/opt/conda/envs/py310/bin/python -m mlc_llm gen_config /models/Mistral-7B-Instruct-v0.3 --quantization q4f32_1 --conv-template mistral_default --output /models/mlc-delivery/hf/mlc-ai/Mistral-7B-Instruct-v0.3-q4f32_1-MLC [2024-06-02 06:23:28] INFO auto_config.py:116: Found model configuration: /models/Mistral-7B-Instruct-v0.3/config.json [2024-06-02 06:23:28] INFO auto_config.py:154: Found model type: mistral. Use `--model-type` to override. [2024-06-02 06:23:28] INFO mistral_model.py:56: prefill_chunk_size defaults to 2048 [2024-06-02 06:23:28] INFO config.py:107: Overriding max_batch_size from 1 to 80 [2024-06-02 06:23:28] INFO gen_config.py:143: [generation_config.json] Setting bos_token_id: 1 [2024-06-02 06:23:28] INFO gen_config.py:143: [generation_config.json] Setting eos_token_id: 2 [2024-06-02 06:23:28] INFO gen_config.py:155: Found tokenizer config: /models/Mistral-7B-Instruct-v0.3/tokenizer.model. Copying to /models/mlc-delivery/hf/mlc-ai/Mistral-7B-Instruct-v0.3-q4f32_1-MLC/tokenizer.model [2024-06-02 06:23:28] INFO gen_config.py:155: Found tokenizer config: /models/Mistral-7B-Instruct-v0.3/tokenizer.json. Copying to /models/mlc-delivery/hf/mlc-ai/Mistral-7B-Instruct-v0.3-q4f32_1-MLC/tokenizer.json [2024-06-02 06:23:28] INFO gen_config.py:157: Not found tokenizer config: /models/Mistral-7B-Instruct-v0.3/vocab.json [2024-06-02 06:23:28] INFO gen_config.py:157: Not found tokenizer config: /models/Mistral-7B-Instruct-v0.3/merges.txt [2024-06-02 06:23:28] INFO gen_config.py:157: Not found tokenizer config: /models/Mistral-7B-Instruct-v0.3/added_tokens.json [2024-06-02 06:23:28] INFO gen_config.py:155: Found tokenizer config: /models/Mistral-7B-Instruct-v0.3/tokenizer_config.json. Copying to /models/mlc-delivery/hf/mlc-ai/Mistral-7B-Instruct-v0.3-q4f32_1-MLC/tokenizer_config.json [2024-06-02 06:23:28] INFO gen_config.py:216: Detected tokenizer info: {'token_postproc_method': 'byte_fallback', 'prepend_space_in_encode': False, 'strip_space_in_decode': True} [2024-06-02 06:23:28] INFO gen_config.py:32: [System default] Setting pad_token_id: 0 [2024-06-02 06:23:28] INFO gen_config.py:32: [System default] Setting temperature: 1.0 [2024-06-02 06:23:28] INFO gen_config.py:32: [System default] Setting presence_penalty: 0.0 [2024-06-02 06:23:28] INFO gen_config.py:32: [System default] Setting frequency_penalty: 0.0 [2024-06-02 06:23:28] INFO gen_config.py:32: [System default] Setting repetition_penalty: 1.0 [2024-06-02 06:23:28] INFO gen_config.py:32: [System default] Setting top_p: 1.0 [2024-06-02 06:23:28] INFO gen_config.py:32: [System default] Setting mean_gen_len: 128 [2024-06-02 06:23:28] INFO gen_config.py:32: [System default] Setting max_gen_len: 512 [2024-06-02 06:23:28] INFO gen_config.py:32: [System default] Setting shift_fill_factor: 0.3 [2024-06-02 06:23:28] INFO gen_config.py:223: Dumping configuration file to: /models/mlc-delivery/hf/mlc-ai/Mistral-7B-Instruct-v0.3-q4f32_1-MLC/mlc-chat-config.json /opt/conda/envs/py310/bin/python -m mlc_llm convert_weight /models/Mistral-7B-Instruct-v0.3 --quantization q4f32_1 --output /models/mlc-delivery/hf/mlc-ai/Mistral-7B-Instruct-v0.3-q4f32_1-MLC [2024-06-02 06:23:29] INFO auto_config.py:116: Found model configuration: /models/Mistral-7B-Instruct-v0.3/config.json [2024-06-02 06:23:31] INFO auto_device.py:79: Found device: cuda:0 [2024-06-02 06:23:32] INFO auto_device.py:88: Not found device: rocm:0 [2024-06-02 06:23:34] INFO auto_device.py:88: Not found device: metal:0 [2024-06-02 06:23:35] INFO auto_device.py:79: Found device: vulkan:0 [2024-06-02 06:23:35] INFO auto_device.py:79: Found device: vulkan:1 [2024-06-02 06:23:35] INFO auto_device.py:79: Found device: vulkan:2 [2024-06-02 06:23:35] INFO auto_device.py:79: Found device: vulkan:3 [2024-06-02 06:23:37] INFO auto_device.py:88: Not found device: opencl:0 [2024-06-02 06:23:37] INFO auto_device.py:35: Using device: cuda:0 [2024-06-02 06:23:37] INFO auto_weight.py:71: Finding weights in: /models/Mistral-7B-Instruct-v0.3 [2024-06-02 06:23:37] INFO auto_weight.py:137: Not found Huggingface PyTorch [2024-06-02 06:23:37] INFO auto_weight.py:144: Found source weight format: huggingface-safetensor. Source configuration: /models/Mistral-7B-Instruct-v0.3/model.safetensors.index.json [2024-06-02 06:23:37] INFO auto_weight.py:107: Using source weight configuration: /models/Mistral-7B-Instruct-v0.3/model.safetensors.index.json. Use `--source` to override. [2024-06-02 06:23:37] INFO auto_weight.py:111: Using source weight format: huggingface-safetensor. Use `--source-format` to override. [2024-06-02 06:23:37] INFO auto_config.py:154: Found model type: mistral. Use `--model-type` to override. [2024-06-02 06:23:37] INFO mistral_model.py:56: prefill_chunk_size defaults to 2048 Weight conversion with arguments: --config /models/Mistral-7B-Instruct-v0.3/config.json --quantization GroupQuantize(name='q4f32_1', kind='group-quant', group_size=32, quantize_dtype='int4', storage_dtype='uint32', model_dtype='float32', linear_weight_layout='NK', quantize_embedding=True, quantize_final_fc=True, num_elem_per_storage=8, num_storage_per_group=4, max_int_value=7) --model-type mistral --device cuda:0 --source /models/Mistral-7B-Instruct-v0.3/model.safetensors.index.json --source-format huggingface-safetensor --output /models/mlc-delivery/hf/mlc-ai/Mistral-7B-Instruct-v0.3-q4f32_1-MLC Start storing to cache /models/mlc-delivery/hf/mlc-ai/Mistral-7B-Instruct-v0.3-q4f32_1-MLC 0%| | 0/195 [00:00 type is zero. setattr(self, word, getattr(machar, word).flat[0]) /home/rickzhou/miniconda3/envs/mlc/lib/python3.11/site-packages/numpy/core/getlimits.py:89: UserWarning: The value of the smallest subnormal for type is zero. return self._float_to_str(self.smallest_subnormal) 1%| | 1/195 [00:03<10:18, 3.19s/it] [2024-06-02 04:15:35] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.22.input_layernorm.weight", shape: (4096,), dtype: float32 1%| | 1/195 [00:03<10:18, 3.19s/it] [2024-06-02 04:15:35] INFO group_quantization.py:217: Compiling quantize function for key: ((4096, 14336), float32, cuda, axis=1, output_transpose=False) 1%| | 1/195 [00:03<10:18, 3.19s/it] [2024-06-02 04:15:36] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.22.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 1%| | 1/195 [00:03<10:18, 3.19s/it] [2024-06-02 04:15:36] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.22.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 1%| | 1/195 [00:03<10:18, 3.19s/it] 2%|▏ | 3/195 [00:03<03:12, 1.00s/it] [2024-06-02 04:15:36] INFO group_quantization.py:217: Compiling quantize function for key: ((28672, 4096), float32, cuda, axis=1, output_transpose=False) 2%|▏ | 3/195 [00:03<03:12, 1.00s/it] [2024-06-02 04:15:36] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.22.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 2%|▏ | 3/195 [00:04<03:12, 1.00s/it] [2024-06-02 04:15:36] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.22.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 2%|▏ | 3/195 [00:04<03:12, 1.00s/it] 2%|▏ | 4/195 [00:04<02:52, 1.11it/s] [2024-06-02 04:15:36] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.22.post_attention_layernorm.weight", shape: (4096,), dtype: float32 2%|▏ | 4/195 [00:04<02:52, 1.11it/s] [2024-06-02 04:15:36] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.23.input_layernorm.weight", shape: (4096,), dtype: float32 2%|▏ | 4/195 [00:04<02:52, 1.11it/s] [2024-06-02 04:15:36] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.23.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 2%|▏ | 4/195 [00:04<02:52, 1.11it/s] [2024-06-02 04:15:37] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.23.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 2%|▏ | 4/195 [00:04<02:52, 1.11it/s] 4%|▎ | 7/195 [00:04<01:15, 2.50it/s] [2024-06-02 04:15:37] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.23.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 4%|▎ | 7/195 [00:04<01:15, 2.50it/s] [2024-06-02 04:15:37] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.23.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 4%|▎ | 7/195 [00:04<01:15, 2.50it/s] 4%|▍ | 8/195 [00:04<01:16, 2.46it/s] [2024-06-02 04:15:37] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.23.post_attention_layernorm.weight", shape: (4096,), dtype: float32 4%|▍ | 8/195 [00:05<01:16, 2.46it/s] [2024-06-02 04:15:37] INFO group_quantization.py:217: Compiling quantize function for key: ((6144, 4096), float32, cuda, axis=1, output_transpose=False) 4%|▍ | 8/195 [00:05<01:16, 2.46it/s] [2024-06-02 04:15:37] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.23.