_ids": null,
8
+ "begin_suppress_tokens": null,
9
+ "bos_token_id": 151643,
10
+ "chunk_size_feed_forward": 0,
11
+ "cross_attention_hidden_size": null,
12
+ "decoder_start_token_id": null,
13
+ "diversity_penalty": 0.0,
14
+ "do_sample": false,
15
+ "early_stopping": false,
16
+ "encoder_no_repeat_ngram_size": 0,
17
+ "eos_token_id": 151645,
18
+ "exponential_decay_length_penalty": null,
19
+ "finetuning_task": null,
20
+ "forced_bos_token_id": null,
21
+ "forced_eos_token_id": null,
22
+ "hidden_act": "silu",
23
+ "hidden_size": 2048,
24
+ "id2label": {
25
+ "0": "LABEL_0",
26
+ "1": "LABEL_1"
27
+ },
28
+ "initializer_range": 0.02,
29
+ "intermediate_size": 11008,
30
+ "is_decoder": false,
31
+ "is_encoder_decoder": false,
32
+ "label2id": {
33
+ "LABEL_0": 0,
34
+ "LABEL_1": 1
35
+ },
36
+ "length_penalty": 1.0,
37
+ "max_length": 20,
38
+ "max_position_embeddings": 32768,
39
+ "max_window_layers": 70,
40
+ "min_length": 0,
41
+ "model_type": "qwen2",
42
+ "no_repeat_ngram_size": 0,
43
+ "num_attention_heads": 16,
44
+ "num_beam_groups": 1,
45
+ "num_beams": 1,
46
+ "num_hidden_layers": 36,
47
+ "num_key_value_heads": 2,
48
+ "num_return_sequences": 1,
49
+ "output_attentions": false,
50
+ "output_hidden_states": false,
51
+ "output_scores": false,
52
+ "pad_token_id": null,
53
+ "prefix": null,
54
+ "problem_type": null,
55
+ "pruned_heads": {},
56
+ "quantization": {
57
+ "group_size": null,
58
+ "bits": null
59
+ },
60
+ "remove_invalid_values": false,
61
+ "repetition_penalty": 1.0,
62
+ "return_dict": true,
63
+ "return_dict_in_generate": false,
64
+ "rms_norm_eps": 1e-06,
65
+ "rope_theta": 1000000.0,
66
+ "sep_token_id": null,
67
+ "sliding_window": null,
68
+ "suppress_tokens": null,
69
+ "task_specific_params": null,
70
+ "temperature": 1.0,
71
+ "tf_legacy_loss": false,
72
+ "tie_encoder_decoder": false,
73
+ "tie_word_embeddings": true,
74
+ "tokenizer_class": null,
75
+ "top_k": 50,
76
+ "top_p": 1.0,
77
+ "torch_dtype": "float16",
78
+ "torchscript": false,
79
+ "transformers_version": "4.44.0",
80
+ "typical_p": 1.0,
81
+ "use_bfloat16": false,
82
+ "use_cache": true,
83
+ "use_sliding_window": false,
84
+ "vocab_size": 151936
85
+ }
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7dcaabcc407811b1185348b4c491f3241ff5f46a217fa1e61e9565313a9bc7c5
3
+ size 2099383813
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
quant_strategy.json ADDED
@@ -0,0 +1,3424 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "measurement": {
3
+ "model.layers.0": {
4
+ "accuracy": 0.9470930099487305,
5
+ "total_bits": 314302464.0,
6
+ "o_proj": {
7
+ "group_size": {
8
+ "4": 128
9
+ },
10
+ "bits": [
11
+ 4
12
+ ],
13
+ "bits_prop": [
14
+ 1
15
+ ],
16
+ "scale_bits": 4,
17
+ "scale_groups:": 32
18
+ },
19
+ "down_proj": {
20
+ "group_size": {
21
+ "4": 128
22
+ },
23
+ "bits": [
24
+ 4
25
+ ],
26
+ "bits_prop": [
27
+ 1
28
+ ],
29
+ "scale_bits": 4,
30
+ "scale_groups:": 32
31
+ },
32
+ "q_proj": {
33
+ "group_size": {
34
+ "4": 128
35
+ },
36
+ "bits": [
37
+ 4
38
+ ],
39
+ "bits_prop": [
40
+ 1
41
+ ],
42
+ "scale_bits": 4,
43
+ "scale_groups:": 32
44
+ },
45
+ "k_proj": {
46
+ "group_size": {
47
+ "4": 128
48
+ },
49
+ "bits": [
50
+ 4
51
+ ],
52
+ "bits_prop": [
53
+ 1
54
+ ],
55
+ "scale_bits": 4,
56
+ "scale_groups:": 32
57
+ },
58
+ "v_proj": {
59
+ "group_size": {
60
+ "4": 128
61
+ },
62
+ "bits": [
63
+ 4
64
+ ],
65
+ "bits_prop": [
66
+ 1
67
+ ],
68
+ "scale_bits": 4,
69
+ "scale_groups:": 32
70
+ },
71
+ "gate_proj": {
72
+ "group_size": {
73
+ "4": 128
74
+ },
75
+ "bits": [
76
+ 4
77
+ ],
78
+ "bits_prop": [
79
+ 1
80
+ ],
81
+ "scale_bits": 4,
82
+ "scale_groups:": 32
83
+ },
84
+ "up_proj": {
85
+ "group_size": {
86
+ "4": 128
87
+ },
88
+ "bits": [
89
+ 4
90
+ ],
91
+ "bits_prop": [
92
+ 1
93
+ ],
94
+ "scale_bits": 4,
95
+ "scale_groups:": 32
96
+ }
97
+ },
98
+ "model.layers.1": {
99
+ "accuracy": 0.44930458068847656,
100
+ "total_bits": 306241536.0,
101
+ "o_proj": {
102
+ "group_size": {
103
+ "4": 128
104
+ },
105
+ "bits": [
106
+ 4
107
+ ],
108
+ "bits_prop": [
109
+ 1
110
+ ],
111
+ "scale_bits": 4,
112
+ "scale_groups:": 32
113
+ },
114
+ "down_proj": {
115
+ "group_size": {
116
+ "4": 128
117
+ },
118
+ "bits": [
119
+ 4
120
+ ],
121
+ "bits_prop": [
122
+ 1
123
+ ],
124
+ "scale_bits": 4,
125
+ "scale_groups:": 32
126
+ },
127
+ "q_proj": {
128
+ "group_size": {
129
+ "2": 64
130
+ },
131
+ "bits": [
132
+ 2
133
+ ],
134
+ "bits_prop": [
135
+ 1
136
+ ],
137
+ "scale_bits": 4,
138
+ "scale_groups:": 32
139
+ },
140
+ "k_proj": {
141
+ "group_size": {
142
+ "4": 128
143
+ },
144
+ "bits": [
145
+ 4
146
+ ],
147
+ "bits_prop": [
148
+ 1
149
+ ],
150
+ "scale_bits": 4,
151
+ "scale_groups:": 32
152
+ },
153
+ "v_proj": {
154
+ "group_size": {
155
+ "4": 128
156
+ },
157
+ "bits": [
158
+ 4
159
+ ],
160
+ "bits_prop": [
161
+ 1
162
+ ],
163
+ "scale_bits": 4,
164
+ "scale_groups:": 32
165
+ },
166
+ "gate_proj": {
167
+ "group_size": {
168
+ "4": 128
169
+ },
170
+ "bits": [
171
+ 4
172
+ ],
173
+ "bits_prop": [
174
+ 1
175
+ ],
176
+ "scale_bits": 4,
177
+ "scale_groups:": 32
178
+ },
179
+ "up_proj": {
180
+ "group_size": {
181
+ "4": 128
182
+ },
183
+ "bits": [
184
+ 4
185
+ ],
186
+ "bits_prop": [
187
+ 1
188
+ ],
189
+ "scale_bits": 4,
190
+ "scale_groups:": 32
191
+ }
192
+ },
193
+ "model.layers.2": {
194
+ "accuracy": 0.9330713748931885,
195
+ "total_bits": 314302464.0,
196
+ "o_proj": {
197
+ "group_size": {
198
+ "4": 128
199
+ },
200
+ "bits": [
201
+ 4
202
+ ],
203
+ "bits_prop": [
204
+ 1
205
+ ],
206
+ "scale_bits": 4,
207
+ "scale_groups:": 32
208
+ },
209
+ "down_proj": {
210
+ "group_size": {
211
+ "4": 128
212
+ },
213
+ "bits": [
214
+ 4
215
+ ],
216
+ "bits_prop": [
217
+ 1
218
+ ],
219
+ "scale_bits": 4,
220
+ "scale_groups:": 32
221
+ },
222
+ "q_proj": {
223
+ "group_size": {
224
+ "4": 128
225
+ },
226
+ "bits": [
227
+ 4
228
+ ],
229
+ "bits_prop": [
230
+ 1
231
+ ],
232
+ "scale_bits": 4,
233
+ "scale_groups:": 32
234
+ },
235
+ "k_proj": {
236
+ "group_size": {
237
+ "4": 128
238
+ },
239
+ "bits": [
240
+ 4
241
+ ],
242
+ "bits_prop": [
243
+ 1
244
+ ],
245
+ "scale_bits": 4,
246
+ "scale_groups:": 32
247
+ },
248
+ "v_proj": {
249
+ "group_size": {
250
+ "4": 128
251
+ },
252
+ "bits": [
253
+ 4
254
+ ],
255
+ "bits_prop": [
256
+ 1
257
+ ],
258
+ "scale_bits": 4,
259
+ "scale_groups:": 32
260
+ },
261
+ "gate_proj": {
262
+ "group_size": {
263
+ "4": 128
264
+ },
265
+ "bits": [
266
+ 4
267
+ ],
268
+ "bits_prop": [
269
+ 1
270
+ ],
271
+ "scale_bits": 4,
272
+ "scale_groups:": 32
273
+ },
274
+ "up_proj": {
275
+ "group_size": {
276
+ "4": 128
277
+ },
278
+ "bits": [
279
+ 4
280
+ ],
281
+ "bits_prop": [
282
+ 1
283
+ ],
284
+ "scale_bits": 4,
285
+ "scale_groups:": 32
286
+ }
287
+ },
288
+ "model.layers.3": {
289
+ "accuracy": 0.970420241355896,
290
+ "total_bits": 314302464.0,
291
+ "o_proj": {
292
+ "group_size": {
293
+ "4": 128
294
+ },
295
+ "bits": [
296
+ 4
297
+ ],
298
+ "bits_prop": [
299
+ 1
300
+ ],
301
+ "scale_bits": 4,
302
+ "scale_groups:": 32
303
+ },
304
+ "down_proj": {
305
+ "group_size": {
306
+ "4": 128
307
+ },
308
+ "bits": [
309
+ 4
310
+ ],
311
+ "bits_prop": [
312
+ 1
313
+ ],
314
+ "scale_bits": 4,
315
+ "scale_groups:": 32
316
+ },
317
+ "q_proj": {
318
+ "group_size": {
319
+ "4": 128
320
+ },
321
+ "bits": [
322
+ 4
323
+ ],
324
+ "bits_prop": [
325
+ 1
326
+ ],
327
+ "scale_bits": 4,
328
+ "scale_groups:": 32
329
+ },
330
+ "k_proj": {
331
+ "group_size": {
332
+ "4": 128
333
+ },
334
+ "bits": [
335
+ 4
336
+ ],
337
+ "bits_prop": [
338
+ 1
339
+ ],
340
+ "scale_bits": 4,
341
+ "scale_groups:": 32
342
+ },
343
+ "v_proj": {
344
+ "group_size": {
345
+ "4": 128
346
+ },
347
+ "bits": [
348
+ 4
349
+ ],
350
+ "bits_prop": [
351
+ 1
352
+ ],
353
+ "scale_bits": 4,
354
+ "scale_groups:": 32
355
+ },
356
+ "gate_proj": {
357
+ "group_size": {
