File size: 67,866 Bytes
494f611 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 |
/home/ruihang/Workspace/miniconda3/envs/python311/bin/python -m mlc_llm gen_config /models/Llama-2-7b-chat-hf --quantization q0f16 --conv-template llama-2 --output /tmp/tmp8ew5r2yr
[2024-05-22 00:30:05] INFO auto_config.py:115: [92mFound[0m model configuration: /models/Llama-2-7b-chat-hf/config.json
[2024-05-22 00:30:05] INFO auto_config.py:153: [92mFound[0m model type: [1mllama[0m. Use `--model-type` to override.
[2024-05-22 00:30:05] INFO llama_model.py:52: [1mcontext_window_size[0m not found in config.json. Falling back to [1mmax_position_embeddings[0m (4096)
[2024-05-22 00:30:05] INFO llama_model.py:72: [1mprefill_chunk_size[0m defaults to 2048
[2024-05-22 00:30:05] INFO config.py:106: Overriding [1mmax_batch_size[0m from 1 to 80
[2024-05-22 00:30:05] INFO gen_config.py:255: [generation_config.json] Setting [1mbos_token_id[0m: 1
[2024-05-22 00:30:05] INFO gen_config.py:255: [generation_config.json] Setting [1meos_token_id[0m: 2
[2024-05-22 00:30:05] INFO gen_config.py:255: [generation_config.json] Setting [1mpad_token_id[0m: 0
[2024-05-22 00:30:05] INFO gen_config.py:255: [generation_config.json] Setting [1mtemperature[0m: 0.6
[2024-05-22 00:30:05] INFO gen_config.py:255: [generation_config.json] Setting [1mtop_p[0m: 0.9
[2024-05-22 00:30:05] INFO gen_config.py:267: [92mFound[0m tokenizer config: /models/Llama-2-7b-chat-hf/tokenizer.model. Copying to [1m/tmp/tmp8ew5r2yr/tokenizer.model[0m
[2024-05-22 00:30:05] INFO gen_config.py:267: [92mFound[0m tokenizer config: /models/Llama-2-7b-chat-hf/tokenizer.json. Copying to [1m/tmp/tmp8ew5r2yr/tokenizer.json[0m
[2024-05-22 00:30:05] INFO gen_config.py:269: [91mNot found[0m tokenizer config: /models/Llama-2-7b-chat-hf/vocab.json
[2024-05-22 00:30:05] INFO gen_config.py:269: [91mNot found[0m tokenizer config: /models/Llama-2-7b-chat-hf/merges.txt
[2024-05-22 00:30:05] INFO gen_config.py:269: [91mNot found[0m tokenizer config: /models/Llama-2-7b-chat-hf/added_tokens.json
[2024-05-22 00:30:05] INFO gen_config.py:267: [92mFound[0m tokenizer config: /models/Llama-2-7b-chat-hf/tokenizer_config.json. Copying to [1m/tmp/tmp8ew5r2yr/tokenizer_config.json[0m
[2024-05-22 00:30:05] INFO gen_config.py:80: [System default] Setting [1mpresence_penalty[0m: 0.0
[2024-05-22 00:30:05] INFO gen_config.py:80: [System default] Setting [1mfrequency_penalty[0m: 0.0
[2024-05-22 00:30:05] INFO gen_config.py:80: [System default] Setting [1mrepetition_penalty[0m: 1.0
[2024-05-22 00:30:05] INFO gen_config.py:80: [System default] Setting [1mmean_gen_len[0m: 128
[2024-05-22 00:30:05] INFO gen_config.py:80: [System default] Setting [1mmax_gen_len[0m: 512
[2024-05-22 00:30:05] INFO gen_config.py:80: [System default] Setting [1mshift_fill_factor[0m: 0.3
[2024-05-22 00:30:05] INFO gen_config.py:335: Dumping configuration file to: [1m/tmp/tmp8ew5r2yr/mlc-chat-config.json[0m
/home/ruihang/Workspace/miniconda3/envs/python311/bin/python -m mlc_llm convert_weight /models/Llama-2-7b-chat-hf --quantization q0f16 --source-format auto --output /tmp/tmp8ew5r2yr
[2024-05-22 00:30:06] INFO auto_config.py:115: [92mFound[0m model configuration: /models/Llama-2-7b-chat-hf/config.json
[2024-05-22 00:30:06] INFO auto_device.py:79: [92mFound[0m device: cuda:0
[2024-05-22 00:30:06] INFO auto_device.py:79: [92mFound[0m device: cuda:1
[2024-05-22 00:30:07] INFO auto_device.py:88: [91mNot found[0m device: rocm:0
[2024-05-22 00:30:08] INFO auto_device.py:88: [91mNot found[0m device: metal:0
[2024-05-22 00:30:09] INFO auto_device.py:79: [92mFound[0m device: vulkan:0
[2024-05-22 00:30:09] INFO auto_device.py:79: [92mFound[0m device: vulkan:1
[2024-05-22 00:30:09] INFO auto_device.py:79: [92mFound[0m device: vulkan:2
[2024-05-22 00:30:10] INFO auto_device.py:79: [92mFound[0m device: opencl:0
[2024-05-22 00:30:10] INFO auto_device.py:79: [92mFound[0m device: opencl:1
[2024-05-22 00:30:10] INFO auto_device.py:35: Using device: [1mcuda:0[0m
[2024-05-22 00:30:10] INFO auto_weight.py:70: Finding weights in: /models/Llama-2-7b-chat-hf
[2024-05-22 00:30:10] INFO auto_weight.py:120: [92mFound[0m source weight format: huggingface-torch. Source configuration: /models/Llama-2-7b-chat-hf/pytorch_model.bin.index.json
[2024-05-22 00:30:10] INFO auto_weight.py:143: [92mFound[0m source weight format: huggingface-safetensor. Source configuration: /models/Llama-2-7b-chat-hf/model.safetensors.index.json
[2024-05-22 00:30:10] INFO auto_weight.py:106: Using source weight configuration: [1m/models/Llama-2-7b-chat-hf/pytorch_model.bin.index.json[0m. Use `--source` to override.
