Hugging Face's logo

rungalileo
/
llama-3.2-3B-instruct-trtllm-ckpt-wq_int4_awq-kv_int8

Text Generation
llama
tensorrt-llm
int4
awq
kv-cache-quantization
Model card Files
xet
 Community
rungalileo commited on
Commit
5fe512a
·
verified ·
1 Parent(s): 3de4399

Upload folder using huggingface_hub

Browse files
Files changed (2) hide show
  1. config.json +147 -0
  2. rank0.safetensors +3 -0
config.json ADDED
@@ -0,0 +1,147 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "producer": {
3
+ "name": "modelopt",
4
+ "version": "0.25.0"
5
+ },
6
+ "architecture": "LlamaForCausalLM",
7
+ "dtype": "float16",
8
+ "logits_dtype": "float16",
9
+ "num_hidden_layers": 28,
10
+ "num_attention_heads": 24,
11
+ "num_key_value_heads": 8,
12
+ "hidden_size": 3072,
13
+ "norm_epsilon": 1e-05,
14
+ "vocab_size": 128256,
15
+ "max_position_embeddings": 131072,
16
+ "hidden_act": "silu",
17
+ "use_parallel_embedding": true,
18
+ "embedding_sharding_dim": 0,
19
+ "head_size": 128,
20
+ "intermediate_size": 8192,
21
+ "position_embedding_type": "rope_gpt_neox",
22
+ "share_embedding_table": false,
23
+ "residual_mlp": false,
24
+ "bias": false,
25
+ "rotary_pct": 1.0,
26
+ "rank": 0,
27
+ "decoder": "llama",
28
+ "rmsnorm": true,
29
+ "lm_head_bias": false,
30
+ "mlp_bias": false,
31
+ "attn_bias": false,
32
+ "rotary_base": 500000.0,
33
+ "rotary_scaling": {
34
+ "factor": 32.0,
35
+ "high_freq_factor": 4.0,
36
+ "low_freq_factor": 1.0,
37
+ "original_max_position_embeddings": 8192,
38
+ "rope_type": "llama3"
39
+ },
40
+ "disable_weight_only_quant_plugin": false,
41
+ "moe": {
42
+ "num_experts": 0,
43
+ "shared_expert_intermediate_size": 0,
44
+ "top_k": 0,
45
+ "normalization_mode": null,
46
+ "sparse_mixer_epsilon": 0.01,
47
+ "tp_mode": 0,
48
+ "device_limited_n_group": 0,
49
+ "device_limited_topk_group": 0,
50
+ "device_limited_routed_scaling_factor": 1.0
51
+ },
52
+ "remove_duplicated_kv_heads": false,
53
+ "fc_after_embed": false,
54
+ "use_input_layernorm_in_first_layer": true,
55
+ "use_last_layernorm": true,
56
+ "layer_idx_offset": 0,
57
+ "embedding_multiplier": 1.0,
58
+ "attention_multiplier": 1.0,
59
+ "residual_multiplier": 1.0,
60
+ "output_multiplier_scale": 1.0,
61
+ "runtime_defaults": null,
62
+ "mapping": {
63
+ "world_size": 1,
64
+ "gpus_per_node": 8,
65
+ "cp_size": 1,
66
+ "tp_size": 1,
67
+ "pp_size": 1,
68
+ "moe_tp_size": 1,
69
+ "moe_ep_size": 1,
70
+ "auto_parallel": false
71
+ },
72
+ "quantization": {
73
+ "quant_algo": "W4A16_AWQ",
74
+ "kv_cache_quant_algo": "INT8",
75
+ "group_size": 128,
76
+ "smoothquant_val": 0.5,
77
+ "clamp_val": null,
78
+ "use_meta_recipe": false,
79
+ "has_zero_point": false,
80
+ "pre_quant_scale": true,
81
+ "exclude_modules": [
82
+ "transformer.layers.17.input_layernorm",
83
+ "transformer.layers.1.input_layernorm",
84
+ "transformer.layers.24.input_layernorm",
85
+ "transformer.layers.1.post_layernorm",
86
+ "transformer.layers.6.input_layernorm",
87
+ "transformer.layers.14.post_layernorm",
88
+ "lm_head",
89
+ "transformer.layers.19.post_layernorm",
90
+ "transformer.layers.22.input_layernorm",
91
+ "transformer.layers.13.input_layernorm",
92
+ "transformer.layers.4.post_layernorm",
93
+ "transformer.layers.16.post_layernorm",
94
+ "transformer.layers.8.input_layernorm",
95
+ "transformer.layers.13.post_layernorm",
96
+ "transformer.layers.20.post_layernorm",
97
+ "transformer.layers.5.input_layernorm",
98
+ "transformer.layers.12.post_layernorm",
99
+ "transformer.layers.27.input_layernorm",
100
+ "transformer.layers.14.input_layernorm",
101
+ "transformer.layers.23.input_layernorm",
102
+ "transformer.layers.10.input_layernorm",
103
+ "transformer.layers.4.input_layernorm",
104
+ "transformer.layers.7.post_layernorm",
105
+ "transformer.layers.18.post_layernorm",
106
+ "transformer.layers.15.post_layernorm",
107
+ "transformer.layers.25.input_layernorm",
108
+ "transformer.layers.21.input_layernorm",
109
+ "transformer.layers.9.post_layernorm",
110
+ "transformer.layers.11.post_layernorm",
111
+ "transformer.layers.24.post_layernorm",
112
+ "transformer.layers.25.post_layernorm",
113
+ "transformer.layers.27.post_layernorm",
114
+ "transformer.layers.26.post_layernorm",
115
+ "transformer.layers.0.post_layernorm",
116
+ "transformer.layers.9.input_layernorm",
117
+ "transformer.layers.10.post_layernorm",
118
+ "transformer.layers.20.input_layernorm",
119
+ "transformer.layers.15.input_layernorm",
120
+ "transformer.layers.7.input_layernorm",
121
+ "transformer.layers.2.input_layernorm",
122
+ "transformer.layers.12.input_layernorm",
123
+ "transformer.layers.5.post_layernorm",
124
+ "transformer.layers.3.post_layernorm",
125
+ "transformer.layers.6.post_layernorm",
126
+ "transformer.layers.17.post_layernorm",
127
+ "transformer.layers.8.post_layernorm",
128
+ "transformer.layers.18.input_layernorm",
129
+ "transformer.layers.2.post_layernorm",
130
+ "transformer.layers.11.input_layernorm",
131
+ "transformer.layers.22.post_layernorm",
132
+ "transformer.ln_f",
133
+ "transformer.layers.21.post_layernorm",
134
+ "transformer.layers.26.input_layernorm",
135
+ "transformer.layers.19.input_layernorm",
136
+ "transformer.layers.23.post_layernorm",
137
+ "transformer.layers.3.input_layernorm",
138
+ "transformer.layers.16.input_layernorm",
139
+ "transformer.layers.0.input_layernorm",
140
+ "transformer.vocab_embedding"
141
+ ]
142
+ },
143
+ "qk_layernorm": false,
144
+ "rotary_embedding_dim": 128,
145
+ "tie_word_embeddings": true,
146
+ "model_type": "llama"
147
+ }
rank0.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2135e23a4d5aa5a0cfe42b58c7bf7b76f83fa566bce3f17240f556a0f4e70d24
3
+ size 3074935136