358
+ "4": 128
359
+ },
360
+ "bits": [
361
+ 4
362
+ ],
363
+ "bits_prop": [
364
+ 1
365
+ ],
366
+ "scale_bits": 4,
367
+ "scale_groups:": 32
368
+ },
369
+ "up_proj": {
370
+ "group_size": {
371
+ "4": 128
372
+ },
373
+ "bits": [
374
+ 4
375
+ ],
376
+ "bits_prop": [
377
+ 1
378
+ ],
379
+ "scale_bits": 4,
380
+ "scale_groups:": 32
381
+ }
382
+ },
383
+ "model.layers.4": {
384
+ "accuracy": 0.9747236967086792,
385
+ "total_bits": 314302464.0,
386
+ "o_proj": {
387
+ "group_size": {
388
+ "4": 128
389
+ },
390
+ "bits": [
391
+ 4
392
+ ],
393
+ "bits_prop": [
394
+ 1
395
+ ],
396
+ "scale_bits": 4,
397
+ "scale_groups:": 32
398
+ },
399
+ "down_proj": {
400
+ "group_size": {
401
+ "4": 128
402
+ },
403
+ "bits": [
404
+ 4
405
+ ],
406
+ "bits_prop": [
407
+ 1
408
+ ],
409
+ "scale_bits": 4,
410
+ "scale_groups:": 32
411
+ },
412
+ "q_proj": {
413
+ "group_size": {
414
+ "4": 128
415
+ },
416
+ "bits": [
417
+ 4
418
+ ],
419
+ "bits_prop": [
420
+ 1
421
+ ],
422
+ "scale_bits": 4,
423
+ "scale_groups:": 32
424
+ },
425
+ "k_proj": {
426
+ "group_size": {
427
+ "4": 128
428
+ },
429
+ "bits": [
430
+ 4
431
+ ],
432
+ "bits_prop": [
433
+ 1
434
+ ],
435
+ "scale_bits": 4,
436
+ "scale_groups:": 32
437
+ },
438
+ "v_proj": {
439
+ "group_size": {
440
+ "4": 128
441
+ },
442
+ "bits": [
443
+ 4
444
+ ],
445
+ "bits_prop": [
446
+ 1
447
+ ],
448
+ "scale_bits": 4,
449
+ "scale_groups:": 32
450
+ },
451
+ "gate_proj": {
452
+ "group_size": {
453
+ "4": 128
454
+ },
455
+ "bits": [
456
+ 4
457
+ ],
458
+ "bits_prop": [
459
+ 1
460
+ ],
461
+ "scale_bits": 4,
462
+ "scale_groups:": 32
463
+ },
464
+ "up_proj": {
465
+ "group_size": {
466
+ "4": 128
467
+ },
468
+ "bits": [
469
+ 4
470
+ ],
471
+ "bits_prop": [
472
+ 1
473
+ ],
474
+ "scale_bits": 4,
475
+ "scale_groups:": 32
476
+ }
477
+ },
478
+ "model.layers.5": {
479
+ "accuracy": 0.9633625745773315,
480
+ "total_bits": 314302464.0,
481
+ "o_proj": {
482
+ "group_size": {
483
+ "4": 128
484
+ },
485
+ "bits": [
486
+ 4
487
+ ],
488
+ "bits_prop": [
489
+ 1
490
+ ],
491
+ "scale_bits": 4,
492
+ "scale_groups:": 32
493
+ },
494
+ "down_proj": {
495
+ "group_size": {
496
+ "4": 128
497
+ },
498
+ "bits": [
499
+ 4
500
+ ],
501
+ "bits_prop": [
502
+ 1
503
+ ],
504
+ "scale_bits": 4,
505
+ "scale_groups:": 32
506
+ },
507
+ "q_proj": {
508
+ "group_size": {
509
+ "4": 128
510
+ },
511
+ "bits": [
512
+ 4
513
+ ],
514
+ "bits_prop": [
515
+ 1
516
+ ],
517
+ "scale_bits": 4,
518
+ "scale_groups:": 32
519
+ },
520
+ "k_proj": {
521
+ "group_size": {
522
+ "4": 128
523
+ },
524
+ "bits": [
525
+ 4
526
+ ],
527
+ "bits_prop": [
528
+ 1
529
+ ],
530
+ "scale_bits": 4,
531
+ "scale_groups:": 32
532
+ },
533
+ "v_proj": {
534
+ "group_size": {
535
+ "4": 128
536
+ },
537
+ "bits": [
538
+ 4
539
+ ],
540
+ "bits_prop": [
541
+ 1
542
+ ],
543
+ "scale_bits": 4,
544
+ "scale_groups:": 32
545
+ },
546
+ "gate_proj": {
547
+ "group_size": {
548
+ "4": 128
549
+ },
550
+ "bits": [
551
+ 4
552
+ ],
553
+ "bits_prop": [
554
+ 1
555
+ ],
556
+ "scale_bits": 4,
557
+ "scale_groups:": 32
558
+ },
559
+ "up_proj": {
560
+ "group_size": {
561
+ "4": 128
562
+ },
563
+ "bits": [
564
+ 4
565
+ ],
566
+ "bits_prop": [
567
+ 1
568
+ ],
569
+ "scale_bits": 4,
570
+ "scale_groups:": 32
571
+ }
572
+ },
573
+ "model.layers.6": {
574
+ "accuracy": 0.9738354682922363,
575
+ "total_bits": 314302464.0,
576
+ "o_proj": {
577
+ "group_size": {
578
+ "4": 128
579
+ },
580
+ "bits": [
581
+ 4
582
+ ],
583
+ "bits_prop": [
584
+ 1
585
+ ],
586
+ "scale_bits": 4,
587
+ "scale_groups:": 32
588
+ },
589
+ "down_proj": {
590
+ "group_size": {
591
+ "4": 128
592
+ },
593
+ "bits": [
594
+ 4
595
+ ],
596
+ "bits_prop": [
597
+ 1
598
+ ],
599
+ "scale_bits": 4,
600
+ "scale_groups:": 32
601
+ },
602
+ "q_proj": {
603
+ "group_size": {
604
+ "4": 128
605
+ },
606
+ "bits": [
607
+ 4
608
+ ],
609
+ "bits_prop": [
610
+ 1
611
+ ],
612
+ "scale_bits": 4,
613
+ "scale_groups:": 32
614
+ },
615
+ "k_proj": {
616
+ "group_size": {
617
+ "4": 128
618
+ },
619
+ "bits": [
620
+ 4
621
+ ],
622
+ "bits_prop": [
623
+ 1
624
+ ],
625
+ "scale_bits": 4,
626
+ "scale_groups:": 32
627
+ },
628
+ "v_proj": {
629
+ "group_size": {
630
+ "4": 128
631
+ },
632
+ "bits": [
633
+ 4
634
+ ],
635
+ "bits_prop": [
636
+ 1
637
+ ],
638
+ "scale_bits": 4,
639
+ "scale_groups:": 32
640
+ },
641
+ "gate_proj": {
642
+ "group_size": {
643
+ "4": 128
644
+ },
645
+ "bits": [
646
+ 4
647
+ ],
648
+ "bits_prop": [
649
+ 1
650
+ ],
651
+ "scale_bits": 4,
652
+ "scale_groups:": 32
653
+ },
654
+ "up_proj": {
655
+ "group_size": {
656
+ "4": 128
657
+ },
658
+ "bits": [
659
+ 4
660
+ ],
661
+ "bits_prop": [
662
+ 1
663
+ ],
664
+ "scale_bits": 4,
665
+ "scale_groups:": 32
666
+ }
667
+ },
668
+ "model.layers.7": {
669
+ "accuracy": 0.9770342111587524,
670
+ "total_bits": 314302464.0,
671
+ "o_proj": {
672
+ "group_size": {
673
+ "4": 128
674
+ },
675
+ "bits": [
676
+ 4
677
+ ],
678
+ "bits_prop": [
679
+ 1
680
+ ],
681
+ "scale_bits": 4,
682
+ "scale_groups:": 32
683
+ },
684
+ "down_proj": {
685
+ "group_size": {
686
+ "4": 128
687
+ },
688
+ "bits": [
689
+ 4
690
+ ],
691
+ "bits_prop": [
692
+ 1
693
+ ],
694
+ "scale_bits": 4,
695
+ "scale_groups:": 32
696
+ },
697
+ "q_proj": {
698
+ "group_size": {
699
+ "4": 128
700
+ },
701
+ "bits": [
702
+ 4
703
+ ],
704
+ "bits_prop": [
705
+ 1
706
+ ],
707
+ "scale_bits": 4,
708
+ "scale_groups:": 32
709
+ },
710
+ "k_proj": {
711
+ "group_size": {
712
+ "4": 128
713
+ },
714
+ "bits": [
715
+ 4
716
+ ],
717
+ "bits_prop": [
718
+ 1
719
+ ],
720
+ "scale_bits": 4,
721
+ "scale_groups:": 32
722
+ },
723
+ "v_proj": {
724
+ "group_size": {
725
+ "4": 128
726
+ },
727
+ "bits": [
728
+ 4
729
+ ],
730
+ "bits_prop": [
731
+ 1
732
+ ],
733
+ "scale_bits": 4,
734
+ "scale_groups:": 32
735
+ },
736
+ "gate_proj": {
737
+ "group_size": {
738
+ "4": 128
739
+ },
740
+ "bits": [
741
+ 4
742
+ ],
743
+ "bits_prop": [
744
+ 1
745
+ ],
746
+ "scale_bits": 4,
747
+ "scale_groups:": 32
748
+ },
749
+ "up_proj": {
750
+ "group_size": {
751
+ "4": 128
752
+ },
753
+ "bits": [
754
+ 4
755
+ ],
756
+ "bits_prop": [
757
+ 1
758
+ ],
759
+ "scale_bits": 4,
760
+ "scale_groups:": 32
761
+ }
762
+ },
763
+ "model.layers.8": {
764
+ "accuracy": 0.9778237342834473,
765
+ "total_bits": 314302464.0,
766
+ "o_proj": {
767
+ "group_size": {
768
+ "4": 128
769
+ },
770
+ "bits": [
771
+ 4
772
+ ],
773
+ "bits_prop": [
774
+ 1
775
+ ],
776
+ "scale_bits": 4,
777
+ "scale_groups:": 32
778
+ },
779
+ "down_proj": {
780
+ "group_size": {
781
+ "4": 128
782
+ },
783
+ "bits": [
784
+ 4
785
+ ],
786
+ "bits_prop": [
787
+ 1
788
+ ],
789
+ "scale_bits": 4,
790
+ "scale_groups:": 32
791
+ },
792
+ "q_proj": {
793
+ "group_size": {
794
+ "4": 128
795
+ },
796
+ "bits": [
797
+ 4
798
+ ],
799
+ "bits_prop": [
800
+ 1
801
+ ],
802
+ "scale_bits": 4,
803
+ "scale_groups:": 32
804
+ },
805
+ "k_proj": {
806
+ "group_size": {
807
+ "4": 128
808
+ },
809
+ "bits": [
810
+ 4
811
+ ],
812
+ "bits_prop": [
813
+ 1
814
+ ],
815
+ "scale_bits": 4,
816
+ "scale_groups:": 32
817
+ },
818
+ "v_proj": {
819
+ "group_size": {
820
+ "4": 128
821
+ },
822
+ "bits": [
823
+ 4
824
+ ],
825
+ "bits_prop": [
826
+ 1
827
+ ],
828
+ "scale_bits": 4,
829
+ "scale_groups:": 32
830
+ },
831
+ "gate_proj": {
832
+ "group_size": {
833
+ "4": 128
834
+ },
835
+ "bits": [
836
+ 4
837
+ ],
838
+ "bits_prop": [
839
+ 1
840
+ ],
841
+ "scale_bits": 4,
842
+ "scale_groups:": 32
843
+ },
844
+ "up_proj": {
845
+ "group_size": {
846
+ "4": 128
847
+ },
848
+ "bits": [
849
+ 4
850
+ ],
851
+ "bits_prop": [
852
+ 1
853
+ ],
854
+ "scale_bits": 4,
855
+ "scale_groups:": 32
856
+ }
857
+ },
858
+ "model.layers.9": {
859
+ "accuracy": 0.9768979549407959,
860
+ "total_bits": 314302464.0,
861
+ "o_proj": {
862
+ "group_size": {
863
+ "4": 128
864
+ },
865
+ "bits": [
866
+ 4
867
+ ],
868
+ "bits_prop": [
