-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathrun_7B_lora.sh
139 lines (123 loc) · 9.25 KB
/
run_7B_lora.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
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
WORKER_GPU=8
WORKER_NUM=1
# pip3 install k-means-constrained -i https://pypi.tuna.tsinghua.edu.cn/simple
torchrun --nproc_per_node $WORKER_GPU --master_addr localhost --node_rank 0 --master_port 12345 --nnodes $WORKER_NUM \
topk_train.py --model_name_or_path /home/hkustadmin/huangzeyu/Llama-2-7b-hf --data_path /home/hkustadmin/huangzeyu/alpaca --bf16 True \
--num_train_epochs 3 --per_device_train_batch_size 1 --per_device_eval_batch_size 1 --gradient_accumulation_steps 16 \
--evaluation_strategy "no" --save_strategy "steps" --save_steps 2000 --save_total_limit 1 \
--learning_rate 2e-5 --weight_decay 0. --warmup_ratio 0.03 \
--lr_scheduler_type "cosine" --logging_steps 2 --tf32 True --report_to tensorboard \
--base_config_dir /home/hkustadmin/huangzeyu/DynamicKTuning/7b_base_configs.json \
--output_dir /home/hkustadmin/huangzeyu/output_7B_DT \
--lora_train True \
--split_start_layer 16 --split_every_layer 2 \
--select 'inter_abs' --dynamic_mode 'l1' \
--threshold 0.1 \
torchrun --nproc_per_node $WORKER_GPU --master_addr localhost --node_rank 0 --master_port 12345 --nnodes $WORKER_NUM \
topk_train.py --model_name_or_path /home/hkustadmin/huangzeyu/Llama-2-7b-hf --data_path /home/hkustadmin/huangzeyu/alpaca --bf16 True \
--num_train_epochs 3 --per_device_train_batch_size 1 --per_device_eval_batch_size 1 --gradient_accumulation_steps 16 \
--evaluation_strategy "no" --save_strategy "steps" --save_steps 2000 --save_total_limit 1 \
--learning_rate 2e-5 --weight_decay 0. --warmup_ratio 0.03 \
--lr_scheduler_type "cosine" --logging_steps 2 --tf32 True --report_to tensorboard \
--base_config_dir /home/hkustadmin/huangzeyu/DynamicKTuning/7b_base_configs.json \
--output_dir /home/hkustadmin/huangzeyu/output_7B_DT \
--lora_train True \
--split_start_layer 16 --split_every_layer 2 \
--select 'inter_abs' --dynamic_mode 'softmax' \
--threshold 0.1 \
torchrun --nproc_per_node $WORKER_GPU --master_addr localhost --node_rank 0 --master_port 12345 --nnodes $WORKER_NUM \
topk_train.py --model_name_or_path /home/hkustadmin/huangzeyu/Llama-2-7b-hf --data_path /home/hkustadmin/huangzeyu/alpaca --bf16 True \
--num_train_epochs 3 --per_device_train_batch_size 1 --per_device_eval_batch_size 1 --gradient_accumulation_steps 16 \
--evaluation_strategy "no" --save_strategy "steps" --save_steps 2000 --save_total_limit 1 \
--learning_rate 2e-5 --weight_decay 0. --warmup_ratio 0.03 \
--lr_scheduler_type "cosine" --logging_steps 2 --tf32 True --report_to tensorboard \
--base_config_dir /home/hkustadmin/huangzeyu/DynamicKTuning/7b_base_configs.json \
--output_dir /home/hkustadmin/huangzeyu/output_7B_DT \
--lora_train True \
--split_start_layer 16 --split_every_layer 2 \
--select 'gate' --dynamic_mode 'l1' \
--threshold 0.1 \
torchrun --nproc_per_node $WORKER_GPU --master_addr localhost --node_rank 0 --master_port 12345 --nnodes $WORKER_NUM \