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 4%|▍ | 8/195 [00:05<01:16, 2.46it/s] [2024-06-02 04:15:37] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.23.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 4%|▍ | 8/195 [00:05<01:16, 2.46it/s] 5%|▌ | 10/195 [00:05<01:02, 2.98it/s] [2024-06-02 04:15:37] INFO group_quantization.py:217: Compiling quantize function for key: ((4096, 4096), float32, cuda, axis=1, output_transpose=False) 5%|▌ | 10/195 [00:05<01:02, 2.98it/s] [2024-06-02 04:15:38] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.23.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 5%|▌ | 10/195 [00:05<01:02, 2.98it/s] [2024-06-02 04:15:38] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.23.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 5%|▌ | 10/195 [00:05<01:02, 2.98it/s] 6%|▌ | 11/195 [00:05<01:00, 3.04it/s] [2024-06-02 04:15:38] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.24.input_layernorm.weight", shape: (4096,), dtype: float32 6%|▌ | 11/195 [00:05<01:00, 3.04it/s] [2024-06-02 04:15:38] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.24.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 6%|▌ | 11/195 [00:05<01:00, 3.04it/s] [2024-06-02 04:15:38] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.24.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 6%|▌ | 11/195 [00:05<01:00, 3.04it/s] 7%|▋ | 13/195 [00:05<00:43, 4.21it/s] [2024-06-02 04:15:38] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.24.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 7%|▋ | 13/195 [00:06<00:43, 4.21it/s] [2024-06-02 04:15:38] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.24.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 7%|▋ | 13/195 [00:06<00:43, 4.21it/s] 7%|▋ | 14/195 [00:06<00:54, 3.33it/s] [2024-06-02 04:15:38] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.24.post_attention_layernorm.weight", shape: (4096,), dtype: float32 7%|▋ | 14/195 [00:06<00:54, 3.33it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.24.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 7%|▋ | 14/195 [00:06<00:54, 3.33it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.24.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 7%|▋ | 14/195 [00:06<00:54, 3.33it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.24.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 7%|▋ | 14/195 [00:06<00:54, 3.33it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.24.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 7%|▋ | 14/195 [00:06<00:54, 3.33it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.25.input_layernorm.weight", shape: (4096,), dtype: float32 7%|▋ | 14/195 [00:06<00:54, 3.33it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.25.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 7%|▋ | 14/195 [00:06<00:54, 3.33it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.25.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 7%|▋ | 14/195 [00:06<00:54, 3.33it/s] 10%|▉ | 19/195 [00:06<00:28, 6.13it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.25.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 10%|▉ | 19/195 [00:06<00:28, 6.13it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.25.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 10%|▉ | 19/195 [00:07<00:28, 6.13it/s] 10%|█ | 20/195 [00:07<00:35, 5.00it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.25.post_attention_layernorm.weight", shape: (4096,), dtype: float32 10%|█ | 20/195 [00:07<00:35, 5.00it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.25.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 10%|█ | 20/195 [00:07<00:35, 5.00it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.25.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 10%|█ | 20/195 [00:07<00:35, 5.00it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.25.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 10%|█ | 20/195 [00:07<00:35, 5.00it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.25.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 10%|█ | 20/195 [00:07<00:35, 5.00it/s] 12%|█▏ | 23/195 [00:07<00:24, 7.00it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.26.input_layernorm.weight", shape: (4096,), dtype: float32 12%|█▏ | 23/195 [00:07<00:24, 7.00it/s] [2024-06-02 04:15:39] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.26.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 12%|█▏ | 23/195 [00:07<00:24, 7.00it/s] [2024-06-02 04:15:40] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.26.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 12%|█▏ | 23/195 [00:07<00:24, 7.00it/s] 13%|█▎ | 25/195 [00:07<00:21, 7.78it/s] [2024-06-02 04:15:40] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.26.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 13%|█▎ | 25/195 [00:07<00:21, 7.78it/s] [2024-06-02 04:15:40] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.26.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 13%|█▎ | 25/195 [00:07<00:21, 7.78it/s] [2024-06-02 04:15:40] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.26.post_attention_layernorm.weight", shape: (4096,), dtype: float32 13%|█▎ | 25/195 [00:07<00:21, 7.78it/s] 14%|█▍ | 27/195 [00:07<00:25, 6.59it/s] [2024-06-02 04:15:40] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.26.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 14%|█▍ | 27/195 [00:08<00:25, 6.59it/s] [2024-06-02 04:15:40] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.26.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 14%|█▍ | 27/195 [00:08<00:25, 6.59it/s] 14%|█▍ | 28/195 [00:08<00:25, 6.54it/s] [2024-06-02 04:15:40] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.26.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 14%|█▍ | 28/195 [00:08<00:25, 6.54it/s] [2024-06-02 04:15:40] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.26.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 14%|█▍ | 28/195 [00:08<00:25, 6.54it/s] [2024-06-02 04:15:40] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.27.input_layernorm.weight", shape: (4096,), dtype: float32 14%|█▍ | 28/195 [00:08<00:25, 6.54it/s] [2024-06-02 04:15:40] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.27.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 14%|█▍ | 28/195 [00:08<00:25, 6.54it/s] [2024-06-02 04:15:40] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.27.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 14%|█▍ | 28/195 [00:08<00:25, 6.54it/s] 16%|█▌ | 31/195 [00:08<00:18, 8.65it/s] [2024-06-02 04:15:41] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.27.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 16%|█▌ | 31/195 [00:08<00:18, 8.65it/s] [2024-06-02 04:15:41] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.27.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 16%|█▌ | 31/195 [00:08<00:18, 8.65it/s] [2024-06-02 04:15:41] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.27.post_attention_layernorm.weight", shape: (4096,), dtype: float32 16%|█▌ | 31/195 [00:08<00:18, 8.65it/s] 17%|█▋ | 33/195 [00:08<00:22, 7.06it/s] [2024-06-02 04:15:41] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.27.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 17%|█▋ | 33/195 [00:08<00:22, 7.06it/s] [2024-06-02 04:15:41] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.27.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 17%|█▋ | 33/195 [00:08<00:22, 7.06it/s] 17%|█▋ | 34/195 [00:08<00:21, 7.34it/s] [2024-06-02 04:15:41] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.27.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 17%|█▋ | 34/195 [00:08<00:21, 7.34it/s] [2024-06-02 04:15:41] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.27.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 17%|█▋ | 34/195 [00:08<00:21, 7.34it/s] [2024-06-02 04:15:41] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.28.input_layernorm.weight", shape: (4096,), dtype: float32 17%|█▋ | 34/195 [00:08<00:21, 7.34it/s] [2024-06-02 04:15:41] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.28.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 17%|█▋ | 34/195 [00:08<00:21, 7.34it/s] [2024-06-02 04:15:41] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.28.