869
+ 1
870
+ ],
871
+ "scale_bits": 4,
872
+ "scale_groups:": 32
873
+ },
874
+ "down_proj": {
875
+ "group_size": {
876
+ "4": 128
877
+ },
878
+ "bits": [
879
+ 4
880
+ ],
881
+ "bits_prop": [
882
+ 1
883
+ ],
884
+ "scale_bits": 4,
885
+ "scale_groups:": 32
886
+ },
887
+ "q_proj": {
888
+ "group_size": {
889
+ "4": 128
890
+ },
891
+ "bits": [
892
+ 4
893
+ ],
894
+ "bits_prop": [
895
+ 1
896
+ ],
897
+ "scale_bits": 4,
898
+ "scale_groups:": 32
899
+ },
900
+ "k_proj": {
901
+ "group_size": {
902
+ "4": 128
903
+ },
904
+ "bits": [
905
+ 4
906
+ ],
907
+ "bits_prop": [
908
+ 1
909
+ ],
910
+ "scale_bits": 4,
911
+ "scale_groups:": 32
912
+ },
913
+ "v_proj": {
914
+ "group_size": {
915
+ "4": 128
916
+ },
917
+ "bits": [
918
+ 4
919
+ ],
920
+ "bits_prop": [
921
+ 1
922
+ ],
923
+ "scale_bits": 4,
924
+ "scale_groups:": 32
925
+ },
926
+ "gate_proj": {
927
+ "group_size": {
928
+ "4": 128
929
+ },
930
+ "bits": [
931
+ 4
932
+ ],
933
+ "bits_prop": [
934
+ 1
935
+ ],
936
+ "scale_bits": 4,
937
+ "scale_groups:": 32
938
+ },
939
+ "up_proj": {
940
+ "group_size": {
941
+ "4": 128
942
+ },
943
+ "bits": [
944
+ 4
945
+ ],
946
+ "bits_prop": [
947
+ 1
948
+ ],
949
+ "scale_bits": 4,
950
+ "scale_groups:": 32
951
+ }
952
+ },
953
+ "model.layers.10": {
954
+ "accuracy": 0.9737787246704102,
955
+ "total_bits": 314302464.0,
956
+ "o_proj": {
957
+ "group_size": {
958
+ "4": 128
959
+ },
960
+ "bits": [
961
+ 4
962
+ ],
963
+ "bits_prop": [
964
+ 1
965
+ ],
966
+ "scale_bits": 4,
967
+ "scale_groups:": 32
968
+ },
969
+ "down_proj": {
970
+ "group_size": {
971
+ "4": 128
972
+ },
973
+ "bits": [
974
+ 4
975
+ ],
976
+ "bits_prop": [
977
+ 1
978
+ ],
979
+ "scale_bits": 4,
980
+ "scale_groups:": 32
981
+ },
982
+ "q_proj": {
983
+ "group_size": {
984
+ "4": 128
985
+ },
986
+ "bits": [
987
+ 4
988
+ ],
989
+ "bits_prop": [
990
+ 1
991
+ ],
992
+ "scale_bits": 4,
993
+ "scale_groups:": 32
994
+ },
995
+ "k_proj": {
996
+ "group_size": {
997
+ "4": 128
998
+ },
999
+ "bits": [
1000
+ 4
1001
+ ],
1002
+ "bits_prop": [
1003
+ 1
1004
+ ],
1005
+ "scale_bits": 4,
1006
+ "scale_groups:": 32
1007
+ },
1008
+ "v_proj": {
1009
+ "group_size": {
1010
+ "4": 128
1011
+ },
1012
+ "bits": [
1013
+ 4
1014
+ ],
1015
+ "bits_prop": [
1016
+ 1
1017
+ ],
1018
+ "scale_bits": 4,
1019
+ "scale_groups:": 32
1020
+ },
1021
+ "gate_proj": {
1022
+ "group_size": {
1023
+ "4": 128
1024
+ },
1025
+ "bits": [
1026
+ 4
1027
+ ],
1028
+ "bits_prop": [
1029
+ 1
1030
+ ],
1031
+ "scale_bits": 4,
1032
+ "scale_groups:": 32
1033
+ },
1034
+ "up_proj": {
1035
+ "group_size": {
1036
+ "4": 128
1037
+ },
1038
+ "bits": [
1039
+ 4
1040
+ ],
1041
+ "bits_prop": [
1042
+ 1
1043
+ ],
1044
+ "scale_bits": 4,
1045
+ "scale_groups:": 32
1046
+ }
1047
+ },
1048
+ "model.layers.11": {
1049
+ "accuracy": 0.9728798866271973,
1050
+ "total_bits": 314302464.0,
1051
+ "o_proj": {
1052
+ "group_size": {
1053
+ "4": 128
1054
+ },
1055
+ "bits": [
1056
+ 4
1057
+ ],
1058
+ "bits_prop": [
1059
+ 1
1060
+ ],
1061
+ "scale_bits": 4,
1062
+ "scale_groups:": 32
1063
+ },
1064
+ "down_proj": {
1065
+ "group_size": {
1066
+ "4": 128
1067
+ },
1068
+ "bits": [
1069
+ 4
1070
+ ],
1071
+ "bits_prop": [
1072
+ 1
1073
+ ],
1074
+ "scale_bits": 4,
1075
+ "scale_groups:": 32
1076
+ },
1077
+ "q_proj": {
1078
+ "group_size": {
1079
+ "4": 128
1080
+ },
1081
+ "bits": [
1082
+ 4
1083
+ ],
1084
+ "bits_prop": [
1085
+ 1
1086
+ ],
1087
+ "scale_bits": 4,
1088
+ "scale_groups:": 32
1089
+ },
1090
+ "k_proj": {
1091
+ "group_size": {
1092
+ "4": 128
1093
+ },
1094
+ "bits": [
1095
+ 4
1096
+ ],
1097
+ "bits_prop": [
1098
+ 1
1099
+ ],
1100
+ "scale_bits": 4,
1101
+ "scale_groups:": 32
1102
+ },
1103
+ "v_proj": {
1104
+ "group_size": {
1105
+ "4": 128
1106
+ },
1107
+ "bits": [
1108
+ 4
1109
+ ],
1110
+ "bits_prop": [
1111
+ 1
1112
+ ],
1113
+ "scale_bits": 4,
1114
+ "scale_groups:": 32
1115
+ },
1116
+ "gate_proj": {
1117
+ "group_size": {
1118
+ "4": 128
1119
+ },
1120
+ "bits": [
1121
+ 4
1122
+ ],
1123
+ "bits_prop": [
1124
+ 1
1125
+ ],
1126
+ "scale_bits": 4,
1127
+ "scale_groups:": 32
1128
+ },
1129
+ "up_proj": {
1130
+ "group_size": {
1131
+ "4": 128
1132
+ },
1133
+ "bits": [
1134
+ 4
1135
+ ],
1136
+ "bits_prop": [
1137
+ 1
1138
+ ],
1139
+ "scale_bits": 4,
1140
+ "scale_groups:": 32
1141
+ }
1142
+ },
1143
+ "model.layers.12": {
1144
+ "accuracy": 0.9736931324005127,
1145
+ "total_bits": 314302464.0,
1146
+ "o_proj": {
1147
+ "group_size": {
1148
+ "4": 128
1149
+ },
1150
+ "bits": [
1151
+ 4
1152
+ ],
1153
+ "bits_prop": [
1154
+ 1
1155
+ ],
1156
+ "scale_bits": 4,
1157
+ "scale_groups:": 32
1158
+ },
1159
+ "down_proj": {
1160
+ "group_size": {
1161
+ "4": 128
1162
+ },
1163
+ "bits": [
1164
+ 4
1165
+ ],
1166
+ "bits_prop": [
1167
+ 1
1168
+ ],
1169
+ "scale_bits": 4,
1170
+ "scale_groups:": 32
1171
+ },
1172
+ "q_proj": {
1173
+ "group_size": {
1174
+ "4": 128
1175
+ },
1176
+ "bits": [
1177
+ 4
1178
+ ],
1179
+ "bits_prop": [
1180
+ 1
1181
+ ],
1182
+ "scale_bits": 4,
1183
+ "scale_groups:": 32
1184
+ },
1185
+ "k_proj": {
1186
+ "group_size": {
1187
+ "4": 128
1188
+ },
1189
+ "bits": [
1190
+ 4
1191
+ ],
1192
+ "bits_prop": [
1193
+ 1
1194
+ ],
1195
+ "scale_bits": 4,
1196
+ "scale_groups:": 32
1197
+ },
1198
+ "v_proj": {
1199
+ "group_size": {
1200
+ "4": 128
1201
+ },
1202
+ "bits": [
1203
+ 4
1204
+ ],
1205
+ "bits_prop": [
1206
+ 1
1207
+ ],
1208
+ "scale_bits": 4,
1209
+ "scale_groups:": 32
1210
+ },
1211
+ "gate_proj": {
1212
+ "group_size": {
1213
+ "4": 128
1214
+ },
1215
+ "bits": [
1216
+ 4
1217
+ ],
1218
+ "bits_prop": [
1219
+ 1
1220
+ ],
1221
+ "scale_bits": 4,
1222
+ "scale_groups:": 32
1223
+ },
1224
+ "up_proj": {
1225
+ "group_size": {
1226
+ "4": 128
1227
+ },
1228
+ "bits": [
1229
+ 4
1230
+ ],
1231
+ "bits_prop": [
1232
+ 1
1233
+ ],
1234
+ "scale_bits": 4,
1235
+ "scale_groups:": 32
1236
+ }
1237
+ },
1238
+ "model.layers.13": {
1239
+ "accuracy": 0.9764463901519775,
1240
+ "total_bits": 314302464.0,
1241
+ "o_proj": {
1242
+ "group_size": {
1243
+ "4": 128
1244
+ },
1245
+ "bits": [
1246
+ 4
1247
+ ],
1248
+ "bits_prop": [
1249
+ 1
1250
+ ],
1251
+ "scale_bits": 4,
1252
+ "scale_groups:": 32
1253
+ },
1254
+ "down_proj": {
1255
+ "group_size": {
1256
+ "4": 128
1257
+ },
1258
+ "bits": [
1259
+ 4
1260
+ ],
1261
+ "bits_prop": [
1262
+ 1
1263
+ ],
1264
+ "scale_bits": 4,
1265
+ "scale_groups:": 32
1266
+ },
1267
+ "q_proj": {
1268
+ "group_size": {
1269
+ "4": 128
1270
+ },
1271
+ "bits": [
1272
+ 4
1273
+ ],
1274
+ "bits_prop": [
1275
+ 1
1276
+ ],
1277
+ "scale_bits": 4,
1278
+ "scale_groups:": 32
1279
+ },
1280
+ "k_proj": {
1281
+ "group_size": {
1282
+ "4": 128
1283
+ },
1284
+ "bits": [
1285
+ 4
1286
+ ],
1287
+ "bits_prop": [
1288
+ 1
1289
+ ],
1290
+ "scale_bits": 4,
1291
+ "scale_groups:": 32
1292
+ },
1293
+ "v_proj": {
1294
+ "group_size": {
1295
+ "4": 128
1296
+ },
1297
+ "bits": [
1298
+ 4
1299
+ ],
1300
+ "bits_prop": [
1301
+ 1
1302
+ ],
1303
+ "scale_bits": 4,
1304
+ "scale_groups:": 32
1305
+ },
1306
+ "gate_proj": {
1307
+ "group_size": {
1308
+ "4": 128
1309
+ },
1310
+ "bits": [
1311
+ 4
1312
+ ],
1313
+ "bits_prop": [
1314
+ 1
1315
+ ],
1316
+ "scale_bits": 4,
1317
+ "scale_groups:": 32
1318
+ },
1319
+ "up_proj": {
1320
+ "group_size": {
1321
+ "4": 128
1322
+ },
1323
+ "bits": [
1324
+ 4
1325
+ ],
1326
+ "bits_prop": [
1327
+ 1
1328
+ ],
1329
+ "scale_bits": 4,
1330
+ "scale_groups:": 32
1331
+ }
1332
+ },
1333
+ "model.layers.14": {
1334
+ "accuracy": 0.9773211479187012,
1335
+ "total_bits": 314302464.0,
1336
+ "o_proj": {
1337
+ "group_size": {
1338
+ "4": 128
1339
+ },
1340
+ "bits": [
1341
+ 4
1342
+ ],
1343
+ "bits_prop": [
1344
+ 1
1345
+ ],
1346
+ "scale_bits": 4,
1347
+ "scale_groups:": 32
1348
+ },
1349
+ "down_proj": {
1350
+ "group_size": {
1351
+ "4": 128
1352
+ },
1353
+ "bits": [
1354
+ 4
1355
+ ],
1356
+ "bits_prop": [
1357
+ 1
1358