[2024-05-22 00:30:10] INFO auto_weight.py:110: Using source weight format: [1mhuggingface-torch[0m. Use `--source-format` to override.
[2024-05-22 00:30:10] INFO auto_config.py:153: [92mFound[0m model type: [1mllama[0m. Use `--model-type` to override.
[2024-05-22 00:30:10] INFO llama_model.py:52: [1mcontext_window_size[0m not found in config.json. Falling back to [1mmax_position_embeddings[0m (4096)
[2024-05-22 00:30:10] INFO llama_model.py:72: [1mprefill_chunk_size[0m defaults to 2048
[1mWeight conversion with arguments:[0m
[1m--config[0m /models/Llama-2-7b-chat-hf/config.json
[1m--quantization[0m NoQuantize(name='q0f16', kind='no-quant', model_dtype='float16')
[1m--model-type[0m llama
[1m--device[0m cuda:0
[1m--source[0m /models/Llama-2-7b-chat-hf/pytorch_model.bin.index.json
[1m--source-format[0m huggingface-torch
[1m--output[0m /tmp/tmp8ew5r2yr
Start storing to cache /tmp/tmp8ew5r2yr
0%| | 0/195 [00:00<?, ?it/s]
[2024-05-22 00:30:11] INFO huggingface_loader.py:184: Loading HF parameters from: /models/Llama-2-7b-chat-hf/pytorch_model-00002-of-00002.bin
0%| | 0/195 [00:00<?, ?it/s]
[2024-05-22 00:30:19] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mlm_head.weight[0m", shape: (32000, 4096), dtype: float16
0%| | 0/195 [00:08<?, ?it/s]
1%| | 1/195 [00:09<29:48, 9.22s/it]
[2024-05-22 00:30:20] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.24.input_layernorm.weight[0m", shape: (4096,), dtype: float16
1%| | 1/195 [00:09<29:48, 9.22s/it]
[2024-05-22 00:30:20] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.24.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
1%| | 1/195 [00:09<29:48, 9.22s/it]
2%|▏ | 3/195 [00:09<07:58, 2.49s/it]
[2024-05-22 00:30:21] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.24.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
2%|▏ | 3/195 [00:09<07:58, 2.49s/it]
2%|▏ | 4/195 [00:10<06:17, 1.98s/it]
[2024-05-22 00:30:21] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.24.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
2%|▏ | 4/195 [00:10<06:17, 1.98s/it]
[2024-05-22 00:30:22] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.24.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
2%|▏ | 4/195 [00:10<06:17, 1.98s/it]
3%|▎ | 6/195 [00:10<03:28, 1.10s/it]
[2024-05-22 00:30:22] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.24.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
3%|▎ | 6/195 [00:10<03:28, 1.10s/it]
4%|▎ | 7/195 [00:11<02:38, 1.19it/s]
[2024-05-22 00:30:22] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.25.input_layernorm.weight[0m", shape: (4096,), dtype: float16
4%|▎ | 7/195 [00:11<02:38, 1.19it/s]
[2024-05-22 00:30:22] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.25.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
4%|▎ | 7/195 [00:11<02:38, 1.19it/s]
5%|▍ | 9/195 [00:11<01:39, 1.87it/s]
[2024-05-22 00:30:23] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.25.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
5%|▍ | 9/195 [00:11<01:39, 1.87it/s]
5%|▌ | 10/195 [00:12<02:00, 1.54it/s]
[2024-05-22 00:30:23] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.25.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
5%|▌ | 10/195 [00:12<02:00, 1.54it/s]
[2024-05-22 00:30:23] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.25.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
5%|▌ | 10/195 [00:12<02:00, 1.54it/s]
6%|▌ | 12/195 [00:12<01:24, 2.15it/s]
[2024-05-22 00:30:24] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.25.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
6%|▌ | 12/195 [00:12<01:24, 2.15it/s]
7%|▋ | 13/195 [00:12<01:09, 2.61it/s]
[2024-05-22 00:30:24] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.26.input_layernorm.weight[0m", shape: (4096,), dtype: float16
7%|▋ | 13/195 [00:12<01:09, 2.61it/s]
[2024-05-22 00:30:24] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.26.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
7%|▋ | 13/195 [00:12<01:09, 2.61it/s]
8%|▊ | 15/195 [00:13<00:50, 3.53it/s]
[2024-05-22 00:30:24] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.26.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
8%|▊ | 15/195 [00:13<00:50, 3.53it/s]
8%|▊ | 16/195 [00:13<01:15, 2.36it/s]
[2024-05-22 00:30:25] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.26.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
8%|▊ | 16/195 [00:13<01:15, 2.36it/s]
[2024-05-22 00:30:25] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.26.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
8%|▊ | 16/195 [00:14<01:15, 2.36it/s]
9%|▉ | 18/195 [00:14<00:58, 3.04it/s]
[2024-05-22 00:30:25] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.26.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
9%|▉ | 18/195 [00:14<00:58, 3.04it/s]
10%|▉ | 19/195 [00:14<00:49, 3.57it/s]
[2024-05-22 00:30:25] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.27.input_layernorm.weight[0m", shape: (4096,), dtype: float16
10%|▉ | 19/195 [00:14<00:49, 3.57it/s]
[2024-05-22 00:30:25] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.27.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
10%|▉ | 19/195 [00:14<00:49, 3.57it/s]
11%|█ | 21/195 [00:14<00:38, 4.53it/s]
[2024-05-22 00:30:26] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.27.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
11%|█ | 21/195 [00:14<00:38, 4.53it/s]
11%|█▏ | 22/195 [00:15<00:55, 3.11it/s]
[2024-05-22 00:30:26] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.27.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
11%|█▏ | 22/195 [00:15<00:55, 3.11it/s]
[2024-05-22 00:30:26] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.27.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
11%|█▏ | 22/195 [00:15<00:55, 3.11it/s]
12%|█▏ | 24/195 [00:15<00:44, 3.87it/s]
[2024-05-22 00:30:27] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.27.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