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 17%|█▋ | 34/195 [00:09<00:21, 7.34it/s] 19%|█▉ | 37/195 [00:09<00:16, 9.38it/s] [2024-06-02 04:15:41] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.28.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 19%|█▉ | 37/195 [00:09<00:16, 9.38it/s] [2024-06-02 04:15:41] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.28.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 19%|█▉ | 37/195 [00:09<00:16, 9.38it/s] [2024-06-02 04:15:42] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.28.post_attention_layernorm.weight", shape: (4096,), dtype: float32 19%|█▉ | 37/195 [00:09<00:16, 9.38it/s] 20%|██ | 39/195 [00:09<00:23, 6.60it/s] [2024-06-02 04:15:42] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.28.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 20%|██ | 39/195 [00:09<00:23, 6.60it/s] [2024-06-02 04:15:42] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.28.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 20%|██ | 39/195 [00:09<00:23, 6.60it/s] [2024-06-02 04:15:42] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.28.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 20%|██ | 39/195 [00:09<00:23, 6.60it/s] [2024-06-02 04:15:42] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.28.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 20%|██ | 39/195 [00:09<00:23, 6.60it/s] [2024-06-02 04:15:42] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.29.input_layernorm.weight", shape: (4096,), dtype: float32 20%|██ | 39/195 [00:09<00:23, 6.60it/s] [2024-06-02 04:15:42] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.29.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 20%|██ | 39/195 [00:09<00:23, 6.60it/s] [2024-06-02 04:15:42] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.29.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 20%|██ | 39/195 [00:09<00:23, 6.60it/s] 22%|██▏ | 43/195 [00:09<00:18, 8.11it/s] [2024-06-02 04:15:42] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.29.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 22%|██▏ | 43/195 [00:10<00:18, 8.11it/s] [2024-06-02 04:15:42] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.29.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 22%|██▏ | 43/195 [00:10<00:18, 8.11it/s] 23%|██▎ | 44/195 [00:10<00:24, 6.16it/s] [2024-06-02 04:15:42] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.29.post_attention_layernorm.weight", shape: (4096,), dtype: float32 23%|██▎ | 44/195 [00:10<00:24, 6.16it/s] [2024-06-02 04:15:42] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.29.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 23%|██▎ | 44/195 [00:10<00:24, 6.16it/s] [2024-06-02 04:15:42] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.29.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 23%|██▎ | 44/195 [00:10<00:24, 6.16it/s] [2024-06-02 04:15:42] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.29.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 23%|██▎ | 44/195 [00:10<00:24, 6.16it/s] [2024-06-02 04:15:43] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.29.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 23%|██▎ | 44/195 [00:10<00:24, 6.16it/s] 24%|██▍ | 47/195 [00:10<00:17, 8.40it/s] [2024-06-02 04:15:43] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.30.input_layernorm.weight", shape: (4096,), dtype: float32 24%|██▍ | 47/195 [00:10<00:17, 8.40it/s] [2024-06-02 04:15:43] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.30.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 24%|██▍ | 47/195 [00:10<00:17, 8.40it/s] [2024-06-02 04:15:43] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.30.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 24%|██▍ | 47/195 [00:10<00:17, 8.40it/s] 25%|██▌ | 49/195 [00:10<00:16, 9.03it/s] [2024-06-02 04:15:43] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.30.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 25%|██▌ | 49/195 [00:10<00:16, 9.03it/s] [2024-06-02 04:15:43] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.30.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 25%|██▌ | 49/195 [00:11<00:16, 9.03it/s] [2024-06-02 04:15:43] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.30.post_attention_layernorm.weight", shape: (4096,), dtype: float32 25%|██▌ | 49/195 [00:11<00:16, 9.03it/s] 26%|██▌ | 51/195 [00:11<00:19, 7.33it/s] [2024-06-02 04:15:43] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.30.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 26%|██▌ | 51/195 [00:11<00:19, 7.33it/s] [2024-06-02 04:15:43] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.30.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 26%|██▌ | 51/195 [00:11<00:19, 7.33it/s] 27%|██▋ | 52/195 [00:11<00:19, 7.18it/s] [2024-06-02 04:15:43] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.30.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 27%|██▋ | 52/195 [00:11<00:19, 7.18it/s] [2024-06-02 04:15:43] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.30.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 27%|██▋ | 52/195 [00:11<00:19, 7.18it/s] [2024-06-02 04:15:43] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.31.input_layernorm.weight", shape: (4096,), dtype: float32 27%|██▋ | 52/195 [00:11<00:19, 7.18it/s] [2024-06-02 04:15:43] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.31.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 27%|██▋ | 52/195 [00:11<00:19, 7.18it/s] [2024-06-02 04:15:43] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.31.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 27%|██▋ | 52/195 [00:11<00:19, 7.18it/s] 28%|██▊ | 55/195 [00:11<00:15, 9.25it/s] [2024-06-02 04:15:44] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.31.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 28%|██▊ | 55/195 [00:11<00:15, 9.25it/s] [2024-06-02 04:15:44] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.31.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 28%|██▊ | 55/195 [00:11<00:15, 9.25it/s] [2024-06-02 04:15:44] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.31.post_attention_layernorm.weight", shape: (4096,), dtype: float32 28%|██▊ | 55/195 [00:11<00:15, 9.25it/s] 29%|██▉ | 57/195 [00:11<00:18, 7.31it/s] [2024-06-02 04:15:44] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.31.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 29%|██▉ | 57/195 [00:11<00:18, 7.31it/s] [2024-06-02 04:15:44] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.31.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 29%|██▉ | 57/195 [00:11<00:18, 7.31it/s] 30%|██▉ | 58/195 [00:11<00:18, 7.57it/s] [2024-06-02 04:15:44] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.31.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 30%|██▉ | 58/195 [00:11<00:18, 7.57it/s] [2024-06-02 04:15:44] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.31.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 30%|██▉ | 58/195 [00:11<00:18, 7.57it/s] [2024-06-02 04:15:44] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.norm.weight", shape: (4096,), dtype: float32 30%|██▉ | 58/195 [00:11<00:18, 7.57it/s] [2024-06-02 04:15:44] INFO huggingface_loader.py:196: Unloading HF weight file: /ssd1/rickzhou/models/Mistral-7B-Instruct-v0.3/model-00003-of-00003.safetensors 30%|██▉ | 58/195 [00:11<00:18, 7.57it/s] [2024-06-02 04:15:44] INFO huggingface_loader.py:184: Loading HF parameters from: /ssd1/rickzhou/models/Mistral-7B-Instruct-v0.3/model-00001-of-00003.safetensors 30%|██▉ | 58/195 [00:12<00:18, 7.57it/s] [2024-06-02 04:15:45] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.embed_tokens.q_weight", shape: (32768, 512), dtype: uint32 30%|██▉ | 58/195 [00:12<00:18, 7.57it/s] [2024-06-02 04:15:45] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.embed_tokens.q_scale", shape: (32768, 128), dtype: float32 30%|██▉ | 58/195 [00:13<00:18, 7.57it/s] 31%|███▏ | 61/195 [00:13<00:36, 3.72it/s] [2024-06-02 04:15:45] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.0.input_layernorm.weight", shape: (4096,), dtype: float32 31%|███▏ | 61/195 [00:13<00:36, 3.72it/s] [2024-06-02 04:15:45] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.0.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 31%|███▏ | 61/195 [00:13<00:36, 3.72it/s] [2024-06-02 04:15:46] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.0.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 31%|███▏ | 61/195 [00:13<00:36, 3.72it/s] 32%|███▏ | 63/195 [00:13<00:30, 4.27it/s] [2024-06-02 04:15:46] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.0.