+ ],
1359
+ "scale_bits": 4,
1360
+ "scale_groups:": 32
1361
+ },
1362
+ "q_proj": {
1363
+ "group_size": {
1364
+ "4": 128
1365
+ },
1366
+ "bits": [
1367
+ 4
1368
+ ],
1369
+ "bits_prop": [
1370
+ 1
1371
+ ],
1372
+ "scale_bits": 4,
1373
+ "scale_groups:": 32
1374
+ },
1375
+ "k_proj": {
1376
+ "group_size": {
1377
+ "4": 128
1378
+ },
1379
+ "bits": [
1380
+ 4
1381
+ ],
1382
+ "bits_prop": [
1383
+ 1
1384
+ ],
1385
+ "scale_bits": 4,
1386
+ "scale_groups:": 32
1387
+ },
1388
+ "v_proj": {
1389
+ "group_size": {
1390
+ "4": 128
1391
+ },
1392
+ "bits": [
1393
+ 4
1394
+ ],
1395
+ "bits_prop": [
1396
+ 1
1397
+ ],
1398
+ "scale_bits": 4,
1399
+ "scale_groups:": 32
1400
+ },
1401
+ "gate_proj": {
1402
+ "group_size": {
1403
+ "4": 128
1404
+ },
1405
+ "bits": [
1406
+ 4
1407
+ ],
1408
+ "bits_prop": [
1409
+ 1
1410
+ ],
1411
+ "scale_bits": 4,
1412
+ "scale_groups:": 32
1413
+ },
1414
+ "up_proj": {
1415
+ "group_size": {
1416
+ "4": 128
1417
+ },
1418
+ "bits": [
1419
+ 4
1420
+ ],
1421
+ "bits_prop": [
1422
+ 1
1423
+ ],
1424
+ "scale_bits": 4,
1425
+ "scale_groups:": 32
1426
+ }
1427
+ },
1428
+ "model.layers.15": {
1429
+ "accuracy": 0.9774775505065918,
1430
+ "total_bits": 314302464.0,
1431
+ "o_proj": {
1432
+ "group_size": {
1433
+ "4": 128
1434
+ },
1435
+ "bits": [
1436
+ 4
1437
+ ],
1438
+ "bits_prop": [
1439
+ 1
1440
+ ],
1441
+ "scale_bits": 4,
1442
+ "scale_groups:": 32
1443
+ },
1444
+ "down_proj": {
1445
+ "group_size": {
1446
+ "4": 128
1447
+ },
1448
+ "bits": [
1449
+ 4
1450
+ ],
1451
+ "bits_prop": [
1452
+ 1
1453
+ ],
1454
+ "scale_bits": 4,
1455
+ "scale_groups:": 32
1456
+ },
1457
+ "q_proj": {
1458
+ "group_size": {
1459
+ "4": 128
1460
+ },
1461
+ "bits": [
1462
+ 4
1463
+ ],
1464
+ "bits_prop": [
1465
+ 1
1466
+ ],
1467
+ "scale_bits": 4,
1468
+ "scale_groups:": 32
1469
+ },
1470
+ "k_proj": {
1471
+ "group_size": {
1472
+ "4": 128
1473
+ },
1474
+ "bits": [
1475
+ 4
1476
+ ],
1477
+ "bits_prop": [
1478
+ 1
1479
+ ],
1480
+ "scale_bits": 4,
1481
+ "scale_groups:": 32
1482
+ },
1483
+ "v_proj": {
1484
+ "group_size": {
1485
+ "4": 128
1486
+ },
1487
+ "bits": [
1488
+ 4
1489
+ ],
1490
+ "bits_prop": [
1491
+ 1
1492
+ ],
1493
+ "scale_bits": 4,
1494
+ "scale_groups:": 32
1495
+ },
1496
+ "gate_proj": {
1497
+ "group_size": {
1498
+ "4": 128
1499
+ },
1500
+ "bits": [
1501
+ 4
1502
+ ],
1503
+ "bits_prop": [
1504
+ 1
1505
+ ],
1506
+ "scale_bits": 4,
1507
+ "scale_groups:": 32
1508
+ },
1509
+ "up_proj": {
1510
+ "group_size": {
1511
+ "4": 128
1512
+ },
1513
+ "bits": [
1514
+ 4
1515
+ ],
1516
+ "bits_prop": [
1517
+ 1
1518
+ ],
1519
+ "scale_bits": 4,
1520
+ "scale_groups:": 32
1521
+ }
1522
+ },
1523
+ "model.layers.16": {
1524
+ "accuracy": 0.9796525239944458,
1525
+ "total_bits": 314302464.0,
1526
+ "o_proj": {
1527
+ "group_size": {
1528
+ "4": 128
1529
+ },
1530
+ "bits": [
1531
+ 4
1532
+ ],
1533
+ "bits_prop": [
1534
+ 1
1535
+ ],
1536
+ "scale_bits": 4,
1537
+ "scale_groups:": 32
1538
+ },
1539
+ "down_proj": {
1540
+ "group_size": {
1541
+ "4": 128
1542
+ },
1543
+ "bits": [
1544
+ 4
1545
+ ],
1546
+ "bits_prop": [
1547
+ 1
1548
+ ],
1549
+ "scale_bits": 4,
1550
+ "scale_groups:": 32
1551
+ },
1552
+ "q_proj": {
1553
+ "group_size": {
1554
+ "4": 128
1555
+ },
1556
+ "bits": [
1557
+ 4
1558
+ ],
1559
+ "bits_prop": [
1560
+ 1
1561
+ ],
1562
+ "scale_bits": 4,
1563
+ "scale_groups:": 32
1564
+ },
1565
+ "k_proj": {
1566
+ "group_size": {
1567
+ "4": 128
1568
+ },
1569
+ "bits": [
1570
+ 4
1571
+ ],
1572
+ "bits_prop": [
1573
+ 1
1574
+ ],
1575
+ "scale_bits": 4,
1576
+ "scale_groups:": 32
1577
+ },
1578
+ "v_proj": {
1579
+ "group_size": {
1580
+ "4": 128
1581
+ },
1582
+ "bits": [
1583
+ 4
1584
+ ],
1585
+ "bits_prop": [
1586
+ 1
1587
+ ],
1588
+ "scale_bits": 4,
1589
+ "scale_groups:": 32
1590
+ },
1591
+ "gate_proj": {
1592
+ "group_size": {
1593
+ "4": 128
1594
+ },
1595
+ "bits": [
1596
+ 4
1597
+ ],
1598
+ "bits_prop": [
1599
+ 1
1600
+ ],
1601
+ "scale_bits": 4,
1602
+ "scale_groups:": 32
1603
+ },
1604
+ "up_proj": {
1605
+ "group_size": {
1606
+ "4": 128
1607
+ },
1608
+ "bits": [
1609
+ 4
1610
+ ],
1611
+ "bits_prop": [
1612
+ 1
1613
+ ],
1614
+ "scale_bits": 4,
1615
+ "scale_groups:": 32
1616
+ }
1617
+ },
1618
+ "model.layers.17": {
1619
+ "accuracy": 0.9784044027328491,
1620
+ "total_bits": 314302464.0,
1621
+ "o_proj": {
1622
+ "group_size": {
1623
+ "4": 128
1624
+ },
1625
+ "bits": [
1626
+ 4
1627
+ ],
1628
+ "bits_prop": [
1629
+ 1
1630
+ ],
1631
+ "scale_bits": 4,
1632
+ "scale_groups:": 32
1633
+ },
1634
+ "down_proj": {
1635
+ "group_size": {
1636
+ "4": 128
1637
+ },
1638
+ "bits": [
1639
+ 4
1640
+ ],
1641
+ "bits_prop": [
1642
+ 1
1643
+ ],
1644
+ "scale_bits": 4,
1645
+ "scale_groups:": 32
1646
+ },
1647
+ "q_proj": {
1648
+ "group_size": {
1649
+ "4": 128
1650
+ },
1651
+ "bits": [
1652
+ 4
1653
+ ],
1654
+ "bits_prop": [
1655
+ 1
1656
+ ],
1657
+ "scale_bits": 4,
1658
+ "scale_groups:": 32
1659
+ },
1660
+ "k_proj": {
1661
+ "group_size": {
1662
+ "4": 128
1663
+ },
1664
+ "bits": [
1665
+ 4
1666
+ ],
1667
+ "bits_prop": [
1668
+ 1
1669
+ ],
1670
+ "scale_bits": 4,
1671
+ "scale_groups:": 32
1672
+ },
1673
+ "v_proj": {
1674
+ "group_size": {
1675
+ "4": 128
1676
+ },
1677
+ "bits": [
1678
+ 4
1679
+ ],
1680
+ "bits_prop": [
1681
+ 1
1682
+ ],
1683
+ "scale_bits": 4,
1684
+ "scale_groups:": 32
1685
+ },
1686
+ "gate_proj": {
1687
+ "group_size": {
1688
+ "4": 128
1689
+ },
1690
+ "bits": [
1691
+ 4
1692
+ ],
1693
+ "bits_prop": [
1694
+ 1
1695
+ ],
1696
+ "scale_bits": 4,
1697
+ "scale_groups:": 32
1698
+ },
1699
+ "up_proj": {
1700
+ "group_size": {
1701
+ "4": 128
1702
+ },
1703
+ "bits": [
1704
+ 4
1705
+ ],
1706
+ "bits_prop": [
1707
+ 1
1708
+ ],
1709
+ "scale_bits": 4,
1710
+ "scale_groups:": 32
1711
+ }
1712
+ },
1713
+ "model.layers.18": {
1714
+ "accuracy": 0.9794589281082153,
1715
+ "total_bits": 314302464.0,
1716
+ "o_proj": {
1717
+ "group_size": {
1718
+ "4": 128
1719
+ },
1720
+ "bits": [
1721
+ 4
1722
+ ],
1723
+ "bits_prop": [
1724
+ 1
1725
+ ],
1726
+ "scale_bits": 4,
1727
+ "scale_groups:": 32
1728
+ },
1729
+ "down_proj": {
1730
+ "group_size": {
1731
+ "4": 128
1732
+ },
1733
+ "bits": [
1734
+ 4
1735
+ ],
1736
+ "bits_prop": [
1737
+ 1
1738
+ ],
1739
+ "scale_bits": 4,
1740
+ "scale_groups:": 32
1741
+ },
1742
+ "q_proj": {
1743
+ "group_size": {
1744
+ "4": 128
1745
+ },
1746
+ "bits": [
1747
+ 4
1748
+ ],
1749
+ "bits_prop": [
1750
+ 1
1751
+ ],
1752
+ "scale_bits": 4,
1753
+ "scale_groups:": 32
1754
+ },
1755
+ "k_proj": {
1756
+ "group_size": {
1757
+ "4": 128
1758
+ },
1759
+ "bits": [
1760
+ 4
1761
+ ],
1762
+ "bits_prop": [
1763
+ 1
1764
+ ],
1765
+ "scale_bits": 4,
1766
+ "scale_groups:": 32
1767
+ },
1768
+ "v_proj": {
1769
+ "group_size": {
1770
+ "4": 128
1771
+ },
1772
+ "bits": [
1773
+ 4
1774
+ ],
1775
+ "bits_prop": [
1776
+ 1
1777
+ ],
1778
+ "scale_bits": 4,
1779
+ "scale_groups:": 32
1780
+ },
1781
+ "gate_proj": {
1782
+ "group_size": {
1783
+ "4": 128
1784
+ },
1785
+ "bits": [
1786
+ 4
1787
+ ],
1788
+ "bits_prop": [
1789
+ 1
1790
+ ],
1791
+ "scale_bits": 4,
1792
+ "scale_groups:": 32
1793
+ },
1794
+ "up_proj": {
1795
+ "group_size": {
1796
+ "4": 128
1797
+ },
1798
+ "bits": [
1799
+ 4
1800
+ ],
1801
+ "bits_prop": [
1802
+ 1
1803
+ ],
1804
+ "scale_bits": 4,
1805
+ "scale_groups:": 32
1806
+ }
1807
+ },
1808
+ "model.layers.19": {
1809
+ "accuracy": 0.979212760925293,
1810
+ "total_bits": 314302464.0,
1811
+ "o_proj": {
1812
+ "group_size": {
1813
+ "4": 128
1814
+ },
1815
+ "bits": [
1816
+ 4
1817
+ ],
1818
+ "bits_prop": [
1819
+ 1
1820
+ ],
1821
+ "scale_bits": 4,
1822
+ "scale_groups:": 32
1823
+ },
1824
+ "down_proj": {
1825
+ "group_size": {
1826
+ "4": 128
1827
+ },
1828
+ "bits": [
1829
+ 4
1830
+ ],
1831
+ "bits_prop": [
1832
+ 1
1833
+ ],
1834
+ "scale_bits": 4,
1835
+ "scale_groups:": 32
1836
+ },
1837
+ "q_proj": {
1838