topk_train.py --model_name_or_path /home/hkustadmin/huangzeyu/Llama-2-7b-hf --data_path /home/hkustadmin/huangzeyu/alpaca --bf16 True \
--num_train_epochs 3 --per_device_train_batch_size 1 --per_device_eval_batch_size 1 --gradient_accumulation_steps 16 \
--evaluation_strategy "no" --save_strategy "steps" --save_steps 2000 --save_total_limit 1 \
--learning_rate 2e-5 --weight_decay 0. --warmup_ratio 0.03 \
--lr_scheduler_type "cosine" --logging_steps 2 --tf32 True --report_to tensorboard \
--base_config_dir /home/hkustadmin/huangzeyu/DynamicKTuning/7b_base_configs.json \
--output_dir /home/hkustadmin/huangzeyu/output_7B_DT \
--lora_train True \
--split_start_layer 16 --split_every_layer 2 \
--select 'gate' --dynamic_mode 'softmax' \
--threshold 0.1 \
torchrun --nproc_per_node $WORKER_GPU --master_addr localhost --node_rank 0 --master_port 12345 --nnodes $WORKER_NUM \
topk_train.py --model_name_or_path /home/hkustadmin/huangzeyu/Llama-2-7b-hf --data_path /home/hkustadmin/huangzeyu/alpaca --bf16 True \
--num_train_epochs 3 --per_device_train_batch_size 1 --per_device_eval_batch_size 1 --gradient_accumulation_steps 16 \
--evaluation_strategy "no" --save_strategy "steps" --save_steps 2000 --save_total_limit 1 \
--learning_rate 2e-5 --weight_decay 0. --warmup_ratio 0.03 \
--lr_scheduler_type "cosine" --logging_steps 2 --tf32 True --report_to tensorboard \
--base_config_dir /home/hkustadmin/huangzeyu/DynamicKTuning/7b_base_configs.json \
--output_dir /home/hkustadmin/huangzeyu/output_7B_DT \
--lora_train True \
--split_start_layer 16 --split_every_layer 1 \
--select 'gate' --dynamic_mode 'l1' \
--threshold 0.1 \
torchrun --nproc_per_node $WORKER_GPU --master_addr localhost --node_rank 0 --master_port 12345 --nnodes $WORKER_NUM \
topk_train.py --model_name_or_path /home/hkustadmin/huangzeyu/Llama-2-7b-hf --data_path /home/hkustadmin/huangzeyu/alpaca --bf16 True \
--num_train_epochs 3 --per_device_train_batch_size 1 --per_device_eval_batch_size 1 --gradient_accumulation_steps 16 \
--evaluation_strategy "no" --save_strategy "steps" --save_steps 2000 --save_total_limit 1 \
--learning_rate 2e-5 --weight_decay 0. --warmup_ratio 0.03 \
--lr_scheduler_type "cosine" --logging_steps 2 --tf32 True --report_to tensorboard \
--base_config_dir /home/hkustadmin/huangzeyu/DynamicKTuning/7b_base_configs.json \
--output_dir /home/hkustadmin/huangzeyu/output_7B_DT \
--lora_train True \
--split_start_layer 16 --split_every_layer 1 \
--select 'gate' --dynamic_mode 'softmax' \
--threshold 0.1 \
torchrun --nproc_per_node $WORKER_GPU --master_addr localhost --node_rank 0 --master_port 12345 --nnodes $WORKER_NUM \
topk_train.py --model_name_or_path /home/hkustadmin/huangzeyu/Llama-2-7b-hf --data_path /home/hkustadmin/huangzeyu/alpaca --bf16 True \
--num_train_epochs 3 --per_device_train_batch_size 1 --per_device_eval_batch_size 1 --gradient_accumulation_steps 16 \
--evaluation_strategy "no" --save_strategy "steps" --save_steps 2000 --save_total_limit 1 \