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 32%|███▏ | 63/195 [00:13<00:30, 4.27it/s] [2024-06-02 04:15:46] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.0.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 32%|███▏ | 63/195 [00:14<00:30, 4.27it/s] 33%|███▎ | 64/195 [00:14<00:34, 3.76it/s] [2024-06-02 04:15:46] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.0.post_attention_layernorm.weight", shape: (4096,), dtype: float32 33%|███▎ | 64/195 [00:14<00:34, 3.76it/s] [2024-06-02 04:15:46] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.0.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 33%|███▎ | 64/195 [00:14<00:34, 3.76it/s] [2024-06-02 04:15:46] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.0.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 33%|███▎ | 64/195 [00:14<00:34, 3.76it/s] [2024-06-02 04:15:46] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.0.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 33%|███▎ | 64/195 [00:14<00:34, 3.76it/s] [2024-06-02 04:15:46] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.0.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 33%|███▎ | 64/195 [00:14<00:34, 3.76it/s] 34%|███▍ | 67/195 [00:14<00:22, 5.72it/s] [2024-06-02 04:15:46] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.1.input_layernorm.weight", shape: (4096,), dtype: float32 34%|███▍ | 67/195 [00:14<00:22, 5.72it/s] [2024-06-02 04:15:46] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.1.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 34%|███▍ | 67/195 [00:14<00:22, 5.72it/s] [2024-06-02 04:15:46] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.1.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 34%|███▍ | 67/195 [00:14<00:22, 5.72it/s] 35%|███▌ | 69/195 [00:14<00:18, 6.66it/s] [2024-06-02 04:15:47] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.1.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 35%|███▌ | 69/195 [00:14<00:18, 6.66it/s] [2024-06-02 04:15:47] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.1.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 35%|███▌ | 69/195 [00:14<00:18, 6.66it/s] [2024-06-02 04:15:47] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.1.post_attention_layernorm.weight", shape: (4096,), dtype: float32 35%|███▌ | 69/195 [00:14<00:18, 6.66it/s] 36%|███▋ | 71/195 [00:14<00:20, 5.95it/s] [2024-06-02 04:15:47] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.1.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 36%|███▋ | 71/195 [00:14<00:20, 5.95it/s] [2024-06-02 04:15:47] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.1.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 36%|███▋ | 71/195 [00:14<00:20, 5.95it/s] 37%|███▋ | 72/195 [00:14<00:20, 5.88it/s] [2024-06-02 04:15:47] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.1.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 37%|███▋ | 72/195 [00:14<00:20, 5.88it/s] [2024-06-02 04:15:47] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.1.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 37%|███▋ | 72/195 [00:14<00:20, 5.88it/s] [2024-06-02 04:15:47] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.10.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 37%|███▋ | 72/195 [00:15<00:20, 5.88it/s] [2024-06-02 04:15:47] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.10.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 37%|███▋ | 72/195 [00:15<00:20, 5.88it/s] 38%|███▊ | 74/195 [00:15<00:22, 5.39it/s] [2024-06-02 04:15:47] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.10.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 38%|███▊ | 74/195 [00:15<00:22, 5.39it/s] [2024-06-02 04:15:47] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.10.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 38%|███▊ | 74/195 [00:15<00:22, 5.39it/s] [2024-06-02 04:15:47] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.10.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 38%|███▊ | 74/195 [00:15<00:22, 5.39it/s] [2024-06-02 04:15:48] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.10.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 38%|███▊ | 74/195 [00:15<00:22, 5.39it/s] 39%|███▉ | 76/195 [00:15<00:18, 6.39it/s] [2024-06-02 04:15:48] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.2.input_layernorm.weight", shape: (4096,), dtype: float32 39%|███▉ | 76/195 [00:15<00:18, 6.39it/s] [2024-06-02 04:15:48] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.2.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 39%|███▉ | 76/195 [00:15<00:18, 6.39it/s] [2024-06-02 04:15:48] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.2.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 39%|███▉ | 76/195 [00:15<00:18, 6.39it/s] 40%|████ | 78/195 [00:15<00:15, 7.44it/s] [2024-06-02 04:15:48] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.2.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 40%|████ | 78/195 [00:15<00:15, 7.44it/s] [2024-06-02 04:15:48] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.2.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 40%|████ | 78/195 [00:16<00:15, 7.44it/s] 41%|████ | 79/195 [00:16<00:21, 5.32it/s] [2024-06-02 04:15:48] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.2.post_attention_layernorm.weight", shape: (4096,), dtype: float32 41%|████ | 79/195 [00:16<00:21, 5.32it/s] [2024-06-02 04:15:48] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.2.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 41%|████ | 79/195 [00:16<00:21, 5.32it/s] [2024-06-02 04:15:48] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.2.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 41%|████ | 79/195 [00:16<00:21, 5.32it/s] 42%|████▏ | 81/195 [00:16<00:17, 6.51it/s] [2024-06-02 04:15:48] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.2.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 42%|████▏ | 81/195 [00:16<00:17, 6.51it/s] [2024-06-02 04:15:48] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.2.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 42%|████▏ | 81/195 [00:16<00:17, 6.51it/s] [2024-06-02 04:15:48] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.3.input_layernorm.weight", shape: (4096,), dtype: float32 42%|████▏ | 81/195 [00:16<00:17, 6.51it/s] [2024-06-02 04:15:48] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.3.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 42%|████▏ | 81/195 [00:16<00:17, 6.51it/s] [2024-06-02 04:15:49] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.3.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 42%|████▏ | 81/195 [00:16<00:17, 6.51it/s] 43%|████▎ | 84/195 [00:16<00:12, 8.72it/s] [2024-06-02 04:15:49] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.3.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 43%|████▎ | 84/195 [00:16<00:12, 8.72it/s] [2024-06-02 04:15:49] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.3.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 43%|████▎ | 84/195 [00:16<00:12, 8.72it/s] [2024-06-02 04:15:49] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.3.post_attention_layernorm.weight", shape: (4096,), dtype: float32 43%|████▎ | 84/195 [00:16<00:12, 8.72it/s] 44%|████▍ | 86/195 [00:16<00:15, 7.08it/s] [2024-06-02 04:15:49] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.3.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 44%|████▍ | 86/195 [00:17<00:15, 7.08it/s] [2024-06-02 04:15:49] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.3.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 44%|████▍ | 86/195 [00:17<00:15, 7.08it/s] 45%|████▍ | 87/195 [00:17<00:15, 7.19it/s] [2024-06-02 04:15:49] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.3.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 45%|████▍ | 87/195 [00:17<00:15, 7.19it/s] [2024-06-02 04:15:49] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.3.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 45%|████▍ | 87/195 [00:17<00:15, 7.19it/s] [2024-06-02 04:15:49] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.4.input_layernorm.weight", shape: (4096,), dtype: float32 45%|████▍ | 87/195 [00:17<00:15, 7.19it/s] [2024-06-02 04:15:49] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.4.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 45%|████▍ | 87/195 [00:17<00:15, 7.19it/s] [2024-06-02 04:15:49] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.4.