+ "group_size": {
1839
+ "4": 128
1840
+ },
1841
+ "bits": [
1842
+ 4
1843
+ ],
1844
+ "bits_prop": [
1845
+ 1
1846
+ ],
1847
+ "scale_bits": 4,
1848
+ "scale_groups:": 32
1849
+ },
1850
+ "k_proj": {
1851
+ "group_size": {
1852
+ "4": 128
1853
+ },
1854
+ "bits": [
1855
+ 4
1856
+ ],
1857
+ "bits_prop": [
1858
+ 1
1859
+ ],
1860
+ "scale_bits": 4,
1861
+ "scale_groups:": 32
1862
+ },
1863
+ "v_proj": {
1864
+ "group_size": {
1865
+ "4": 128
1866
+ },
1867
+ "bits": [
1868
+ 4
1869
+ ],
1870
+ "bits_prop": [
1871
+ 1
1872
+ ],
1873
+ "scale_bits": 4,
1874
+ "scale_groups:": 32
1875
+ },
1876
+ "gate_proj": {
1877
+ "group_size": {
1878
+ "4": 128
1879
+ },
1880
+ "bits": [
1881
+ 4
1882
+ ],
1883
+ "bits_prop": [
1884
+ 1
1885
+ ],
1886
+ "scale_bits": 4,
1887
+ "scale_groups:": 32
1888
+ },
1889
+ "up_proj": {
1890
+ "group_size": {
1891
+ "4": 128
1892
+ },
1893
+ "bits": [
1894
+ 4
1895
+ ],
1896
+ "bits_prop": [
1897
+ 1
1898
+ ],
1899
+ "scale_bits": 4,
1900
+ "scale_groups:": 32
1901
+ }
1902
+ },
1903
+ "model.layers.20": {
1904
+ "accuracy": 0.9766635894775391,
1905
+ "total_bits": 314302464.0,
1906
+ "o_proj": {
1907
+ "group_size": {
1908
+ "4": 128
1909
+ },
1910
+ "bits": [
1911
+ 4
1912
+ ],
1913
+ "bits_prop": [
1914
+ 1
1915
+ ],
1916
+ "scale_bits": 4,
1917
+ "scale_groups:": 32
1918
+ },
1919
+ "down_proj": {
1920
+ "group_size": {
1921
+ "4": 128
1922
+ },
1923
+ "bits": [
1924
+ 4
1925
+ ],
1926
+ "bits_prop": [
1927
+ 1
1928
+ ],
1929
+ "scale_bits": 4,
1930
+ "scale_groups:": 32
1931
+ },
1932
+ "q_proj": {
1933
+ "group_size": {
1934
+ "4": 128
1935
+ },
1936
+ "bits": [
1937
+ 4
1938
+ ],
1939
+ "bits_prop": [
1940
+ 1
1941
+ ],
1942
+ "scale_bits": 4,
1943
+ "scale_groups:": 32
1944
+ },
1945
+ "k_proj": {
1946
+ "group_size": {
1947
+ "4": 128
1948
+ },
1949
+ "bits": [
1950
+ 4
1951
+ ],
1952
+ "bits_prop": [
1953
+ 1
1954
+ ],
1955
+ "scale_bits": 4,
1956
+ "scale_groups:": 32
1957
+ },
1958
+ "v_proj": {
1959
+ "group_size": {
1960
+ "4": 128
1961
+ },
1962
+ "bits": [
1963
+ 4
1964
+ ],
1965
+ "bits_prop": [
1966
+ 1
1967
+ ],
1968
+ "scale_bits": 4,
1969
+ "scale_groups:": 32
1970
+ },
1971
+ "gate_proj": {
1972
+ "group_size": {
1973
+ "4": 128
1974
+ },
1975
+ "bits": [
1976
+ 4
1977
+ ],
1978
+ "bits_prop": [
1979
+ 1
1980
+ ],
1981
+ "scale_bits": 4,
1982
+ "scale_groups:": 32
1983
+ },
1984
+ "up_proj": {
1985
+ "group_size": {
1986
+ "4": 128
1987
+ },
1988
+ "bits": [
1989
+ 4
1990
+ ],
1991
+ "bits_prop": [
1992
+ 1
1993
+ ],
1994
+ "scale_bits": 4,
1995
+ "scale_groups:": 32
1996
+ }
1997
+ },
1998
+ "model.layers.21": {
1999
+ "accuracy": 0.9790831208229065,
2000
+ "total_bits": 314302464.0,
2001
+ "o_proj": {
2002
+ "group_size": {
2003
+ "4": 128
2004
+ },
2005
+ "bits": [
2006
+ 4
2007
+ ],
2008
+ "bits_prop": [
2009
+ 1
2010
+ ],
2011
+ "scale_bits": 4,
2012
+ "scale_groups:": 32
2013
+ },
2014
+ "down_proj": {
2015
+ "group_size": {
2016
+ "4": 128
2017
+ },
2018
+ "bits": [
2019
+ 4
2020
+ ],
2021
+ "bits_prop": [
2022
+ 1
2023
+ ],
2024
+ "scale_bits": 4,
2025
+ "scale_groups:": 32
2026
+ },
2027
+ "q_proj": {
2028
+ "group_size": {
2029
+ "4": 128
2030
+ },
2031
+ "bits": [
2032
+ 4
2033
+ ],
2034
+ "bits_prop": [
2035
+ 1
2036
+ ],
2037
+ "scale_bits": 4,
2038
+ "scale_groups:": 32
2039
+ },
2040
+ "k_proj": {
2041
+ "group_size": {
2042
+ "4": 128
2043
+ },
2044
+ "bits": [
2045
+ 4
2046
+ ],
2047
+ "bits_prop": [
2048
+ 1
2049
+ ],
2050
+ "scale_bits": 4,
2051
+ "scale_groups:": 32
2052
+ },
2053
+ "v_proj": {
2054
+ "group_size": {
2055
+ "4": 128
2056
+ },
2057
+ "bits": [
2058
+ 4
2059
+ ],
2060
+ "bits_prop": [
2061
+ 1
2062
+ ],
2063
+ "scale_bits": 4,
2064
+ "scale_groups:": 32
2065
+ },
2066
+ "gate_proj": {
2067
+ "group_size": {
2068
+ "4": 128
2069
+ },
2070
+ "bits": [
2071
+ 4
2072
+ ],
2073
+ "bits_prop": [
2074
+ 1
2075
+ ],
2076
+ "scale_bits": 4,
2077
+ "scale_groups:": 32
2078
+ },
2079
+ "up_proj": {
2080
+ "group_size": {
2081
+ "4": 128
2082
+ },
2083
+ "bits": [
2084
+ 4
2085
+ ],
2086
+ "bits_prop": [
2087
+ 1
2088
+ ],
2089
+ "scale_bits": 4,
2090
+ "scale_groups:": 32
2091
+ }
2092
+ },
2093
+ "model.layers.22": {
2094
+ "accuracy": 0.9801523685455322,
2095
+ "total_bits": 314302464.0,
2096
+ "o_proj": {
2097
+ "group_size": {
2098
+ "4": 128
2099
+ },
2100
+ "bits": [
2101
+ 4
2102
+ ],
2103
+ "bits_prop": [
2104
+ 1
2105
+ ],
2106
+ "scale_bits": 4,
2107
+ "scale_groups:": 32
2108
+ },
2109
+ "down_proj": {
2110
+ "group_size": {
2111
+ "4": 128
2112
+ },
2113
+ "bits": [
2114
+ 4
2115
+ ],
2116
+ "bits_prop": [
2117
+ 1
2118
+ ],
2119
+ "scale_bits": 4,
2120
+ "scale_groups:": 32
2121
+ },
2122
+ "q_proj": {
2123
+ "group_size": {
2124
+ "4": 128
2125
+ },
2126
+ "bits": [
2127
+ 4
2128
+ ],
2129
+ "bits_prop": [
2130
+ 1
2131
+ ],
2132
+ "scale_bits": 4,
2133
+ "scale_groups:": 32
2134
+ },
2135
+ "k_proj": {
2136
+ "group_size": {
2137
+ "4": 128
2138
+ },
2139
+ "bits": [
2140
+ 4
2141
+ ],
2142
+ "bits_prop": [
2143
+ 1
2144
+ ],
2145
+ "scale_bits": 4,
2146
+ "scale_groups:": 32
2147
+ },
2148
+ "v_proj": {
2149
+ "group_size": {
2150
+ "4": 128
2151
+ },
2152
+ "bits": [
2153
+ 4
2154
+ ],
2155
+ "bits_prop": [
2156
+ 1
2157
+ ],
2158
+ "scale_bits": 4,
2159
+ "scale_groups:": 32
2160
+ },
2161
+ "gate_proj": {
2162
+ "group_size": {
2163
+ "4": 128
2164
+ },
2165
+ "bits": [
2166
+ 4
2167
+ ],
2168
+ "bits_prop": [
2169
+ 1
2170
+ ],
2171
+ "scale_bits": 4,
2172
+ "scale_groups:": 32
2173
+ },
2174
+ "up_proj": {
2175
+ "group_size": {
2176
+ "4": 128
2177
+ },
2178
+ "bits": [
2179
+ 4
2180
+ ],
2181
+ "bits_prop": [
2182
+ 1
2183
+ ],
2184
+ "scale_bits": 4,
2185
+ "scale_groups:": 32
2186
+ }
2187
+ },
2188
+ "model.layers.23": {
2189
+ "accuracy": 0.9790130853652954,
2190
+ "total_bits": 314302464.0,
2191
+ "o_proj": {
2192
+ "group_size": {
2193
+ "4": 128
2194
+ },
2195
+ "bits": [
2196
+ 4
2197
+ ],
2198
+ "bits_prop": [
2199
+ 1
2200
+ ],
2201
+ "scale_bits": 4,
2202
+ "scale_groups:": 32
2203
+ },
2204
+ "down_proj": {
2205
+ "group_size": {
2206
+ "4": 128
2207
+ },
2208
+ "bits": [
2209
+ 4
2210
+ ],
2211
+ "bits_prop": [
2212
+ 1
2213
+ ],
2214
+ "scale_bits": 4,
2215
+ "scale_groups:": 32
2216
+ },
2217
+ "q_proj": {
2218
+ "group_size": {
2219
+ "4": 128
2220
+ },
2221
+ "bits": [
2222
+ 4
2223
+ ],
2224
+ "bits_prop": [
2225
+ 1
2226
+ ],
2227
+ "scale_bits": 4,
2228
+ "scale_groups:": 32
2229
+ },
2230
+ "k_proj": {
2231
+ "group_size": {
2232
+ "4": 128
2233
+ },
2234
+ "bits": [
2235
+ 4
2236
+ ],
2237
+ "bits_prop": [
2238
+ 1
2239
+ ],
2240
+ "scale_bits": 4,
2241
+ "scale_groups:": 32
2242
+ },
2243
+ "v_proj": {
2244
+ "group_size": {
2245
+ "4": 128
2246
+ },
2247
+ "bits": [
2248
+ 4
2249
+ ],
2250
+ "bits_prop": [
2251
+ 1
2252
+ ],
2253
+ "scale_bits": 4,
2254
+ "scale_groups:": 32
2255
+ },
2256
+ "gate_proj": {
2257
+ "group_size": {
2258
+ "4": 128
2259
+ },
2260
+ "bits": [
2261
+ 4
2262
+ ],
2263
+ "bits_prop": [
2264
+ 1
2265
+ ],
2266
+ "scale_bits": 4,
2267
+ "scale_groups:": 32
2268
+ },
2269
+ "up_proj": {
2270
+ "group_size": {
2271
+ "4": 128
2272
+ },
2273
+ "bits": [
2274
+ 4
2275
+ ],
2276
+ "bits_prop": [
2277
+ 1
2278
+ ],
2279
+ "scale_bits": 4,
2280
+ "scale_groups:": 32
2281
+ }
2282
+ },
2283
+ "model.layers.24": {
2284
+ "accuracy": 0.9794877767562866,
2285
+ "total_bits": 314302464.0,
2286
+ "o_proj": {
2287
+ "group_size": {
2288
+ "4": 128
2289
+ },
2290
+ "bits": [
2291
+ 4
2292
+ ],
2293
+ "bits_prop": [
2294
+ 1
2295
+ ],
2296
+ "scale_bits": 4,
2297
+ "scale_groups:": 32
2298
+ },
2299
+ "down_proj": {
2300
+ "group_size": {
2301
+ "4": 128
2302
+ },
2303
+ "bits": [
2304
+ 4
2305
+ ],
2306
+ "bits_prop": [
2307
+ 1
2308
+ ],
2309
+ "scale_bits": 4,
2310
+ "scale_groups:": 32
2311
+ },
2312
+ "q_proj": {
2313
+ "group_size": {
2314
+ "4": 128
2315
+ },
2316
+ "bits": [
2317
+ 4
2318
+ ],
2319
+ "bits_prop": [