--learning_rate 2e-5 --weight_decay 0. --warmup_ratio 0.03 \
--lr_scheduler_type "cosine" --logging_steps 2 --tf32 True --report_to tensorboard \
--base_config_dir /home/hkustadmin/huangzeyu/DynamicKTuning/7b_base_configs.json \
--output_dir /home/hkustadmin/huangzeyu/output_7B_DT \
--lora_train True \
--split_start_layer 16 --split_every_layer 1 \
--select 'gate' --dynamic_mode 'l1' \
--threshold 0.05 \
torchrun --nproc_per_node $WORKER_GPU --master_addr localhost --node_rank 0 --master_port 12345 --nnodes $WORKER_NUM \
topk_train.py --model_name_or_path /home/hkustadmin/huangzeyu/Llama-2-7b-hf --data_path /home/hkustadmin/huangzeyu/alpaca --bf16 True \
--num_train_epochs 3 --per_device_train_batch_size 1 --per_device_eval_batch_size 1 --gradient_accumulation_steps 16 \
--evaluation_strategy "no" --save_strategy "steps" --save_steps 2000 --save_total_limit 1 \
--learning_rate 2e-5 --weight_decay 0. --warmup_ratio 0.03 \
--lr_scheduler_type "cosine" --logging_steps 2 --tf32 True --report_to tensorboard \
--base_config_dir /home/hkustadmin/huangzeyu/DynamicKTuning/7b_base_configs.json \
--output_dir /home/hkustadmin/huangzeyu/output_7B_DT \
--lora_train True \
--split_start_layer 16 --split_every_layer 1 \
--select 'inter_abs' --dynamic_mode 'l1' \
--threshold 0.05 \
torchrun --nproc_per_node $WORKER_GPU --master_addr localhost --node_rank 0 --master_port 12345 --nnodes $WORKER_NUM \
topk_train.py --model_name_or_path /home/hkustadmin/huangzeyu/Llama-2-7b-hf --data_path /home/hkustadmin/huangzeyu/alpaca --bf16 True \
--num_train_epochs 3 --per_device_train_batch_size 1 --per_device_eval_batch_size 1 --gradient_accumulation_steps 16 \
--evaluation_strategy "no" --save_strategy "steps" --save_steps 2000 --save_total_limit 1 \
--learning_rate 2e-5 --weight_decay 0. --warmup_ratio 0.03 \
--lr_scheduler_type "cosine" --logging_steps 2 --tf32 True --report_to tensorboard \
--base_config_dir /home/hkustadmin/huangzeyu/DynamicKTuning/7b_base_configs.json \
--output_dir /home/hkustadmin/huangzeyu/output_7B_DT \
--lora_train True \
--split_start_layer 16 --split_every_layer 1 \
--select 'gate' --dynamic_mode 'l1' \
--threshold 0.15 \
torchrun --nproc_per_node $WORKER_GPU --master_addr localhost --node_rank 0 --master_port 12345 --nnodes $WORKER_NUM \
topk_train.py --model_name_or_path /home/hkustadmin/huangzeyu/Llama-2-7b-hf --data_path /home/hkustadmin/huangzeyu/alpaca --bf16 True \
--num_train_epochs 3 --per_device_train_batch_size 1 --per_device_eval_batch_size 1 --gradient_accumulation_steps 16 \
--evaluation_strategy "no" --save_strategy "steps" --save_steps 2000 --save_total_limit 1 \
--learning_rate 2e-5 --weight_decay 0. --warmup_ratio 0.03 \
--lr_scheduler_type "cosine" --logging_steps 2 --tf32 True --report_to tensorboard \
--base_config_dir /home/hkustadmin/huangzeyu/DynamicKTuning/7b_base_configs.json \
--output_dir /home/hkustadmin/huangzeyu/output_7B_DT \
--lora_train True \
--split_start_layer 16 --split_every_layer 1 \
--select 'inter_abs' --dynamic_mode 'l1' \
--threshold 0.15 \