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 45%|████▍ | 87/195 [00:17<00:15, 7.19it/s] 46%|████▌ | 90/195 [00:17<00:11, 9.27it/s] [2024-06-02 04:15:49] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.4.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 46%|████▌ | 90/195 [00:17<00:11, 9.27it/s] [2024-06-02 04:15:50] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.4.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 46%|████▌ | 90/195 [00:17<00:11, 9.27it/s] [2024-06-02 04:15:50] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.4.post_attention_layernorm.weight", shape: (4096,), dtype: float32 46%|████▌ | 90/195 [00:17<00:11, 9.27it/s] 47%|████▋ | 92/195 [00:17<00:15, 6.59it/s] [2024-06-02 04:15:50] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.4.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 47%|████▋ | 92/195 [00:17<00:15, 6.59it/s] [2024-06-02 04:15:50] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.4.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 47%|████▋ | 92/195 [00:17<00:15, 6.59it/s] [2024-06-02 04:15:50] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.4.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 47%|████▋ | 92/195 [00:17<00:15, 6.59it/s] [2024-06-02 04:15:50] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.4.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 47%|████▋ | 92/195 [00:17<00:15, 6.59it/s] [2024-06-02 04:15:50] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.5.input_layernorm.weight", shape: (4096,), dtype: float32 47%|████▋ | 92/195 [00:17<00:15, 6.59it/s] [2024-06-02 04:15:50] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.5.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 47%|████▋ | 92/195 [00:17<00:15, 6.59it/s] [2024-06-02 04:15:50] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.5.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 47%|████▋ | 92/195 [00:18<00:15, 6.59it/s] 49%|████▉ | 96/195 [00:18<00:12, 8.05it/s] [2024-06-02 04:15:50] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.5.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 49%|████▉ | 96/195 [00:18<00:12, 8.05it/s] [2024-06-02 04:15:51] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.5.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 49%|████▉ | 96/195 [00:18<00:12, 8.05it/s] 50%|████▉ | 97/195 [00:18<00:16, 6.10it/s] [2024-06-02 04:15:51] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.5.post_attention_layernorm.weight", shape: (4096,), dtype: float32 50%|████▉ | 97/195 [00:18<00:16, 6.10it/s] [2024-06-02 04:15:51] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.5.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 50%|████▉ | 97/195 [00:18<00:16, 6.10it/s] [2024-06-02 04:15:51] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.5.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 50%|████▉ | 97/195 [00:18<00:16, 6.10it/s] [2024-06-02 04:15:51] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.5.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 50%|████▉ | 97/195 [00:18<00:16, 6.10it/s] [2024-06-02 04:15:51] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.5.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 50%|████▉ | 97/195 [00:18<00:16, 6.10it/s] 51%|█████▏ | 100/195 [00:18<00:11, 8.16it/s] [2024-06-02 04:15:51] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.6.input_layernorm.weight", shape: (4096,), dtype: float32 51%|█████▏ | 100/195 [00:18<00:11, 8.16it/s] [2024-06-02 04:15:51] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.6.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 51%|█████▏ | 100/195 [00:18<00:11, 8.16it/s] [2024-06-02 04:15:51] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.6.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 51%|█████▏ | 100/195 [00:18<00:11, 8.16it/s] 52%|█████▏ | 102/195 [00:18<00:10, 8.85it/s] [2024-06-02 04:15:51] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.6.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 52%|█████▏ | 102/195 [00:19<00:10, 8.85it/s] [2024-06-02 04:15:51] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.6.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 52%|█████▏ | 102/195 [00:19<00:10, 8.85it/s] [2024-06-02 04:15:51] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.6.post_attention_layernorm.weight", shape: (4096,), dtype: float32 52%|█████▏ | 102/195 [00:19<00:10, 8.85it/s] 53%|█████▎ | 104/195 [00:19<00:12, 7.23it/s] [2024-06-02 04:15:51] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.6.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 53%|█████▎ | 104/195 [00:19<00:12, 7.23it/s] [2024-06-02 04:15:51] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.6.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 53%|█████▎ | 104/195 [00:19<00:12, 7.23it/s] 54%|█████▍ | 105/195 [00:19<00:12, 6.94it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.6.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 54%|█████▍ | 105/195 [00:19<00:12, 6.94it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.6.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 54%|█████▍ | 105/195 [00:19<00:12, 6.94it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.7.input_layernorm.weight", shape: (4096,), dtype: float32 54%|█████▍ | 105/195 [00:19<00:12, 6.94it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.7.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 54%|█████▍ | 105/195 [00:19<00:12, 6.94it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.7.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 54%|█████▍ | 105/195 [00:19<00:12, 6.94it/s] 55%|█████▌ | 108/195 [00:19<00:09, 9.19it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.7.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 55%|█████▌ | 108/195 [00:19<00:09, 9.19it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.7.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 55%|█████▌ | 108/195 [00:20<00:09, 9.19it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.7.post_attention_layernorm.weight", shape: (4096,), dtype: float32 55%|█████▌ | 108/195 [00:20<00:09, 9.19it/s] 56%|█████▋ | 110/195 [00:20<00:11, 7.31it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.7.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 56%|█████▋ | 110/195 [00:20<00:11, 7.31it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.7.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 56%|█████▋ | 110/195 [00:20<00:11, 7.31it/s] 57%|█████▋ | 111/195 [00:20<00:11, 7.44it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.7.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 57%|█████▋ | 111/195 [00:20<00:11, 7.44it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.7.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 57%|█████▋ | 111/195 [00:20<00:11, 7.44it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.8.input_layernorm.weight", shape: (4096,), dtype: float32 57%|█████▋ | 111/195 [00:20<00:11, 7.44it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.8.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 57%|█████▋ | 111/195 [00:20<00:11, 7.44it/s] [2024-06-02 04:15:52] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.8.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 57%|█████▋ | 111/195 [00:20<00:11, 7.44it/s] 58%|█████▊ | 114/195 [00:20<00:08, 9.49it/s] [2024-06-02 04:15:53] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.8.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 58%|█████▊ | 114/195 [00:20<00:08, 9.49it/s] [2024-06-02 04:15:53] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.8.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 58%|█████▊ | 114/195 [00:20<00:08, 9.49it/s] [2024-06-02 04:15:53] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.8.post_attention_layernorm.weight", shape: (4096,), dtype: float32 58%|█████▊ | 114/195 [00:20<00:08, 9.49it/s] 59%|█████▉ | 116/195 [00:20<00:12, 6.58it/s] [2024-06-02 04:15:53] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.8.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 59%|█████▉ | 116/195 [00:20<00:12, 6.58it/s] [2024-06-02 04:15:53] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.8.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 59%|█████▉ | 116/195 [00:20<00:12, 6.58it/s] [2024-06-02 04:15:53] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.8.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 59%|█████▉ | 116/195 [00:21<00:12, 6.58it/s] [2024-06-02 04:15:53] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.8.