2320
+ 1
2321
+ ],
2322
+ "scale_bits": 4,
2323
+ "scale_groups:": 32
2324
+ },
2325
+ "k_proj": {
2326
+ "group_size": {
2327
+ "4": 128
2328
+ },
2329
+ "bits": [
2330
+ 4
2331
+ ],
2332
+ "bits_prop": [
2333
+ 1
2334
+ ],
2335
+ "scale_bits": 4,
2336
+ "scale_groups:": 32
2337
+ },
2338
+ "v_proj": {
2339
+ "group_size": {
2340
+ "4": 128
2341
+ },
2342
+ "bits": [
2343
+ 4
2344
+ ],
2345
+ "bits_prop": [
2346
+ 1
2347
+ ],
2348
+ "scale_bits": 4,
2349
+ "scale_groups:": 32
2350
+ },
2351
+ "gate_proj": {
2352
+ "group_size": {
2353
+ "4": 128
2354
+ },
2355
+ "bits": [
2356
+ 4
2357
+ ],
2358
+ "bits_prop": [
2359
+ 1
2360
+ ],
2361
+ "scale_bits": 4,
2362
+ "scale_groups:": 32
2363
+ },
2364
+ "up_proj": {
2365
+ "group_size": {
2366
+ "4": 128
2367
+ },
2368
+ "bits": [
2369
+ 4
2370
+ ],
2371
+ "bits_prop": [
2372
+ 1
2373
+ ],
2374
+ "scale_bits": 4,
2375
+ "scale_groups:": 32
2376
+ }
2377
+ },
2378
+ "model.layers.25": {
2379
+ "accuracy": 0.9761180877685547,
2380
+ "total_bits": 314302464.0,
2381
+ "o_proj": {
2382
+ "group_size": {
2383
+ "4": 128
2384
+ },
2385
+ "bits": [
2386
+ 4
2387
+ ],
2388
+ "bits_prop": [
2389
+ 1
2390
+ ],
2391
+ "scale_bits": 4,
2392
+ "scale_groups:": 32
2393
+ },
2394
+ "down_proj": {
2395
+ "group_size": {
2396
+ "4": 128
2397
+ },
2398
+ "bits": [
2399
+ 4
2400
+ ],
2401
+ "bits_prop": [
2402
+ 1
2403
+ ],
2404
+ "scale_bits": 4,
2405
+ "scale_groups:": 32
2406
+ },
2407
+ "q_proj": {
2408
+ "group_size": {
2409
+ "4": 128
2410
+ },
2411
+ "bits": [
2412
+ 4
2413
+ ],
2414
+ "bits_prop": [
2415
+ 1
2416
+ ],
2417
+ "scale_bits": 4,
2418
+ "scale_groups:": 32
2419
+ },
2420
+ "k_proj": {
2421
+ "group_size": {
2422
+ "4": 128
2423
+ },
2424
+ "bits": [
2425
+ 4
2426
+ ],
2427
+ "bits_prop": [
2428
+ 1
2429
+ ],
2430
+ "scale_bits": 4,
2431
+ "scale_groups:": 32
2432
+ },
2433
+ "v_proj": {
2434
+ "group_size": {
2435
+ "4": 128
2436
+ },
2437
+ "bits": [
2438
+ 4
2439
+ ],
2440
+ "bits_prop": [
2441
+ 1
2442
+ ],
2443
+ "scale_bits": 4,
2444
+ "scale_groups:": 32
2445
+ },
2446
+ "gate_proj": {
2447
+ "group_size": {
2448
+ "4": 128
2449
+ },
2450
+ "bits": [
2451
+ 4
2452
+ ],
2453
+ "bits_prop": [
2454
+ 1
2455
+ ],
2456
+ "scale_bits": 4,
2457
+ "scale_groups:": 32
2458
+ },
2459
+ "up_proj": {
2460
+ "group_size": {
2461
+ "4": 128
2462
+ },
2463
+ "bits": [
2464
+ 4
2465
+ ],
2466
+ "bits_prop": [
2467
+ 1
2468
+ ],
2469
+ "scale_bits": 4,
2470
+ "scale_groups:": 32
2471
+ }
2472
+ },
2473
+ "model.layers.26": {
2474
+ "accuracy": 0.9733923673629761,
2475
+ "total_bits": 314302464.0,
2476
+ "o_proj": {
2477
+ "group_size": {
2478
+ "4": 128
2479
+ },
2480
+ "bits": [
2481
+ 4
2482
+ ],
2483
+ "bits_prop": [
2484
+ 1
2485
+ ],
2486
+ "scale_bits": 4,
2487
+ "scale_groups:": 32
2488
+ },
2489
+ "down_proj": {
2490
+ "group_size": {
2491
+ "4": 128
2492
+ },
2493
+ "bits": [
2494
+ 4
2495
+ ],
2496
+ "bits_prop": [
2497
+ 1
2498
+ ],
2499
+ "scale_bits": 4,
2500
+ "scale_groups:": 32
2501
+ },
2502
+ "q_proj": {
2503
+ "group_size": {
2504
+ "4": 128
2505
+ },
2506
+ "bits": [
2507
+ 4
2508
+ ],
2509
+ "bits_prop": [
2510
+ 1
2511
+ ],
2512
+ "scale_bits": 4,
2513
+ "scale_groups:": 32
2514
+ },
2515
+ "k_proj": {
2516
+ "group_size": {
2517
+ "4": 128
2518
+ },
2519
+ "bits": [
2520
+ 4
2521
+ ],
2522
+ "bits_prop": [
2523
+ 1
2524
+ ],
2525
+ "scale_bits": 4,
2526
+ "scale_groups:": 32
2527
+ },
2528
+ "v_proj": {
2529
+ "group_size": {
2530
+ "4": 128
2531
+ },
2532
+ "bits": [
2533
+ 4
2534
+ ],
2535
+ "bits_prop": [
2536
+ 1
2537
+ ],
2538
+ "scale_bits": 4,
2539
+ "scale_groups:": 32
2540
+ },
2541
+ "gate_proj": {
2542
+ "group_size": {
2543
+ "4": 128
2544
+ },
2545
+ "bits": [
2546
+ 4
2547
+ ],
2548
+ "bits_prop": [
2549
+ 1
2550
+ ],
2551
+ "scale_bits": 4,
2552
+ "scale_groups:": 32
2553
+ },
2554
+ "up_proj": {
2555
+ "group_size": {
2556
+ "4": 128
2557
+ },
2558
+ "bits": [
2559
+ 4
2560
+ ],
2561
+ "bits_prop": [
2562
+ 1
2563
+ ],
2564
+ "scale_bits": 4,
2565
+ "scale_groups:": 32
2566
+ }
2567
+ },
2568
+ "model.layers.27": {
2569
+ "accuracy": 0.9720484018325806,
2570
+ "total_bits": 314302464.0,
2571
+ "o_proj": {
2572
+ "group_size": {
2573
+ "4": 128
2574
+ },
2575
+ "bits": [
2576
+ 4
2577
+ ],
2578
+ "bits_prop": [
2579
+ 1
2580
+ ],
2581
+ "scale_bits": 4,
2582
+ "scale_groups:": 32
2583
+ },
2584
+ "down_proj": {
2585
+ "group_size": {
2586
+ "4": 128
2587
+ },
2588
+ "bits": [
2589
+ 4
2590
+ ],
2591
+ "bits_prop": [
2592
+ 1
2593
+ ],
2594
+ "scale_bits": 4,
2595
+ "scale_groups:": 32
2596
+ },
2597
+ "q_proj": {
2598
+ "group_size": {
2599
+ "4": 128
2600
+ },
2601
+ "bits": [
2602
+ 4
2603
+ ],
2604
+ "bits_prop": [
2605
+ 1
2606
+ ],
2607
+ "scale_bits": 4,
2608
+ "scale_groups:": 32
2609
+ },
2610
+ "k_proj": {
2611
+ "group_size": {
2612
+ "4": 128
2613
+ },
2614
+ "bits": [
2615
+ 4
2616
+ ],
2617
+ "bits_prop": [
2618
+ 1
2619
+ ],
2620
+ "scale_bits": 4,
2621
+ "scale_groups:": 32
2622
+ },
2623
+ "v_proj": {
2624
+ "group_size": {
2625
+ "4": 128
2626
+ },
2627
+ "bits": [
2628
+ 4
2629
+ ],
2630
+ "bits_prop": [
2631
+ 1
2632
+ ],
2633
+ "scale_bits": 4,
2634
+ "scale_groups:": 32
2635
+ },
2636
+ "gate_proj": {
2637
+ "group_size": {
2638
+ "4": 128
2639
+ },
2640
+ "bits": [
2641
+ 4
2642
+ ],
2643
+ "bits_prop": [
2644
+ 1
2645
+ ],
2646
+ "scale_bits": 4,
2647
+ "scale_groups:": 32
2648
+ },
2649
+ "up_proj": {
2650
+ "group_size": {
2651
+ "4": 128
2652
+ },
2653
+ "bits": [
2654
+ 4
2655
+ ],
2656
+ "bits_prop": [
2657
+ 1
2658
+ ],
2659
+ "scale_bits": 4,
2660
+ "scale_groups:": 32
2661
+ }
2662
+ },
2663
+ "model.layers.28": {
2664
+ "accuracy": 0.9757145643234253,
2665
+ "total_bits": 314302464.0,
2666
+ "o_proj": {
2667
+ "group_size": {
2668
+ "4": 128
2669
+ },
2670
+ "bits": [
2671
+ 4
2672
+ ],
2673
+ "bits_prop": [
2674
+ 1
2675
+ ],
2676
+ "scale_bits": 4,
2677
+ "scale_groups:": 32
2678
+ },
2679
+ "down_proj": {
2680
+ "group_size": {
2681
+ "4": 128
2682
+ },
2683
+ "bits": [
2684
+ 4
2685
+ ],
2686
+ "bits_prop": [
2687
+ 1
2688
+ ],
2689
+ "scale_bits": 4,
2690
+ "scale_groups:": 32
2691
+ },
2692
+ "q_proj": {
2693
+ "group_size": {
2694
+ "4": 128
2695
+ },
2696
+ "bits": [
2697
+ 4
2698
+ ],
2699
+ "bits_prop": [
2700
+ 1
2701
+ ],
2702
+ "scale_bits": 4,
2703
+ "scale_groups:": 32
2704
+ },
2705
+ "k_proj": {
2706
+ "group_size": {
2707
+ "4": 128
2708
+ },
2709
+ "bits": [
2710
+ 4
2711
+ ],
2712
+ "bits_prop": [
2713
+ 1
2714
+ ],
2715
+ "scale_bits": 4,
2716
+ "scale_groups:": 32
2717
+ },
2718
+ "v_proj": {
2719
+ "group_size": {
2720
+ "4": 128
2721
+ },
2722
+ "bits": [
2723
+ 4
2724
+ ],
2725
+ "bits_prop": [
2726
+ 1
2727
+ ],
2728
+ "scale_bits": 4,
2729
+ "scale_groups:": 32
2730
+ },
2731
+ "gate_proj": {
2732
+ "group_size": {
2733
+ "4": 128
2734
+ },
2735
+ "bits": [
2736
+ 4
2737
+ ],
2738
+ "bits_prop": [
2739
+ 1
2740
+ ],
2741
+ "scale_bits": 4,
2742
+ "scale_groups:": 32
2743
+ },
2744
+ "up_proj": {
2745
+ "group_size": {
2746
+ "4": 128
2747
+ },
2748
+ "bits": [
2749
+ 4
2750
+ ],
2751
+ "bits_prop": [
2752
+ 1
2753
+ ],
2754
+ "scale_bits": 4,
2755
+ "scale_groups:": 32
2756
+ }
2757
+ },
2758
+ "model.layers.29": {
2759
+ "accuracy": 0.975440263748169,
2760
+ "total_bits": 314302464.0,
2761
+ "o_proj": {
2762
+ "group_size": {
2763
+ "4": 128
2764
+ },
2765
+ "bits": [
2766
+ 4
2767
+ ],
2768
+ "bits_prop": [
2769
+ 1
2770
+ ],
2771
+ "scale_bits": 4,
2772
+ "scale_groups:": 32
2773
+ },
2774
+ "down_proj": {
2775
+ "group_size": {
2776
+ "4": 128
2777
+ },
2778
+ "bits": [
2779
+ 4
2780
+ ],
2781
+ "bits_prop": [
2782
+ 1
2783
+ ],
2784
+ "scale_bits": 4,
2785
+ "scale_groups:": 32
2786
+ },
2787
+ "q_proj": {
2788
+ "group_size": {
2789
+ "4": 128
2790
+ },
2791
+ "bits": [
2792
+ 4
2793
+ ],
2794
+ "bits_prop": [
2795
+ 1
2796
+ ],
2797
+ "scale_bits": 4,
2798
+ "scale_groups:": 32
2799
+ },
2800