12%|█▏ | 24/195 [00:15<00:44, 3.87it/s]
13%|█▎ | 25/195 [00:15<00:38, 4.45it/s]
[2024-05-22 00:30:27] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.28.input_layernorm.weight[0m", shape: (4096,), dtype: float16
13%|█▎ | 25/195 [00:15<00:38, 4.45it/s]
[2024-05-22 00:30:27] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.28.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
13%|█▎ | 25/195 [00:15<00:38, 4.45it/s]
14%|█▍ | 27/195 [00:16<00:31, 5.38it/s]
[2024-05-22 00:30:27] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.28.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
14%|█▍ | 27/195 [00:16<00:31, 5.38it/s]
14%|█▍ | 28/195 [00:16<00:53, 3.14it/s]
[2024-05-22 00:30:28] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.28.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
14%|█▍ | 28/195 [00:16<00:53, 3.14it/s]
[2024-05-22 00:30:28] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.28.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
14%|█▍ | 28/195 [00:16<00:53, 3.14it/s]
15%|█▌ | 30/195 [00:17<00:42, 3.89it/s]
[2024-05-22 00:30:28] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.28.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
15%|█▌ | 30/195 [00:17<00:42, 3.89it/s]
16%|█▌ | 31/195 [00:17<00:36, 4.47it/s]
[2024-05-22 00:30:28] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.29.input_layernorm.weight[0m", shape: (4096,), dtype: float16
16%|█▌ | 31/195 [00:17<00:36, 4.47it/s]
[2024-05-22 00:30:28] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.29.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
16%|█▌ | 31/195 [00:17<00:36, 4.47it/s]
17%|█▋ | 33/195 [00:17<00:32, 4.97it/s]
[2024-05-22 00:30:29] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.29.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
17%|█▋ | 33/195 [00:17<00:32, 4.97it/s]
17%|█▋ | 34/195 [00:18<00:51, 3.12it/s]
[2024-05-22 00:30:29] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.29.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
17%|█▋ | 34/195 [00:18<00:51, 3.12it/s]
[2024-05-22 00:30:29] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.29.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
17%|█▋ | 34/195 [00:18<00:51, 3.12it/s]
18%|█▊ | 36/195 [00:18<00:42, 3.77it/s]
[2024-05-22 00:30:30] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.29.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
18%|█▊ | 36/195 [00:18<00:42, 3.77it/s]
[2024-05-22 00:30:30] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.30.input_layernorm.weight[0m", shape: (4096,), dtype: float16
18%|█▊ | 36/195 [00:18<00:42, 3.77it/s]
19%|█▉ | 38/195 [00:18<00:29, 5.26it/s]
[2024-05-22 00:30:30] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.30.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
19%|█▉ | 38/195 [00:18<00:29, 5.26it/s]
20%|██ | 39/195 [00:19<00:31, 5.02it/s]
[2024-05-22 00:30:30] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.30.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
20%|██ | 39/195 [00:19<00:31, 5.02it/s]
21%|██ | 40/195 [00:19<00:46, 3.35it/s]
[2024-05-22 00:30:31] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.30.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
21%|██ | 40/195 [00:19<00:46, 3.35it/s]
[2024-05-22 00:30:31] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.30.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
21%|██ | 40/195 [00:19<00:46, 3.35it/s]
22%|██▏ | 42/195 [00:19<00:37, 4.12it/s]
[2024-05-22 00:30:31] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.30.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
22%|██▏ | 42/195 [00:19<00:37, 4.12it/s]
22%|██▏ | 43/195 [00:20<00:32, 4.72it/s]
[2024-05-22 00:30:31] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.31.input_layernorm.weight[0m", shape: (4096,), dtype: float16
22%|██▏ | 43/195 [00:20<00:32, 4.72it/s]
[2024-05-22 00:30:31] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.31.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
22%|██▏ | 43/195 [00:20<00:32, 4.72it/s]
23%|██▎ | 45/195 [00:20<00:26, 5.62it/s]
[2024-05-22 00:30:31] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.31.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
23%|██▎ | 45/195 [00:20<00:26, 5.62it/s]
24%|██▎ | 46/195 [00:21<00:44, 3.35it/s]
[2024-05-22 00:30:32] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.31.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
24%|██▎ | 46/195 [00:21<00:44, 3.35it/s]
[2024-05-22 00:30:32] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.31.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
24%|██▎ | 46/195 [00:21<00:44, 3.35it/s]
25%|██▍ | 48/195 [00:21<00:35, 4.11it/s]
[2024-05-22 00:30:32] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.31.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
25%|██▍ | 48/195 [00:21<00:35, 4.11it/s]
25%|██▌ | 49/195 [00:21<00:31, 4.70it/s]
[2024-05-22 00:30:32] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.norm.weight[0m", shape: (4096,), dtype: float16
25%|██▌ | 49/195 [00:21<00:31, 4.70it/s]
[2024-05-22 00:30:32] INFO huggingface_loader.py:196: Unloading HF weight file: /models/Llama-2-7b-chat-hf/pytorch_model-00002-of-00002.bin
25%|██▌ | 49/195 [00:21<00:31, 4.70it/s]
[2024-05-22 00:30:32] INFO huggingface_loader.py:184: Loading HF parameters from: /models/Llama-2-7b-chat-hf/pytorch_model-00001-of-00002.bin
25%|██▌ | 49/195 [00:21<00:31, 4.70it/s]
[2024-05-22 00:31:09] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.embed_tokens.weight[0m", shape: (32000, 4096), dtype: float16
25%|██▌ | 49/195 [00:58<00:31, 4.70it/s]
26%|██▌ | 51/195 [00:58<17:18, 7.21s/it]
[2024-05-22 00:31:10] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.0.input_layernorm.weight[0m", shape: (4096,), dtype: float16
26%|██▌ | 51/195 [00:58<17:18, 7.21s/it]
[2024-05-22 00:31:10] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.0.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