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 59%|█████▉ | 116/195 [00:21<00:12, 6.58it/s] [2024-06-02 04:15:53] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.9.input_layernorm.weight", shape: (4096,), dtype: float32 59%|█████▉ | 116/195 [00:21<00:12, 6.58it/s] [2024-06-02 04:15:53] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.9.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 59%|█████▉ | 116/195 [00:21<00:12, 6.58it/s] [2024-06-02 04:15:53] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.9.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 59%|█████▉ | 116/195 [00:21<00:12, 6.58it/s] 62%|██████▏ | 120/195 [00:21<00:09, 8.01it/s] [2024-06-02 04:15:53] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.9.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 62%|██████▏ | 120/195 [00:21<00:09, 8.01it/s] [2024-06-02 04:15:54] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.9.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 62%|██████▏ | 120/195 [00:21<00:09, 8.01it/s] 62%|██████▏ | 121/195 [00:21<00:12, 6.09it/s] [2024-06-02 04:15:54] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.9.post_attention_layernorm.weight", shape: (4096,), dtype: float32 62%|██████▏ | 121/195 [00:21<00:12, 6.09it/s] [2024-06-02 04:15:54] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.9.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 62%|██████▏ | 121/195 [00:21<00:12, 6.09it/s] [2024-06-02 04:15:54] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.9.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 62%|██████▏ | 121/195 [00:21<00:12, 6.09it/s] [2024-06-02 04:15:54] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.9.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 62%|██████▏ | 121/195 [00:21<00:12, 6.09it/s] [2024-06-02 04:15:54] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.9.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 62%|██████▏ | 121/195 [00:21<00:12, 6.09it/s] 64%|██████▎ | 124/195 [00:21<00:08, 8.22it/s] [2024-06-02 04:15:54] INFO huggingface_loader.py:196: Unloading HF weight file: /ssd1/rickzhou/models/Mistral-7B-Instruct-v0.3/model-00001-of-00003.safetensors 64%|██████▎ | 124/195 [00:21<00:08, 8.22it/s] [2024-06-02 04:15:54] INFO huggingface_loader.py:184: Loading HF parameters from: /ssd1/rickzhou/models/Mistral-7B-Instruct-v0.3/model-00002-of-00003.safetensors 64%|██████▎ | 124/195 [00:22<00:08, 8.22it/s] [2024-06-02 04:15:55] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.10.input_layernorm.weight", shape: (4096,), dtype: float32 64%|██████▎ | 124/195 [00:22<00:08, 8.22it/s] [2024-06-02 04:15:55] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.10.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 64%|██████▎ | 124/195 [00:22<00:08, 8.22it/s] [2024-06-02 04:15:55] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.10.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 64%|██████▎ | 124/195 [00:22<00:08, 8.22it/s] 65%|██████▍ | 126/195 [00:22<00:16, 4.24it/s] [2024-06-02 04:15:55] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.10.post_attention_layernorm.weight", shape: (4096,), dtype: float32 65%|██████▍ | 126/195 [00:22<00:16, 4.24it/s] [2024-06-02 04:15:55] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.11.input_layernorm.weight", shape: (4096,), dtype: float32 65%|██████▍ | 126/195 [00:23<00:16, 4.24it/s] [2024-06-02 04:15:55] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.11.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 65%|██████▍ | 126/195 [00:23<00:16, 4.24it/s] [2024-06-02 04:15:55] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.11.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 65%|██████▍ | 126/195 [00:23<00:16, 4.24it/s] 66%|██████▌ | 129/195 [00:23<00:11, 5.79it/s] [2024-06-02 04:15:55] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.11.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 66%|██████▌ | 129/195 [00:23<00:11, 5.79it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.11.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 66%|██████▌ | 129/195 [00:23<00:11, 5.79it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.11.post_attention_layernorm.weight", shape: (4096,), dtype: float32 66%|██████▌ | 129/195 [00:23<00:11, 5.79it/s] 67%|██████▋ | 131/195 [00:23<00:11, 5.59it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.11.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 67%|██████▋ | 131/195 [00:23<00:11, 5.59it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.11.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 67%|██████▋ | 131/195 [00:23<00:11, 5.59it/s] 68%|██████▊ | 132/195 [00:23<00:10, 5.90it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.11.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 68%|██████▊ | 132/195 [00:23<00:10, 5.90it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.11.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 68%|██████▊ | 132/195 [00:23<00:10, 5.90it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.12.input_layernorm.weight", shape: (4096,), dtype: float32 68%|██████▊ | 132/195 [00:23<00:10, 5.90it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.12.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 68%|██████▊ | 132/195 [00:23<00:10, 5.90it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.12.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 68%|██████▊ | 132/195 [00:23<00:10, 5.90it/s] 69%|██████▉ | 135/195 [00:23<00:07, 7.73it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.12.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 69%|██████▉ | 135/195 [00:24<00:07, 7.73it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.12.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 69%|██████▉ | 135/195 [00:24<00:07, 7.73it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.12.post_attention_layernorm.weight", shape: (4096,), dtype: float32 69%|██████▉ | 135/195 [00:24<00:07, 7.73it/s] 70%|███████ | 137/195 [00:24<00:09, 6.16it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.12.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 70%|███████ | 137/195 [00:24<00:09, 6.16it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.12.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 70%|███████ | 137/195 [00:24<00:09, 6.16it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.12.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 70%|███████ | 137/195 [00:24<00:09, 6.16it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.12.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 70%|███████ | 137/195 [00:24<00:09, 6.16it/s] [2024-06-02 04:15:56] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.13.input_layernorm.weight", shape: (4096,), dtype: float32 70%|███████ | 137/195 [00:24<00:09, 6.16it/s] [2024-06-02 04:15:57] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.13.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 70%|███████ | 137/195 [00:24<00:09, 6.16it/s] [2024-06-02 04:15:57] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.13.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 70%|███████ | 137/195 [00:24<00:09, 6.16it/s] 72%|███████▏ | 141/195 [00:24<00:07, 7.68it/s] [2024-06-02 04:15:57] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.13.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 72%|███████▏ | 141/195 [00:24<00:07, 7.68it/s] [2024-06-02 04:15:57] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.13.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 72%|███████▏ | 141/195 [00:25<00:07, 7.68it/s] 73%|███████▎ | 142/195 [00:25<00:08, 6.03it/s] [2024-06-02 04:15:57] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.13.post_attention_layernorm.weight", shape: (4096,), dtype: float32 73%|███████▎ | 142/195 [00:25<00:08, 6.03it/s] [2024-06-02 04:15:57] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.13.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 73%|███████▎ | 142/195 [00:25<00:08, 6.03it/s] [2024-06-02 04:15:57] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.13.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 73%|███████▎ | 142/195 [00:25<00:08, 6.03it/s] [2024-06-02 04:15:57] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.13.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 73%|███████▎ | 142/195 [00:25<00:08, 6.03it/s] [2024-06-02 04:15:57] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.13.