+ "k_proj": {
2801
+ "group_size": {
2802
+ "4": 128
2803
+ },
2804
+ "bits": [
2805
+ 4
2806
+ ],
2807
+ "bits_prop": [
2808
+ 1
2809
+ ],
2810
+ "scale_bits": 4,
2811
+ "scale_groups:": 32
2812
+ },
2813
+ "v_proj": {
2814
+ "group_size": {
2815
+ "4": 128
2816
+ },
2817
+ "bits": [
2818
+ 4
2819
+ ],
2820
+ "bits_prop": [
2821
+ 1
2822
+ ],
2823
+ "scale_bits": 4,
2824
+ "scale_groups:": 32
2825
+ },
2826
+ "gate_proj": {
2827
+ "group_size": {
2828
+ "4": 128
2829
+ },
2830
+ "bits": [
2831
+ 4
2832
+ ],
2833
+ "bits_prop": [
2834
+ 1
2835
+ ],
2836
+ "scale_bits": 4,
2837
+ "scale_groups:": 32
2838
+ },
2839
+ "up_proj": {
2840
+ "group_size": {
2841
+ "4": 128
2842
+ },
2843
+ "bits": [
2844
+ 4
2845
+ ],
2846
+ "bits_prop": [
2847
+ 1
2848
+ ],
2849
+ "scale_bits": 4,
2850
+ "scale_groups:": 32
2851
+ }
2852
+ },
2853
+ "model.layers.30": {
2854
+ "accuracy": 0.9650356769561768,
2855
+ "total_bits": 314302464.0,
2856
+ "o_proj": {
2857
+ "group_size": {
2858
+ "4": 128
2859
+ },
2860
+ "bits": [
2861
+ 4
2862
+ ],
2863
+ "bits_prop": [
2864
+ 1
2865
+ ],
2866
+ "scale_bits": 4,
2867
+ "scale_groups:": 32
2868
+ },
2869
+ "down_proj": {
2870
+ "group_size": {
2871
+ "4": 128
2872
+ },
2873
+ "bits": [
2874
+ 4
2875
+ ],
2876
+ "bits_prop": [
2877
+ 1
2878
+ ],
2879
+ "scale_bits": 4,
2880
+ "scale_groups:": 32
2881
+ },
2882
+ "q_proj": {
2883
+ "group_size": {
2884
+ "4": 128
2885
+ },
2886
+ "bits": [
2887
+ 4
2888
+ ],
2889
+ "bits_prop": [
2890
+ 1
2891
+ ],
2892
+ "scale_bits": 4,
2893
+ "scale_groups:": 32
2894
+ },
2895
+ "k_proj": {
2896
+ "group_size": {
2897
+ "4": 128
2898
+ },
2899
+ "bits": [
2900
+ 4
2901
+ ],
2902
+ "bits_prop": [
2903
+ 1
2904
+ ],
2905
+ "scale_bits": 4,
2906
+ "scale_groups:": 32
2907
+ },
2908
+ "v_proj": {
2909
+ "group_size": {
2910
+ "4": 128
2911
+ },
2912
+ "bits": [
2913
+ 4
2914
+ ],
2915
+ "bits_prop": [
2916
+ 1
2917
+ ],
2918
+ "scale_bits": 4,
2919
+ "scale_groups:": 32
2920
+ },
2921
+ "gate_proj": {
2922
+ "group_size": {
2923
+ "4": 128
2924
+ },
2925
+ "bits": [
2926
+ 4
2927
+ ],
2928
+ "bits_prop": [
2929
+ 1
2930
+ ],
2931
+ "scale_bits": 4,
2932
+ "scale_groups:": 32
2933
+ },
2934
+ "up_proj": {
2935
+ "group_size": {
2936
+ "4": 128
2937
+ },
2938
+ "bits": [
2939
+ 4
2940
+ ],
2941
+ "bits_prop": [
2942
+ 1
2943
+ ],
2944
+ "scale_bits": 4,
2945
+ "scale_groups:": 32
2946
+ }
2947
+ },
2948
+ "model.layers.31": {
2949
+ "accuracy": 0.9708733558654785,
2950
+ "total_bits": 314302464.0,
2951
+ "o_proj": {
2952
+ "group_size": {
2953
+ "4": 128
2954
+ },
2955
+ "bits": [
2956
+ 4
2957
+ ],
2958
+ "bits_prop": [
2959
+ 1
2960
+ ],
2961
+ "scale_bits": 4,
2962
+ "scale_groups:": 32
2963
+ },
2964
+ "down_proj": {
2965
+ "group_size": {
2966
+ "4": 128
2967
+ },
2968
+ "bits": [
2969
+ 4
2970
+ ],
2971
+ "bits_prop": [
2972
+ 1
2973
+ ],
2974
+ "scale_bits": 4,
2975
+ "scale_groups:": 32
2976
+ },
2977
+ "q_proj": {
2978
+ "group_size": {
2979
+ "4": 128
2980
+ },
2981
+ "bits": [
2982
+ 4
2983
+ ],
2984
+ "bits_prop": [
2985
+ 1
2986
+ ],
2987
+ "scale_bits": 4,
2988
+ "scale_groups:": 32
2989
+ },
2990
+ "k_proj": {
2991
+ "group_size": {
2992
+ "4": 128
2993
+ },
2994
+ "bits": [
2995
+ 4
2996
+ ],
2997
+ "bits_prop": [
2998
+ 1
2999
+ ],
3000
+ "scale_bits": 4,
3001
+ "scale_groups:": 32
3002
+ },
3003
+ "v_proj": {
3004
+ "group_size": {
3005
+ "4": 128
3006
+ },
3007
+ "bits": [
3008
+ 4
3009
+ ],
3010
+ "bits_prop": [
3011
+ 1
3012
+ ],
3013
+ "scale_bits": 4,
3014
+ "scale_groups:": 32
3015
+ },
3016
+ "gate_proj": {
3017
+ "group_size": {
3018
+ "4": 128
3019
+ },
3020
+ "bits": [
3021
+ 4
3022
+ ],
3023
+ "bits_prop": [
3024
+ 1
3025
+ ],
3026
+ "scale_bits": 4,
3027
+ "scale_groups:": 32
3028
+ },
3029
+ "up_proj": {
3030
+ "group_size": {
3031
+ "4": 128
3032
+ },
3033
+ "bits": [
3034
+ 4
3035
+ ],
3036
+ "bits_prop": [
3037
+ 1
3038
+ ],
3039
+ "scale_bits": 4,
3040
+ "scale_groups:": 32
3041
+ }
3042
+ },
3043
+ "model.layers.32": {
3044
+ "accuracy": 0.9723145961761475,
3045
+ "total_bits": 314302464.0,
3046
+ "o_proj": {
3047
+ "group_size": {
3048
+ "4": 128
3049
+ },
3050
+ "bits": [
3051
+ 4
3052
+ ],
3053
+ "bits_prop": [
3054
+ 1
3055
+ ],
3056
+ "scale_bits": 4,
3057
+ "scale_groups:": 32
3058
+ },
3059
+ "down_proj": {
3060
+ "group_size": {
3061
+ "4": 128
3062
+ },
3063
+ "bits": [
3064
+ 4
3065
+ ],
3066
+ "bits_prop": [
3067
+ 1
3068
+ ],
3069
+ "scale_bits": 4,
3070
+ "scale_groups:": 32
3071
+ },
3072
+ "q_proj": {
3073
+ "group_size": {
3074
+ "4": 128
3075
+ },
3076
+ "bits": [
3077
+ 4
3078
+ ],
3079
+ "bits_prop": [
3080
+ 1
3081
+ ],
3082
+ "scale_bits": 4,
3083
+ "scale_groups:": 32
3084
+ },
3085
+ "k_proj": {
3086
+ "group_size": {
3087
+ "4": 128
3088
+ },
3089
+ "bits": [
3090
+ 4
3091
+ ],
3092
+ "bits_prop": [
3093
+ 1
3094
+ ],
3095
+ "scale_bits": 4,
3096
+ "scale_groups:": 32
3097
+ },
3098
+ "v_proj": {
3099
+ "group_size": {
3100
+ "4": 128
3101
+ },
3102
+ "bits": [
3103
+ 4
3104
+ ],
3105
+ "bits_prop": [
3106
+ 1
3107
+ ],
3108
+ "scale_bits": 4,
3109
+ "scale_groups:": 32
3110
+ },
3111
+ "gate_proj": {
3112
+ "group_size": {
3113
+ "4": 128
3114
+ },
3115
+ "bits": [
3116
+ 4
3117
+ ],
3118
+ "bits_prop": [
3119
+ 1
3120
+ ],
3121
+ "scale_bits": 4,
3122
+ "scale_groups:": 32
3123
+ },
3124
+ "up_proj": {
3125
+ "group_size": {
3126
+ "4": 128
3127
+ },
3128
+ "bits": [
3129
+ 4
3130
+ ],
3131
+ "bits_prop": [
3132
+ 1
3133
+ ],
3134
+ "scale_bits": 4,
3135
+ "scale_groups:": 32
3136
+ }
3137
+ },
3138
+ "model.layers.33": {
3139
+ "accuracy": 0.9625141620635986,
3140
+ "total_bits": 314302464.0,
3141
+ "o_proj": {
3142
+ "group_size": {
3143
+ "4": 128
3144
+ },
3145
+ "bits": [
3146
+ 4
3147
+ ],
3148
+ "bits_prop": [
3149
+ 1
3150
+ ],
3151
+ "scale_bits": 4,
3152
+ "scale_groups:": 32
3153
+ },
3154
+ "down_proj": {
3155
+ "group_size": {
3156
+ "4": 128
3157
+ },
3158
+ "bits": [
3159
+ 4
3160
+ ],
3161
+ "bits_prop": [
3162
+ 1
3163
+ ],
3164
+ "scale_bits": 4,
3165
+ "scale_groups:": 32
3166
+ },
3167
+ "q_proj": {
3168
+ "group_size": {
3169
+ "4": 128
3170
+ },
3171
+ "bits": [
3172
+ 4
3173
+ ],
3174
+ "bits_prop": [
3175
+ 1
3176
+ ],
3177
+ "scale_bits": 4,
3178
+ "scale_groups:": 32
3179
+ },
3180
+ "k_proj": {
3181
+ "group_size": {
3182
+ "4": 128
3183
+ },
3184
+ "bits": [
3185
+ 4
3186
+ ],
3187
+ "bits_prop": [
3188
+ 1
3189
+ ],
3190
+ "scale_bits": 4,
3191
+ "scale_groups:": 32
3192
+ },
3193
+ "v_proj": {
3194
+ "group_size": {
3195
+ "4": 128
3196
+ },
3197
+ "bits": [
3198
+ 4
3199
+ ],
3200
+ "bits_prop": [
3201
+ 1
3202
+ ],
3203
+ "scale_bits": 4,
3204
+ "scale_groups:": 32
3205
+ },
3206
+ "gate_proj": {
3207
+ "group_size": {
3208
+ "4": 128
3209
+ },
3210
+ "bits": [
3211
+ 4
3212
+ ],
3213
+ "bits_prop": [
3214
+ 1
3215
+ ],
3216
+ "scale_bits": 4,
3217
+ "scale_groups:": 32
3218
+ },
3219
+ "up_proj": {
3220
+ "group_size": {
3221
+ "4": 128
3222
+ },
3223
+ "bits": [
3224
+ 4
3225
+ ],
3226
+ "bits_prop": [
3227
+ 1
3228
+ ],
3229
+ "scale_bits": 4,
3230
+ "scale_groups:": 32
3231
+ }
3232
+ },
3233
+ "model.layers.34": {
3234
+ "accuracy": 0.9762517213821411,
3235
+ "total_bits": 314302464.0,
3236
+ "o_proj": {
3237
+ "group_size": {
3238
+ "4": 128
3239
+ },
3240
+ "bits": [
3241
+ 4
3242
+ ],
3243
+ "bits_prop": [
3244
+ 1
3245
+ ],
3246
+ "scale_bits": 4,
3247
+ "scale_groups:": 32
3248
+ },
3249
+ "down_proj": {
3250
+ "group_size": {
3251
+ "4": 128
3252
+ },
3253
+ "bits": [
3254
+ 4
3255
+ ],
3256
+ "bits_prop": [
3257
+ 1
3258
+ ],
3259
+ "scale_bits": 4,
3260
+ "scale_groups:": 32
3261
+ },
3262
+ "q_proj": {
3263
+ "group_size": {
3264
+ "4": 128
3265
+ },
3266
+ "bits": [
3267
+ 4
3268
+ ],
3269
+ "bits_prop": [
3270
+ 1
3271
+ ],
3272
+ "scale_bits": 4,
3273
+ "scale_groups:": 32
3274
+ },
3275
+ "k_proj": {
3276
+ "group_size": {
3277
+ "4": 128
3278
+ },
3279
+ "bits": [
3280
+ 4
3281
