Hugging Face's logo

Sebesky
/
MiniMax-M3-EAGLE3-RTN-INT4

Safetensors
llama
eagle3
speculative-decoding
compressed-tensors
int4
Model card Files
xet
 Community
Sebesky commited on
Commit
a727847
·
verified ·
1 Parent(s): b9201e5

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. config.json +80 -0
  2. model.safetensors +3 -0
  3. model.safetensors.index.json +44 -0
config.json ADDED
@@ -0,0 +1,80 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "LlamaForCausalLMEagle3"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "fc_norm": true,
8
+ "norm_output": true,
9
+ "head_dim": 128,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 6144,
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 18432,
14
+ "max_position_embeddings": 1048576,
15
+ "model_type": "llama",
16
+ "num_attention_heads": 64,
17
+ "num_hidden_layers": 1,
18
+ "num_key_value_heads": 64,
19
+ "rms_norm_eps": 1e-06,
20
+ "rope_scaling": null,
21
+ "rope_theta": 5000000,
22
+ "sliding_window": null,
23
+ "tie_word_embeddings": false,
24
+ "torch_dtype": "bfloat16",
25
+ "transformers_version": "4.51.0",
26
+ "use_cache": true,
27
+ "use_sliding_window": false,
28
+ "vocab_size": 200064,
29
+ "draft_vocab_size": 200064,
30
+ "_torchspec_version": "0.1.0",
31
+ "quantization_config": {
32
+ "quant_method": "compressed-tensors",
33
+ "format": "pack-quantized",
34
+ "quantization_status": "compressed",
35
+ "config_groups": {
36
+ "group_0": {
37
+ "targets": [
38
+ "Linear"
39
+ ],
40
+ "weights": {
41
+ "num_bits": 4,
42
+ "type": "int",
43
+ "symmetric": true,
44
+ "strategy": "group",
45
+ "group_size": 128,
46
+ "dynamic": false
47
+ }
48
+ },
49
+ "group_embed": {
50
+ "targets": [
51
+ "Embedding"
52
+ ],
53
+ "weights": {
54
+ "num_bits": 4,
55
+ "type": "int",
56
+ "symmetric": true,
57
+ "strategy": "group",
58
+ "group_size": 64,
59
+ "dynamic": false
60
+ }
61
+ },
62
+ "group_lmhead": {
63
+ "targets": [
64
+ "re:.*lm_head.*"
65
+ ],
66
+ "weights": {
67
+ "num_bits": 4,
68
+ "type": "int",
69
+ "symmetric": true,
70
+ "strategy": "group",
71
+ "group_size": 128,
72
+ "dynamic": false
73
+ }
74
+ }
75
+ },
76
+ "ignore": [
77
+ "re:.*norm.*"
78
+ ]
79
+ }
80
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a707a4302ce0becee147d2817d3bd3d9d8230711fde8ccd5fbd140cf19b56bb5
3
+ size 1702137680
model.safetensors.index.json ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 1702137680
4
+ },
5
+ "weight_map": {
6
+ "embed_tokens.weight_packed": "model.safetensors",
7
+ "embed_tokens.weight_scale": "model.safetensors",
8
+ "embed_tokens.weight_shape": "model.safetensors",
9
+ "fc.weight_packed": "model.safetensors",
10
+ "fc.weight_scale": "model.safetensors",
11
+ "fc.weight_shape": "model.safetensors",
12
+ "fc_norm.0.weight": "model.safetensors",
13
+ "fc_norm.1.weight": "model.safetensors",
14
+ "fc_norm.2.weight": "model.safetensors",
15
+ "layers.0.hidden_norm.weight": "model.safetensors",
16
+ "layers.0.input_layernorm.weight": "model.safetensors",
17
+ "layers.0.mlp.down_proj.weight_packed": "model.safetensors",
18
+ "layers.0.mlp.down_proj.weight_scale": "model.safetensors",
19
+ "layers.0.mlp.down_proj.weight_shape": "model.safetensors",
20
+ "layers.0.mlp.gate_proj.weight_packed": "model.safetensors",
21
+ "layers.0.mlp.gate_proj.weight_scale": "model.safetensors",
22
+ "layers.0.mlp.gate_proj.weight_shape": "model.safetensors",
23
+ "layers.0.mlp.up_proj.weight_packed": "model.safetensors",
24
+ "layers.0.mlp.up_proj.weight_scale": "model.safetensors",
25
+ "layers.0.mlp.up_proj.weight_shape": "model.safetensors",
26
+ "layers.0.post_attention_layernorm.weight": "model.safetensors",
27
+ "layers.0.self_attn.k_proj.weight_packed": "model.safetensors",
28
+ "layers.0.self_attn.k_proj.weight_scale": "model.safetensors",
29
+ "layers.0.self_attn.k_proj.weight_shape": "model.safetensors",
30
+ "layers.0.self_attn.o_proj.weight_packed": "model.safetensors",
31
+ "layers.0.self_attn.o_proj.weight_scale": "model.safetensors",
32
+ "layers.0.self_attn.o_proj.weight_shape": "model.safetensors",
33
+ "layers.0.self_attn.q_proj.weight_packed": "model.safetensors",
34
+ "layers.0.self_attn.q_proj.weight_scale": "model.safetensors",
35
+ "layers.0.self_attn.q_proj.weight_shape": "model.safetensors",
36
+ "layers.0.self_attn.v_proj.weight_packed": "model.safetensors",
37
+ "layers.0.self_attn.v_proj.weight_scale": "model.safetensors",
38
+ "layers.0.self_attn.v_proj.weight_shape": "model.safetensors",
39
+ "lm_head.weight_packed": "model.safetensors",
40
+ "lm_head.weight_scale": "model.safetensors",
41
+ "lm_head.weight_shape": "model.safetensors",
42
+ "norm.weight": "model.safetensors"
43
+ }
44
+ }