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 73%|███████▎ | 142/195 [00:25<00:08, 6.03it/s] 74%|███████▍ | 145/195 [00:25<00:06, 8.14it/s] [2024-06-02 04:15:57] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.14.input_layernorm.weight", shape: (4096,), dtype: float32 74%|███████▍ | 145/195 [00:25<00:06, 8.14it/s] [2024-06-02 04:15:57] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.14.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 74%|███████▍ | 145/195 [00:25<00:06, 8.14it/s] [2024-06-02 04:15:57] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.14.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 74%|███████▍ | 145/195 [00:25<00:06, 8.14it/s] 75%|███████▌ | 147/195 [00:25<00:05, 8.72it/s] [2024-06-02 04:15:58] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.14.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 75%|███████▌ | 147/195 [00:25<00:05, 8.72it/s] [2024-06-02 04:15:58] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.14.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 75%|███████▌ | 147/195 [00:25<00:05, 8.72it/s] [2024-06-02 04:15:58] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.14.post_attention_layernorm.weight", shape: (4096,), dtype: float32 75%|███████▌ | 147/195 [00:25<00:05, 8.72it/s] 76%|███████▋ | 149/195 [00:25<00:06, 7.15it/s] [2024-06-02 04:15:58] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.14.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 76%|███████▋ | 149/195 [00:25<00:06, 7.15it/s] [2024-06-02 04:15:58] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.14.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 76%|███████▋ | 149/195 [00:25<00:06, 7.15it/s] 77%|███████▋ | 150/195 [00:26<00:06, 6.93it/s] [2024-06-02 04:15:58] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.14.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 77%|███████▋ | 150/195 [00:26<00:06, 6.93it/s] [2024-06-02 04:15:58] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.14.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 77%|███████▋ | 150/195 [00:26<00:06, 6.93it/s] [2024-06-02 04:15:58] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.15.input_layernorm.weight", shape: (4096,), dtype: float32 77%|███████▋ | 150/195 [00:26<00:06, 6.93it/s] [2024-06-02 04:15:58] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.15.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 77%|███████▋ | 150/195 [00:26<00:06, 6.93it/s] [2024-06-02 04:15:58] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.15.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 77%|███████▋ | 150/195 [00:26<00:06, 6.93it/s] 78%|███████▊ | 153/195 [00:26<00:04, 8.92it/s] [2024-06-02 04:15:58] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.15.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 78%|███████▊ | 153/195 [00:26<00:04, 8.92it/s] [2024-06-02 04:15:59] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.15.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 78%|███████▊ | 153/195 [00:26<00:04, 8.92it/s] [2024-06-02 04:15:59] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.15.post_attention_layernorm.weight", shape: (4096,), dtype: float32 78%|███████▊ | 153/195 [00:26<00:04, 8.92it/s] 79%|███████▉ | 155/195 [00:26<00:05, 7.19it/s] [2024-06-02 04:15:59] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.15.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 79%|███████▉ | 155/195 [00:26<00:05, 7.19it/s] [2024-06-02 04:15:59] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.15.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 79%|███████▉ | 155/195 [00:26<00:05, 7.19it/s] 80%|████████ | 156/195 [00:26<00:05, 7.30it/s] [2024-06-02 04:15:59] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.15.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 80%|████████ | 156/195 [00:26<00:05, 7.30it/s] [2024-06-02 04:15:59] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.15.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 80%|████████ | 156/195 [00:26<00:05, 7.30it/s] [2024-06-02 04:15:59] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.16.input_layernorm.weight", shape: (4096,), dtype: float32 80%|████████ | 156/195 [00:26<00:05, 7.30it/s] [2024-06-02 04:15:59] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.16.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 80%|████████ | 156/195 [00:26<00:05, 7.30it/s] [2024-06-02 04:15:59] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.16.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 80%|████████ | 156/195 [00:26<00:05, 7.30it/s] 82%|████████▏ | 159/195 [00:26<00:03, 9.34it/s] [2024-06-02 04:15:59] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.16.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 82%|████████▏ | 159/195 [00:27<00:03, 9.34it/s] [2024-06-02 04:15:59] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.16.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 82%|████████▏ | 159/195 [00:27<00:03, 9.34it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.16.post_attention_layernorm.weight", shape: (4096,), dtype: float32 82%|████████▏ | 159/195 [00:27<00:03, 9.34it/s] 83%|████████▎ | 161/195 [00:27<00:05, 6.71it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.16.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 83%|████████▎ | 161/195 [00:27<00:05, 6.71it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.16.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 83%|████████▎ | 161/195 [00:27<00:05, 6.71it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.16.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 83%|████████▎ | 161/195 [00:27<00:05, 6.71it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.16.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 83%|████████▎ | 161/195 [00:27<00:05, 6.71it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.17.input_layernorm.weight", shape: (4096,), dtype: float32 83%|████████▎ | 161/195 [00:27<00:05, 6.71it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.17.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 83%|████████▎ | 161/195 [00:27<00:05, 6.71it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.17.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 83%|████████▎ | 161/195 [00:27<00:05, 6.71it/s] 85%|████████▍ | 165/195 [00:27<00:03, 8.28it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.17.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 85%|████████▍ | 165/195 [00:27<00:03, 8.28it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.17.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 85%|████████▍ | 165/195 [00:28<00:03, 8.28it/s] 85%|████████▌ | 166/195 [00:28<00:04, 6.32it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.17.post_attention_layernorm.weight", shape: (4096,), dtype: float32 85%|████████▌ | 166/195 [00:28<00:04, 6.32it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.17.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 85%|████████▌ | 166/195 [00:28<00:04, 6.32it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.17.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 85%|████████▌ | 166/195 [00:28<00:04, 6.32it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.17.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 85%|████████▌ | 166/195 [00:28<00:04, 6.32it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.17.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 85%|████████▌ | 166/195 [00:28<00:04, 6.32it/s] 87%|████████▋ | 169/195 [00:28<00:03, 8.44it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.18.input_layernorm.weight", shape: (4096,), dtype: float32 87%|████████▋ | 169/195 [00:28<00:03, 8.44it/s] [2024-06-02 04:16:00] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.18.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 87%|████████▋ | 169/195 [00:28<00:03, 8.44it/s] [2024-06-02 04:16:01] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.18.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 87%|████████▋ | 169/195 [00:28<00:03, 8.44it/s] 88%|████████▊ | 171/195 [00:28<00:02, 9.01it/s] [2024-06-02 04:16:01] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.18.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 88%|████████▊ | 171/195 [00:28<00:02, 9.01it/s] [2024-06-02 04:16:01] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.18.