26%|██▌ | 51/195 [00:58<17:18, 7.21s/it]
27%|██▋ | 53/195 [00:59<11:13, 4.74s/it]
[2024-05-22 00:31:10] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.0.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
27%|██▋ | 53/195 [00:59<11:13, 4.74s/it]
28%|██▊ | 54/195 [01:00<09:23, 4.00s/it]
[2024-05-22 00:31:11] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.0.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
28%|██▊ | 54/195 [01:00<09:23, 4.00s/it]
[2024-05-22 00:31:11] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.0.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
28%|██▊ | 54/195 [01:00<09:23, 4.00s/it]
29%|██▊ | 56/195 [01:00<06:03, 2.61s/it]
[2024-05-22 00:31:11] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.0.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
29%|██▊ | 56/195 [01:00<06:03, 2.61s/it]
29%|██▉ | 57/195 [01:00<04:50, 2.10s/it]
[2024-05-22 00:31:12] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.1.input_layernorm.weight[0m", shape: (4096,), dtype: float16
29%|██▉ | 57/195 [01:00<04:50, 2.10s/it]
[2024-05-22 00:31:12] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.1.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
29%|██▉ | 57/195 [01:00<04:50, 2.10s/it]
30%|███ | 59/195 [01:01<03:07, 1.38s/it]
[2024-05-22 00:31:12] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.1.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
30%|███ | 59/195 [01:01<03:07, 1.38s/it]
31%|███ | 60/195 [01:01<02:47, 1.24s/it]
[2024-05-22 00:31:13] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.1.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
31%|███ | 60/195 [01:01<02:47, 1.24s/it]
[2024-05-22 00:31:13] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.1.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
31%|███ | 60/195 [01:01<02:47, 1.24s/it]
32%|███▏ | 62/195 [01:02<01:52, 1.18it/s]
[2024-05-22 00:31:13] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.1.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
32%|███▏ | 62/195 [01:02<01:52, 1.18it/s]
32%|███▏ | 63/195 [01:02<01:31, 1.44it/s]
[2024-05-22 00:31:13] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.10.input_layernorm.weight[0m", shape: (4096,), dtype: float16
32%|███▏ | 63/195 [01:02<01:31, 1.44it/s]
[2024-05-22 00:31:13] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.10.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
32%|███▏ | 63/195 [01:02<01:31, 1.44it/s]
33%|███▎ | 65/195 [01:02<01:03, 2.04it/s]
[2024-05-22 00:31:14] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.10.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
33%|███▎ | 65/195 [01:02<01:03, 2.04it/s]
34%|███▍ | 66/195 [01:03<01:08, 1.90it/s]
[2024-05-22 00:31:14] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.10.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
34%|███▍ | 66/195 [01:03<01:08, 1.90it/s]
[2024-05-22 00:31:14] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.10.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
34%|███▍ | 66/195 [01:03<01:08, 1.90it/s]
35%|███▍ | 68/195 [01:03<00:48, 2.61it/s]
[2024-05-22 00:31:14] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.10.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
35%|███▍ | 68/195 [01:03<00:48, 2.61it/s]
[2024-05-22 00:31:14] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.11.input_layernorm.weight[0m", shape: (4096,), dtype: float16
35%|███▍ | 68/195 [01:03<00:48, 2.61it/s]
[2024-05-22 00:31:15] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.11.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
35%|███▍ | 68/195 [01:03<00:48, 2.61it/s]
36%|███▋ | 71/195 [01:03<00:32, 3.82it/s]
[2024-05-22 00:31:15] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.11.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
36%|███▋ | 71/195 [01:03<00:32, 3.82it/s]
37%|███▋ | 72/195 [01:04<00:37, 3.31it/s]
[2024-05-22 00:31:15] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.11.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
37%|███▋ | 72/195 [01:04<00:37, 3.31it/s]
[2024-05-22 00:31:15] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.11.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
37%|███▋ | 72/195 [01:04<00:37, 3.31it/s]
38%|███▊ | 74/195 [01:04<00:30, 4.02it/s]
[2024-05-22 00:31:16] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.11.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
38%|███▊ | 74/195 [01:04<00:30, 4.02it/s]
[2024-05-22 00:31:16] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.12.input_layernorm.weight[0m", shape: (4096,), dtype: float16
38%|███▊ | 74/195 [01:04<00:30, 4.02it/s]
[2024-05-22 00:31:16] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.12.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
38%|███▊ | 74/195 [01:04<00:30, 4.02it/s]
39%|███▉ | 77/195 [01:04<00:22, 5.26it/s]
[2024-05-22 00:31:16] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.12.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
39%|███▉ | 77/195 [01:05<00:22, 5.26it/s]
40%|████ | 78/195 [01:05<00:27, 4.18it/s]
[2024-05-22 00:31:16] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.12.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
40%|████ | 78/195 [01:05<00:27, 4.18it/s]
[2024-05-22 00:31:16] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.12.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
40%|████ | 78/195 [01:05<00:27, 4.18it/s]
41%|████ | 80/195 [01:05<00:23, 4.83it/s]
[2024-05-22 00:31:17] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.12.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
41%|████ | 80/195 [01:05<00:23, 4.83it/s]
[2024-05-22 00:31:17] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.13.input_layernorm.weight[0m", shape: (4096,), dtype: float16
41%|████ | 80/195 [01:05<00:23, 4.83it/s]
[2024-05-22 00:31:17] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.13.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
