-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathgptq_deepseek_eval.sh
111 lines (81 loc) · 4.29 KB
/
gptq_deepseek_eval.sh
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
#!/bin/bash
export DEBUG=0
export PYTHONPATH=/home/LeiFeng/pingzhi/moe_quantize/optimum/:$PYTHONPATH:/home/LeiFeng/pingzhi/moe_quantize/auto_gptq/:$PYTHONPATH
export CUDA_VISIBLE_DEVICES=0
python quantize_gptq_deepseek_eval.py \
--model_name deepseek-ai/deepseek-moe-16b-chat \
--quant_model_path autogptq_quantize_model/deepseek-moe-16b-chat-gptq_w_bit_all_4 \
--bits all_4 \
--is_quantized
export DEBUG=0
export PYTHONPATH=/home/LeiFeng/pingzhi/moe_quantize/optimum/:$PYTHONPATH:/home/LeiFeng/pingzhi/moe_quantize/auto_gptq/:$PYTHONPATH
export CUDA_VISIBLE_DEVICES=1
python quantize_gptq_deepseek_eval.py \
--model_name deepseek-ai/deepseek-moe-16b-chat \
--quant_model_path autogptq_quantize_model/deepseek-moe-16b-chat-gptq_w_bit_all_2 \
--bits all_2 \
--is_quantized \
export DEBUG=0
export CUDA_VISIBLE_DEVICES=2
export PYTHONPATH=/home/LeiFeng/pingzhi/moe_quantize/optimum/:$PYTHONPATH:/home/LeiFeng/pingzhi/moe_quantize/auto_gptq/:$PYTHONPATH
python quantize_gptq_deepseek_eval.py \
--model_name deepseek-ai/deepseek-moe-16b-chat \
--quant_model_path autogptq_quantize_model/deepseek-moe-16b-chat-gptq_w_bit_moe.all_mlp.2+other_block.4 \
--bits moe.all_mlp.2+other_block.4 \
--is_quantized \
export DEBUG=0
export CUDA_VISIBLE_DEVICES=3
export PYTHONPATH=/home/LeiFeng/pingzhi/moe_quantize/optimum/:$PYTHONPATH:/home/LeiFeng/pingzhi/moe_quantize/auto_gptq/:$PYTHONPATH
python quantize_gptq_deepseek_eval.py \
--model_name deepseek-ai/deepseek-moe-16b-chat \
--quant_model_path autogptq_quantize_model/deepseek-moe-16b-chat-gptq_w_bit_moe.shared_4.other.2+other_block_4 \
--bits moe.shared_4.other.2+other_block_4 \
--is_quantized \
export DEBUG=0
export CUDA_VISIBLE_DEVICES=4
export PYTHONPATH=/home/LeiFeng/pingzhi/moe_quantize/optimum/:$PYTHONPATH:/home/LeiFeng/pingzhi/moe_quantize/auto_gptq/:$PYTHONPATH
python quantize_gptq_deepseek_eval.py \
--model_name deepseek-ai/deepseek-moe-16b-chat \
--quant_model_path autogptq_quantize_model/deepseek-moe-16b-chat-gptq_w_bit_moe.shared_2.other.4+other_block_4 \
--bits moe.shared_2.other.4+other_block_4 \
--is_quantized
export DEBUG=0
export CUDA_VISIBLE_DEVICES=5
export PYTHONPATH=/home/LeiFeng/pingzhi/moe_quantize/optimum/:$PYTHONPATH:/home/LeiFeng/pingzhi/moe_quantize/auto_gptq/:$PYTHONPATH
python quantize_gptq_deepseek_eval.py \
--model_name deepseek-ai/deepseek-moe-16b-chat \
--quant_model_path autogptq_quantize_model/deepseek-moe-16b-chat-gptq_w_bit_all_8 \
--bits all_8 \
--is_quantized
export DEBUG=0
export CUDA_VISIBLE_DEVICES=6
export PYTHONPATH=/home/LeiFeng/pingzhi/moe_quantize/optimum/:$PYTHONPATH:/home/LeiFeng/pingzhi/moe_quantize/auto_gptq/:$PYTHONPATH
python quantize_gptq_deepseek_eval.py \
--model_name deepseek-ai/deepseek-moe-16b-chat \
--quant_model_path autogptq_quantize_model/deepseek-moe-16b-chat-gptq_w_bit_moe.all_mlp.4+other_block.8 \
--bits moe.all_mlp.4+other_block.8 \
--is_quantized
export DEBUG=0
export CUDA_VISIBLE_DEVICES=7
export PYTHONPATH=/home/LeiFeng/pingzhi/moe_quantize/optimum/:$PYTHONPATH:/home/LeiFeng/pingzhi/moe_quantize/auto_gptq/:$PYTHONPATH
python quantize_gptq_deepseek_eval.py \
--model_name deepseek-ai/deepseek-moe-16b-chat \
--quant_model_path autogptq_quantize_model/deepseek-moe-16b-chat-gptq_w_bit_moe.shared_4.other.2+other_block.8 \
--bits moe.shared_4.other.2+other_block.8 \
--is_quantized
export DEBUG=0
export CUDA_VISIBLE_DEVICES=0
export PYTHONPATH=/home/LeiFeng/pingzhi/moe_quantize/optimum/:$PYTHONPATH:/home/LeiFeng/pingzhi/moe_quantize/auto_gptq/:$PYTHONPATH
python quantize_gptq_deepseek_eval.py \
--model_name deepseek-ai/deepseek-moe-16b-chat \
--quant_model_path autogptq_quantize_model/deepseek-moe-16b-chat-gptq_w_bit_moe.shared_2.other.4+other_block.8 \
--bits moe.shared_2.other.4+other_block.8 \
--is_quantized
export DEBUG=0
export CUDA_VISIBLE_DEVICES=1
export PYTHONPATH=/home/LeiFeng/pingzhi/moe_quantize/optimum/:$PYTHONPATH:/home/LeiFeng/pingzhi/moe_quantize/auto_gptq/:$PYTHONPATH
python quantize_gptq_deepseek_eval.py \
--model_name deepseek-ai/deepseek-moe-16b-chat \
--quant_model_path autogptq_quantize_model/deepseek-moe-16b-chat-gptq_w_bit_moe.shared_2.other.4+other_block.8 \
--bits moe.shared_2.other.4+other_block.8 \
--is_quantized