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 88%|████████▊ | 171/195 [00:28<00:02, 9.01it/s] [2024-06-02 04:16:01] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.18.post_attention_layernorm.weight", shape: (4096,), dtype: float32 88%|████████▊ | 171/195 [00:28<00:02, 9.01it/s] 89%|████████▊ | 173/195 [00:28<00:03, 7.29it/s] [2024-06-02 04:16:01] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.18.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 89%|████████▊ | 173/195 [00:29<00:03, 7.29it/s] [2024-06-02 04:16:01] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.18.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 89%|████████▊ | 173/195 [00:29<00:03, 7.29it/s] 89%|████████▉ | 174/195 [00:29<00:03, 7.00it/s] [2024-06-02 04:16:01] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.18.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 89%|████████▉ | 174/195 [00:29<00:03, 7.00it/s] [2024-06-02 04:16:01] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.18.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 89%|████████▉ | 174/195 [00:29<00:03, 7.00it/s] [2024-06-02 04:16:01] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.19.input_layernorm.weight", shape: (4096,), dtype: float32 89%|████████▉ | 174/195 [00:29<00:03, 7.00it/s] [2024-06-02 04:16:01] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.19.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 89%|████████▉ | 174/195 [00:29<00:03, 7.00it/s] [2024-06-02 04:16:01] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.19.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 89%|████████▉ | 174/195 [00:29<00:03, 7.00it/s] 91%|█████████ | 177/195 [00:29<00:02, 8.96it/s] [2024-06-02 04:16:02] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.19.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 91%|█████████ | 177/195 [00:29<00:02, 8.96it/s] [2024-06-02 04:16:02] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.19.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 91%|█████████ | 177/195 [00:29<00:02, 8.96it/s] [2024-06-02 04:16:02] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.19.post_attention_layernorm.weight", shape: (4096,), dtype: float32 91%|█████████ | 177/195 [00:29<00:02, 8.96it/s] 92%|█████████▏| 179/195 [00:29<00:02, 7.18it/s] [2024-06-02 04:16:02] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.19.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 92%|█████████▏| 179/195 [00:29<00:02, 7.18it/s] [2024-06-02 04:16:02] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.19.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 92%|█████████▏| 179/195 [00:29<00:02, 7.18it/s] 92%|█████████▏| 180/195 [00:29<00:02, 7.29it/s] [2024-06-02 04:16:02] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.19.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 92%|█████████▏| 180/195 [00:29<00:02, 7.29it/s] [2024-06-02 04:16:02] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.19.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 92%|█████████▏| 180/195 [00:29<00:02, 7.29it/s] [2024-06-02 04:16:02] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.20.input_layernorm.weight", shape: (4096,), dtype: float32 92%|█████████▏| 180/195 [00:29<00:02, 7.29it/s] [2024-06-02 04:16:02] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.20.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 92%|█████████▏| 180/195 [00:29<00:02, 7.29it/s] [2024-06-02 04:16:02] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.20.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 92%|█████████▏| 180/195 [00:30<00:02, 7.29it/s] 94%|█████████▍| 183/195 [00:30<00:01, 9.33it/s] [2024-06-02 04:16:02] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.20.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 94%|█████████▍| 183/195 [00:30<00:01, 9.33it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.20.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 94%|█████████▍| 183/195 [00:30<00:01, 9.33it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.20.post_attention_layernorm.weight", shape: (4096,), dtype: float32 94%|█████████▍| 183/195 [00:30<00:01, 9.33it/s] 95%|█████████▍| 185/195 [00:30<00:01, 6.73it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.20.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 95%|█████████▍| 185/195 [00:30<00:01, 6.73it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.20.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 95%|█████████▍| 185/195 [00:30<00:01, 6.73it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.20.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 95%|█████████▍| 185/195 [00:30<00:01, 6.73it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.20.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 95%|█████████▍| 185/195 [00:30<00:01, 6.73it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.21.input_layernorm.weight", shape: (4096,), dtype: float32 95%|█████████▍| 185/195 [00:30<00:01, 6.73it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.21.mlp.down_proj.q_weight", shape: (4096, 1792), dtype: uint32 95%|█████████▍| 185/195 [00:30<00:01, 6.73it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.21.mlp.down_proj.q_scale", shape: (4096, 448), dtype: float32 95%|█████████▍| 185/195 [00:30<00:01, 6.73it/s] 97%|█████████▋| 189/195 [00:30<00:00, 8.29it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.21.mlp.gate_up_proj.q_weight", shape: (28672, 512), dtype: uint32 97%|█████████▋| 189/195 [00:31<00:00, 8.29it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.21.mlp.gate_up_proj.q_scale", shape: (28672, 128), dtype: float32 97%|█████████▋| 189/195 [00:31<00:00, 8.29it/s] 97%|█████████▋| 190/195 [00:31<00:00, 6.26it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:174: [Not quantized] Parameter: "model.layers.21.post_attention_layernorm.weight", shape: (4096,), dtype: float32 97%|█████████▋| 190/195 [00:31<00:00, 6.26it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.21.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 97%|█████████▋| 190/195 [00:31<00:00, 6.26it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.21.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 97%|█████████▋| 190/195 [00:31<00:00, 6.26it/s] [2024-06-02 04:16:03] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.21.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 97%|█████████▋| 190/195 [00:31<00:00, 6.26it/s] [2024-06-02 04:16:04] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.21.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 97%|█████████▋| 190/195 [00:31<00:00, 6.26it/s] 99%|█████████▉| 193/195 [00:31<00:00, 8.38it/s] [2024-06-02 04:16:04] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.22.self_attn.qkv_proj.q_weight", shape: (6144, 512), dtype: uint32 99%|█████████▉| 193/195 [00:31<00:00, 8.38it/s] [2024-06-02 04:16:04] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.22.self_attn.qkv_proj.q_scale", shape: (6144, 128), dtype: float32 99%|█████████▉| 193/195 [00:31<00:00, 8.38it/s] [2024-06-02 04:16:04] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.22.self_attn.o_proj.q_weight", shape: (4096, 512), dtype: uint32 99%|█████████▉| 193/195 [00:31<00:00, 8.38it/s] [2024-06-02 04:16:04] INFO huggingface_loader.py:166: [Quantized] Parameter: "model.layers.22.self_attn.o_proj.q_scale", shape: (4096, 128), dtype: float32 99%|█████████▉| 193/195 [00:31<00:00, 8.38it/s] 100%|██████████| 195/195 [00:31<00:00, 6.18it/s] [2024-06-02 04:16:04] INFO huggingface_loader.py:196: Unloading HF weight file: /ssd1/rickzhou/models/Mistral-7B-Instruct-v0.3/model-00002-of-00003.safetensors [2024-06-02 04:16:04] INFO stats.py:76: Time usage: HF loading: 3.729 sec; Pre-quantization mapping: 5.336 sec; Quantization: 2.465 sec [2024-06-02 04:16:04] INFO stats.py:90: RAM usage: Peak RAM: 9.313 GB. Total bytes loaded from disk: 27.001 GB [2024-06-02 04:16:04] INFO convert_weight.py:155: Parameter size after quantization: 4.220 GB [2024-06-02 04:16:04] INFO convert_weight.py:160: Total parameters: 7,248,023,552 [2024-06-02 04:16:04] INFO convert_weight.py:161: Bits per parameter: 5.001 [2024-06-02 04:16:04] INFO convert_weight.py:166: Saved to directory: /ssd2/models/mlc-delivery/hf/mlc-ai/Mistral-7B-Instruct-v0.3-q4f32_1-MLC All finished, 106 total shards committed, record saved to /ssd2/models/mlc-delivery/hf/mlc-ai/Mistral-7B-Instruct-v0.3-q4f32_1-MLC/ndarray-cache.json Also saved a bf16 record to /ssd2/models/mlc-delivery/hf/mlc-ai/Mistral-7B-Instruct-v0.3-q4f32_1-MLC/ndarray-cache-b16.json