41%|████ | 80/195 [01:05<00:23, 4.83it/s]
43%|████▎ | 83/195 [01:06<00:18, 6.01it/s]
[2024-05-22 00:31:17] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.13.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
43%|████▎ | 83/195 [01:06<00:18, 6.01it/s]
43%|████▎ | 84/195 [01:06<00:24, 4.57it/s]
[2024-05-22 00:31:17] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.13.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
43%|████▎ | 84/195 [01:06<00:24, 4.57it/s]
[2024-05-22 00:31:17] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.13.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
43%|████▎ | 84/195 [01:06<00:24, 4.57it/s]
44%|████▍ | 86/195 [01:06<00:20, 5.19it/s]
[2024-05-22 00:31:18] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.13.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
44%|████▍ | 86/195 [01:06<00:20, 5.19it/s]
[2024-05-22 00:31:18] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.14.input_layernorm.weight[0m", shape: (4096,), dtype: float16
44%|████▍ | 86/195 [01:06<00:20, 5.19it/s]
[2024-05-22 00:31:18] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.14.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
44%|████▍ | 86/195 [01:06<00:20, 5.19it/s]
46%|████▌ | 89/195 [01:07<00:16, 6.37it/s]
[2024-05-22 00:31:18] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.14.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
46%|████▌ | 89/195 [01:07<00:16, 6.37it/s]
46%|████▌ | 90/195 [01:07<00:21, 4.80it/s]
[2024-05-22 00:31:18] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.14.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
46%|████▌ | 90/195 [01:07<00:21, 4.80it/s]
[2024-05-22 00:31:19] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.14.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
46%|████▌ | 90/195 [01:07<00:21, 4.80it/s]
47%|████▋ | 92/195 [01:07<00:19, 5.41it/s]
[2024-05-22 00:31:19] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.14.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
47%|████▋ | 92/195 [01:07<00:19, 5.41it/s]
[2024-05-22 00:31:19] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.15.input_layernorm.weight[0m", shape: (4096,), dtype: float16
47%|████▋ | 92/195 [01:07<00:19, 5.41it/s]
[2024-05-22 00:31:19] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.15.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
47%|████▋ | 92/195 [01:07<00:19, 5.41it/s]
49%|████▊ | 95/195 [01:08<00:15, 6.51it/s]
[2024-05-22 00:31:19] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.15.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
49%|████▊ | 95/195 [01:08<00:15, 6.51it/s]
49%|████▉ | 96/195 [01:08<00:20, 4.81it/s]
[2024-05-22 00:31:20] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.15.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
49%|████▉ | 96/195 [01:08<00:20, 4.81it/s]
[2024-05-22 00:31:20] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.15.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
49%|████▉ | 96/195 [01:08<00:20, 4.81it/s]
50%|█████ | 98/195 [01:08<00:17, 5.41it/s]
[2024-05-22 00:31:20] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.15.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
50%|█████ | 98/195 [01:08<00:17, 5.41it/s]
[2024-05-22 00:31:20] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.16.input_layernorm.weight[0m", shape: (4096,), dtype: float16
50%|█████ | 98/195 [01:09<00:17, 5.41it/s]
[2024-05-22 00:31:20] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.16.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
50%|█████ | 98/195 [01:09<00:17, 5.41it/s]
52%|█████▏ | 101/195 [01:09<00:14, 6.58it/s]
[2024-05-22 00:31:20] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.16.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
52%|█████▏ | 101/195 [01:09<00:14, 6.58it/s]
52%|█████▏ | 102/195 [01:09<00:18, 4.92it/s]
[2024-05-22 00:31:21] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.16.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
52%|█████▏ | 102/195 [01:09<00:18, 4.92it/s]
[2024-05-22 00:31:21] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.16.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
52%|█████▏ | 102/195 [01:09<00:18, 4.92it/s]
53%|█████▎ | 104/195 [01:10<00:16, 5.44it/s]
[2024-05-22 00:31:21] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.16.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
53%|█████▎ | 104/195 [01:10<00:16, 5.44it/s]
[2024-05-22 00:31:21] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.17.input_layernorm.weight[0m", shape: (4096,), dtype: float16
53%|█████▎ | 104/195 [01:10<00:16, 5.44it/s]
[2024-05-22 00:31:21] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.17.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
53%|█████▎ | 104/195 [01:10<00:16, 5.44it/s]
55%|█████▍ | 107/195 [01:10<00:13, 6.60it/s]
[2024-05-22 00:31:21] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.17.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
55%|█████▍ | 107/195 [01:10<00:13, 6.60it/s]
55%|█████▌ | 108/195 [01:10<00:17, 4.94it/s]
[2024-05-22 00:31:22] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.17.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
55%|█████▌ | 108/195 [01:10<00:17, 4.94it/s]
[2024-05-22 00:31:22] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.17.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
55%|█████▌ | 108/195 [01:10<00:17, 4.94it/s]
56%|█████▋ | 110/195 [01:11<00:15, 5.47it/s]
[2024-05-22 00:31:22] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.17.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
56%|█████▋ | 110/195 [01:11<00:15, 5.47it/s]
[2024-05-22 00:31:22] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.18.input_layernorm.weight[0m", shape: (4096,), dtype: float16
56%|█████▋ | 110/195 [01:11<00:15, 5.47it/s]
[2024-05-22 00:31:22] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.18.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