+ ],
3282
+ "bits_prop": [
3283
+ 1
3284
+ ],
3285
+ "scale_bits": 4,
3286
+ "scale_groups:": 32
3287
+ },
3288
+ "v_proj": {
3289
+ "group_size": {
3290
+ "4": 128
3291
+ },
3292
+ "bits": [
3293
+ 4
3294
+ ],
3295
+ "bits_prop": [
3296
+ 1
3297
+ ],
3298
+ "scale_bits": 4,
3299
+ "scale_groups:": 32
3300
+ },
3301
+ "gate_proj": {
3302
+ "group_size": {
3303
+ "4": 128
3304
+ },
3305
+ "bits": [
3306
+ 4
3307
+ ],
3308
+ "bits_prop": [
3309
+ 1
3310
+ ],
3311
+ "scale_bits": 4,
3312
+ "scale_groups:": 32
3313
+ },
3314
+ "up_proj": {
3315
+ "group_size": {
3316
+ "4": 128
3317
+ },
3318
+ "bits": [
3319
+ 4
3320
+ ],
3321
+ "bits_prop": [
3322
+ 1
3323
+ ],
3324
+ "scale_bits": 4,
3325
+ "scale_groups:": 32
3326
+ }
3327
+ },
3328
+ "model.layers.35": {
3329
+ "accuracy": 0.9684627056121826,
3330
+ "total_bits": 314302464.0,
3331
+ "o_proj": {
3332
+ "group_size": {
3333
+ "4": 128
3334
+ },
3335
+ "bits": [
3336
+ 4
3337
+ ],
3338
+ "bits_prop": [
3339
+ 1
3340
+ ],
3341
+ "scale_bits": 4,
3342
+ "scale_groups:": 32
3343
+ },
3344
+ "down_proj": {
3345
+ "group_size": {
3346
+ "4": 128
3347
+ },
3348
+ "bits": [
3349
+ 4
3350
+ ],
3351
+ "bits_prop": [
3352
+ 1
3353
+ ],
3354
+ "scale_bits": 4,
3355
+ "scale_groups:": 32
3356
+ },
3357
+ "q_proj": {
3358
+ "group_size": {
3359
+ "4": 128
3360
+ },
3361
+ "bits": [
3362
+ 4
3363
+ ],
3364
+ "bits_prop": [
3365
+ 1
3366
+ ],
3367
+ "scale_bits": 4,
3368
+ "scale_groups:": 32
3369
+ },
3370
+ "k_proj": {
3371
+ "group_size": {
3372
+ "4": 128
3373
+ },
3374
+ "bits": [
3375
+ 4
3376
+ ],
3377
+ "bits_prop": [
3378
+ 1
3379
+ ],
3380
+ "scale_bits": 4,
3381
+ "scale_groups:": 32
3382
+ },
3383
+ "v_proj": {
3384
+ "group_size": {
3385
+ "4": 128
3386
+ },
3387
+ "bits": [
3388
+ 4
3389
+ ],
3390
+ "bits_prop": [
3391
+ 1
3392
+ ],
3393
+ "scale_bits": 4,
3394
+ "scale_groups:": 32
3395
+ },
3396
+ "gate_proj": {
3397
+ "group_size": {
3398
+ "4": 128
3399
+ },
3400
+ "bits": [
3401
+ 4
3402
+ ],
3403
+ "bits_prop": [
3404
+ 1
3405
+ ],
3406
+ "scale_bits": 4,
3407
+ "scale_groups:": 32
3408
+ },
3409
+ "up_proj": {
3410
+ "group_size": {
3411
+ "4": 128
3412
+ },
3413
+ "bits": [
3414
+ 4
3415
+ ],
3416
+ "bits_prop": [
3417
+ 1
3418
+ ],
3419
+ "scale_bits": 4,
3420
+ "scale_groups:": 32
3421
+ }
3422
+ }
3423
+ }
3424
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>",
5
+ "<|object_ref_start|>",
6
+ "<|object_ref_end|>",
7
+ "<|box_start|>",
8
+ "<|box_end|>",
9
+ "<|quad_start|>",
10
+ "<|quad_end|>",
11
+ "<|vision_start|>",
12
+ "<|vision_end|>",
13
+ "<|vision_pad|>",
14
+ "<|image_pad|>",
15
+ "<|video_pad|>"
16
+ ],
17
+ "eos_token": {
18
+ "content": "<|im_end|>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ },
24
+ "pad_token": {
25
+ "content": "<|endoftext|>",
26
+ "lstrip": false,
27
+ "normalized": false,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ }
31
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,207 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "151643": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "151644": {
14
+ "content": "<|im_start|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "151645": {
22
+ "content": "<|im_end|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "151646": {
30
+ "content": "<|object_ref_start|>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "151647": {
38
+ "content": "<|object_ref_end|>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": true
44
+ },
45
+ "151648": {
46
+ "content": "<|box_start|>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": true
52
+ },
53
+ "151649": {
54
+ "content": "<|box_end|>",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": true
60
+ },
61
+ "151650": {
62
+ "content": "<|quad_start|>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": true
68
+ },
69
+ "151651": {
70
+ "content": "<|quad_end|>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": true
76
+ },
77
+ "151652": {
78
+ "content": "<|vision_start|>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": true
84
+ },
85
+ "151653": {
86
+ "content": "<|vision_end|>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": true
92
+ },
93
+ "151654": {
94
+ "content": "<|vision_pad|>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": true
100
+ },
101
+ "151655": {
102
+ "content": "<|image_pad|>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": true
108
+ },
109
+ "151656": {
110
+ "content": "<|video_pad|>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": true
116
+ },
117
+ "151657": {
118
+ "content": "<tool_call>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "151658": {
126
+ "content": "</tool_call>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "151659": {
134
+ "content": "<|fim_prefix|>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "151660": {
142
+ "content": "<|fim_middle|>",
143
+ "lstrip": false,
144
+ "normalized": false,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "151661": {
150
+ "content": "<|fim_suffix|>",
151
+ "lstrip": false,
152
+ "normalized": false,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "151662": {
158
+ "content": "<|fim_pad|>",
159
+ "lstrip": false,
160
+ "normalized": false,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "151663": {
166
+ "content": "<|repo_name|>",
167
+ "lstrip": false,
168
+ "normalized": false,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "151664": {
174
+ "content": "<|file_sep|>",
175
+ "lstrip": false,
176
+ "normalized": false,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ }
181
+ },
182
+ "additional_special_tokens": [
183
+ "<|im_start|>",
184
+ "<|im_end|>",
185
+ "<|object_ref_start|>",
186
+ "<|object_ref_end|>",
187
+ "<|box_start|>",
188
+ "<|box_end|>",
189
+ "<|quad_start|>",
190
+ "<|quad_end|>",
191
+ "<|vision_start|>",
192
+ "<|vision_end|>",
193
+ "<|vision_pad|>",
194
+ "<|image_pad|>",
195
+ "<|video_pad|>"
196
+ ],
197
+ "bos_token": null,
198
+ "chat_template": "{%- if tools %}\n {{- '<|im_start|>system\\n' }}\n {%- if messages[0]['role'] == 'system' %}\n {{- messages[0]['content'] }}\n {%- else %}\n {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}\n {%- endif %}\n {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n {%- for tool in tools %}\n {{- \"\\n\" }}\n {{- tool | tojson }}\n {%- endfor %}\n {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n {%- if messages[0]['role'] == 'system' %}\n {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n {%- else %}\n {{- '<|im_start|>system\\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\\n' }}\n {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n {%- elif message.role == \"assistant\" %}\n {{- '<|im_start|>' + message.role }}\n {%- if message.content %}\n {{- '\\n' + message.content }}\n {%- endif %}\n {%- for tool_call in message.tool_calls %}\n {%- if tool_call.function is defined %}\n {%- set tool_call = tool_call.function %}\n {%- endif %}\n {{- '\\n<tool_call>\\n{\"name\": \"' }}\n {{- tool_call.name }}\n {{- '\", \"arguments\": ' }}\n {{- tool_call.arguments | tojson }}\n {{- '}\\n</tool_call>' }}\n {%- endfor %}\n {{- '<|im_end|>\\n' }}\n {%- elif message.role == \"tool\" %}\n {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n {{- '<|im_start|>user' }}\n {%- endif %}\n {{- '\\n<tool_response>\\n' }}\n {{- message.content }}\n {{- '\\n</tool_response>' }}\n {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n {{- '<|im_end|>\\n' }}\n {%- endif %}\n {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
199
+ "clean_up_tokenization_spaces": false,
200
+ "eos_token": "<|im_end|>",
201
+ "errors": "replace",
202
+ "model_max_length": 131072,
203
+ "pad_token": "<|endoftext|>",
204
+ "split_special_tokens": false,
205
+ "tokenizer_class": "Qwen2Tokenizer",
206
+ "unk_token": null
207
+ }
vocab.json ADDED
The diff for this file is too large to render. See raw diff