56%|█████▋ | 110/195 [01:11<00:15, 5.47it/s]
58%|█████▊ | 113/195 [01:11<00:12, 6.62it/s]
[2024-05-22 00:31:22] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.18.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
58%|█████▊ | 113/195 [01:11<00:12, 6.62it/s]
58%|█████▊ | 114/195 [01:11<00:16, 4.95it/s]
[2024-05-22 00:31:23] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.18.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
58%|█████▊ | 114/195 [01:11<00:16, 4.95it/s]
[2024-05-22 00:31:23] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.18.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
58%|█████▊ | 114/195 [01:11<00:16, 4.95it/s]
59%|█████▉ | 116/195 [01:12<00:14, 5.55it/s]
[2024-05-22 00:31:23] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.18.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
59%|█████▉ | 116/195 [01:12<00:14, 5.55it/s]
[2024-05-22 00:31:23] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.19.input_layernorm.weight[0m", shape: (4096,), dtype: float16
59%|█████▉ | 116/195 [01:12<00:14, 5.55it/s]
[2024-05-22 00:31:23] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.19.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
59%|█████▉ | 116/195 [01:12<00:14, 5.55it/s]
61%|██████ | 119/195 [01:12<00:11, 6.63it/s]
[2024-05-22 00:31:23] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.19.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
61%|██████ | 119/195 [01:12<00:11, 6.63it/s]
62%|██████▏ | 120/195 [01:12<00:15, 4.78it/s]
[2024-05-22 00:31:24] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.19.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
62%|██████▏ | 120/195 [01:12<00:15, 4.78it/s]
[2024-05-22 00:31:24] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.19.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
62%|██████▏ | 120/195 [01:13<00:15, 4.78it/s]
63%|██████▎ | 122/195 [01:13<00:13, 5.33it/s]
[2024-05-22 00:31:24] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.19.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
63%|██████▎ | 122/195 [01:13<00:13, 5.33it/s]
[2024-05-22 00:31:24] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.2.input_layernorm.weight[0m", shape: (4096,), dtype: float16
63%|██████▎ | 122/195 [01:13<00:13, 5.33it/s]
[2024-05-22 00:31:24] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.2.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
63%|██████▎ | 122/195 [01:13<00:13, 5.33it/s]
64%|██████▍ | 125/195 [01:13<00:10, 6.42it/s]
[2024-05-22 00:31:24] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.2.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
64%|██████▍ | 125/195 [01:13<00:10, 6.42it/s]
65%|██████▍ | 126/195 [01:14<00:14, 4.77it/s]
[2024-05-22 00:31:25] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.2.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
65%|██████▍ | 126/195 [01:14<00:14, 4.77it/s]
[2024-05-22 00:31:25] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.2.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
65%|██████▍ | 126/195 [01:14<00:14, 4.77it/s]
66%|██████▌ | 128/195 [01:14<00:12, 5.32it/s]
[2024-05-22 00:31:25] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.2.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
66%|██████▌ | 128/195 [01:14<00:12, 5.32it/s]
[2024-05-22 00:31:25] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.20.input_layernorm.weight[0m", shape: (4096,), dtype: float16
66%|██████▌ | 128/195 [01:14<00:12, 5.32it/s]
[2024-05-22 00:31:25] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.20.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
66%|██████▌ | 128/195 [01:14<00:12, 5.32it/s]
67%|██████▋ | 131/195 [01:14<00:09, 6.50it/s]
[2024-05-22 00:31:26] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.20.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
67%|██████▋ | 131/195 [01:14<00:09, 6.50it/s]
68%|██████▊ | 132/195 [01:15<00:12, 4.89it/s]
[2024-05-22 00:31:26] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.20.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
68%|██████▊ | 132/195 [01:15<00:12, 4.89it/s]
[2024-05-22 00:31:26] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.20.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
68%|██████▊ | 132/195 [01:15<00:12, 4.89it/s]
69%|██████▊ | 134/195 [01:15<00:11, 5.50it/s]
[2024-05-22 00:31:26] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.20.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
69%|██████▊ | 134/195 [01:15<00:11, 5.50it/s]
[2024-05-22 00:31:26] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.21.input_layernorm.weight[0m", shape: (4096,), dtype: float16
69%|██████▊ | 134/195 [01:15<00:11, 5.50it/s]
[2024-05-22 00:31:26] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.21.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
69%|██████▊ | 134/195 [01:15<00:11, 5.50it/s]
70%|███████ | 137/195 [01:15<00:08, 6.58it/s]
[2024-05-22 00:31:27] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.21.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
70%|███████ | 137/195 [01:15<00:08, 6.58it/s]
71%|███████ | 138/195 [01:16<00:11, 4.80it/s]
[2024-05-22 00:31:27] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.21.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
71%|███████ | 138/195 [01:16<00:11, 4.80it/s]
[2024-05-22 00:31:27] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.21.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
71%|███████ | 138/195 [01:16<00:11, 4.80it/s]
72%|███████▏ | 140/195 [01:16<00:10, 5.42it/s]
[2024-05-22 00:31:27] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.21.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
72%|███████▏ | 140/195 [01:16<00:10, 5.42it/s]
[2024-05-22 00:31:27] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.22.input_layernorm.weight[0m", shape: (4096,), dtype: float16
72%|███████▏ | 140/195 [01:16<00:10, 5.42it/s]
[2024-05-22 00:31:27] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.22.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
72%|███████▏ | 140/195 [01:16<00:10, 5.42it/s]
73%|███████▎ | 143/195 [01:16<00:07, 6.51it/s]
[2024-05-22 00:31:28] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.22.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
73%|███████▎ | 143/195 [01:16<00:07, 6.51it/s]
74%|███████▍ | 144/195 [01:17<00:10, 4.87it/s]
[2024-05-22 00:31:28] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.22.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
74%|███████▍ | 144/195 [01:17<00:10, 4.87it/s]
[2024-05-22 00:31:28] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.22.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
74%|███████▍ | 144/195 [01:17<00:10, 4.87it/s]
75%|███████▍ | 146/195 [01:17<00:09, 5.41it/s]
[2024-05-22 00:31:28] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.22.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
75%|███████▍ | 146/195 [01:17<00:09, 5.41it/s]
[2024-05-22 00:31:28] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.23.input_layernorm.weight[0m", shape: (4096,), dtype: float16
75%|███████▍ | 146/195 [01:17<00:09, 5.41it/s]
[2024-05-22 00:31:29] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.23.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
75%|███████▍ | 146/195 [01:17<00:09, 5.41it/s]
76%|███████▋ | 149/195 [01:17<00:06, 6.58it/s]
[2024-05-22 00:31:29] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.23.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
76%|███████▋ | 149/195 [01:17<00:06, 6.58it/s]
77%|███████▋ | 150/195 [01:18<00:09, 4.82it/s]
[2024-05-22 00:31:29] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.23.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
77%|███████▋ | 150/195 [01:18<00:09, 4.82it/s]
[2024-05-22 00:31:29] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.23.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
77%|███████▋ | 150/195 [01:18<00:09, 4.82it/s]
78%|███████▊ | 152/195 [01:18<00:07, 5.44it/s]
[2024-05-22 00:31:29] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.23.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
78%|███████▊ | 152/195 [01:18<00:07, 5.44it/s]
[2024-05-22 00:31:30] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.3.input_layernorm.weight[0m", shape: (4096,), dtype: float16
78%|███████▊ | 152/195 [01:18<00:07, 5.44it/s]
[2024-05-22 00:31:30] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.3.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
78%|███████▊ | 152/195 [01:18<00:07, 5.44it/s]
79%|███████▉ | 155/195 [01:18<00:06, 6.53it/s]
[2024-05-22 00:31:30] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.3.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
79%|███████▉ | 155/195 [01:19<00:06, 6.53it/s]
80%|████████ | 156/195 [01:19<00:08, 4.72it/s]
[2024-05-22 00:31:30] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.3.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
80%|████████ | 156/195 [01:19<00:08, 4.72it/s]
[2024-05-22 00:31:30] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.3.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
80%|████████ | 156/195 [01:19<00:08, 4.72it/s]
81%|████████ | 158/195 [01:19<00:06, 5.33it/s]
[2024-05-22 00:31:31] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.3.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
81%|████████ | 158/195 [01:19<00:06, 5.33it/s]
[2024-05-22 00:31:31] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.4.input_layernorm.weight[0m", shape: (4096,), dtype: float16
81%|████████ | 158/195 [01:19<00:06, 5.33it/s]
[2024-05-22 00:31:31] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.4.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
81%|████████ | 158/195 [01:19<00:06, 5.33it/s]
83%|████████▎ | 161/195 [01:20<00:05, 6.51it/s]
[2024-05-22 00:31:31] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.4.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
83%|████████▎ | 161/195 [01:20<00:05, 6.51it/s]
83%|████████▎ | 162/195 [01:20<00:06, 4.83it/s]
[2024-05-22 00:31:31] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.4.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
83%|████████▎ | 162/195 [01:20<00:06, 4.83it/s]
[2024-05-22 00:31:31] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.4.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
83%|████████▎ | 162/195 [01:20<00:06, 4.83it/s]
84%|████████▍ | 164/195 [01:20<00:05, 5.43it/s]
[2024-05-22 00:31:32] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.4.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
84%|████████▍ | 164/195 [01:20<00:05, 5.43it/s]
[2024-05-22 00:31:32] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.5.input_layernorm.weight[0m", shape: (4096,), dtype: float16
84%|████████▍ | 164/195 [01:20<00:05, 5.43it/s]
[2024-05-22 00:31:32] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.5.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
84%|████████▍ | 164/195 [01:20<00:05, 5.43it/s]
86%|████████▌ | 167/195 [01:21<00:04, 6.52it/s]
[2024-05-22 00:31:32] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.5.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
86%|████████▌ | 167/195 [01:21<00:04, 6.52it/s]
86%|████████▌ | 168/195 [01:21<00:05, 4.74it/s]
[2024-05-22 00:31:32] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.5.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
86%|████████▌ | 168/195 [01:21<00:05, 4.74it/s]
[2024-05-22 00:31:33] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.5.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
86%|████████▌ | 168/195 [01:21<00:05, 4.74it/s]
87%|████████▋ | 170/195 [01:21<00:04, 5.35it/s]
[2024-05-22 00:31:33] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.5.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
87%|████████▋ | 170/195 [01:21<00:04, 5.35it/s]
[2024-05-22 00:31:33] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.6.input_layernorm.weight[0m", shape: (4096,), dtype: float16
87%|████████▋ | 170/195 [01:21<00:04, 5.35it/s]
[2024-05-22 00:31:33] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.6.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
87%|████████▋ | 170/195 [01:22<00:04, 5.35it/s]
89%|████████▊ | 173/195 [01:22<00:03, 6.53it/s]
[2024-05-22 00:31:33] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.6.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
89%|████████▊ | 173/195 [01:22<00:03, 6.53it/s]
89%|████████▉ | 174/195 [01:22<00:04, 4.84it/s]
[2024-05-22 00:31:34] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.6.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
89%|████████▉ | 174/195 [01:22<00:04, 4.84it/s]
[2024-05-22 00:31:34] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.6.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
89%|████████▉ | 174/195 [01:22<00:04, 4.84it/s]
90%|█████████ | 176/195 [01:22<00:03, 5.38it/s]
[2024-05-22 00:31:34] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.6.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
90%|█████████ | 176/195 [01:22<00:03, 5.38it/s]
[2024-05-22 00:31:34] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.7.input_layernorm.weight[0m", shape: (4096,), dtype: float16
90%|█████████ | 176/195 [01:23<00:03, 5.38it/s]
[2024-05-22 00:31:34] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.7.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
90%|█████████ | 176/195 [01:23<00:03, 5.38it/s]
92%|█████████▏| 179/195 [01:23<00:02, 6.56it/s]
[2024-05-22 00:31:34] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.7.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
92%|█████████▏| 179/195 [01:23<00:02, 6.56it/s]
92%|█████████▏| 180/195 [01:23<00:03, 4.82it/s]
[2024-05-22 00:31:35] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.7.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
92%|█████████▏| 180/195 [01:23<00:03, 4.82it/s]
[2024-05-22 00:31:35] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.7.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
92%|█████████▏| 180/195 [01:23<00:03, 4.82it/s]
93%|█████████▎| 182/195 [01:24<00:02, 5.43it/s]
[2024-05-22 00:31:35] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.7.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
93%|█████████▎| 182/195 [01:24<00:02, 5.43it/s]
[2024-05-22 00:31:35] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.8.input_layernorm.weight[0m", shape: (4096,), dtype: float16
93%|█████████▎| 182/195 [01:24<00:02, 5.43it/s]
[2024-05-22 00:31:35] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.8.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
93%|█████████▎| 182/195 [01:24<00:02, 5.43it/s]
95%|█████████▍| 185/195 [01:24<00:01, 5.69it/s]
[2024-05-22 00:31:35] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.8.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
95%|█████████▍| 185/195 [01:24<00:01, 5.69it/s]
95%|█████████▌| 186/195 [01:25<00:02, 4.42it/s]
[2024-05-22 00:31:36] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.8.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
95%|█████████▌| 186/195 [01:25<00:02, 4.42it/s]
[2024-05-22 00:31:36] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.8.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
95%|█████████▌| 186/195 [01:25<00:02, 4.42it/s]
96%|█████████▋| 188/195 [01:25<00:01, 5.07it/s]
[2024-05-22 00:31:36] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.8.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
96%|█████████▋| 188/195 [01:25<00:01, 5.07it/s]
[2024-05-22 00:31:36] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.9.input_layernorm.weight[0m", shape: (4096,), dtype: float16
96%|█████████▋| 188/195 [01:25<00:01, 5.07it/s]
[2024-05-22 00:31:36] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.9.mlp.down_proj.weight[0m", shape: (4096, 11008), dtype: float16
96%|█████████▋| 188/195 [01:25<00:01, 5.07it/s]
98%|█████████▊| 191/195 [01:25<00:00, 6.28it/s]
[2024-05-22 00:31:37] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.9.mlp.gate_up_proj.weight[0m", shape: (22016, 4096), dtype: float16
98%|█████████▊| 191/195 [01:25<00:00, 6.28it/s]
98%|█████████▊| 192/195 [01:26<00:00, 4.79it/s]
[2024-05-22 00:31:37] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.9.post_attention_layernorm.weight[0m", shape: (4096,), dtype: float16
98%|█████████▊| 192/195 [01:26<00:00, 4.79it/s]
[2024-05-22 00:31:37] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.9.self_attn.qkv_proj.weight[0m", shape: (12288, 4096), dtype: float16
98%|█████████▊| 192/195 [01:26<00:00, 4.79it/s]
99%|█████████▉| 194/195 [01:26<00:00, 5.34it/s]
[2024-05-22 00:31:37] INFO huggingface_loader.py:174: [Not quantized] Parameter: "[1mmodel.layers.9.self_attn.o_proj.weight[0m", shape: (4096, 4096), dtype: float16
99%|█████████▉| 194/195 [01:26<00:00, 5.34it/s]
100%|██████████| 195/195 [01:26<00:00, 2.26it/s]
[2024-05-22 00:31:37] INFO huggingface_loader.py:196: Unloading HF weight file: /models/Llama-2-7b-chat-hf/pytorch_model-00001-of-00002.bin
[2024-05-22 00:31:38] INFO stats.py:76: [92mTime usage[0m: HF loading: 44.115 sec; Pre-quantization mapping: 7.475 sec; Quantization: 0.000 sec
[2024-05-22 00:31:38] INFO stats.py:90: [92mRAM usage[0m: Peak RAM: 9.291 GB. Total bytes loaded from disk: 12.551 GB
[2024-05-22 00:31:38] INFO convert_weight.py:155: [92mParameter size[0m after quantization: 12.551 GB
[2024-05-22 00:31:38] INFO convert_weight.py:160: [92mTotal parameters[0m: 6,738,415,616
[2024-05-22 00:31:38] INFO convert_weight.py:161: [92mBits per parameter[0m: 16.000
[2024-05-22 00:31:38] INFO convert_weight.py:166: Saved to directory: [1m/tmp/tmp8ew5r2yr[0m
All finished, 131 total shards committed, record saved to /tmp/tmp8ew5r2yr/ndarray-cache.json
|