p04798526
/
LLaMA-Factory-Mirror
1
0
Fork 1
Code Issues Pull Requests Packages Projects Releases Wiki Activity

change: change for step500 run

This commit is contained in:
wql 2024-08-24 00:00:56 +08:00
parent d5b1161eff
commit bffeeddf29
25 changed files with 98 additions and 96 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

98
batch_run.sh
View File

@ -1,56 +1,58 @@
FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Llama2-7B/llama2_lora_sft_1.yaml | tee results/lora_sft_2/Llama2-7B/llama2_lora_sft_1_log.txt FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Llama2-7B/llama2_lora_sft_1.yaml | tee results/lora_sft_2/Llama2-7B/llama2_lora_sft_1_step500_log.txt
FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Llama2-7B/llama2_lora_sft_2.yaml | tee results/lora_sft_2/Llama2-7B/llama2_lora_sft_2_log.txt FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Llama2-7B/llama2_lora_sft_2.yaml | tee results/lora_sft_2/Llama2-7B/llama2_lora_sft_2_step500_log.txt
FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Llama2-7B/llama2_lora_sft_3.yaml | tee results/lora_sft_2/Llama2-7B/llama2_lora_sft_3_log.txt FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Llama2-7B/llama2_lora_sft_3.yaml | tee results/lora_sft_2/Llama2-7B/llama2_lora_sft_3_step500_log.txt
CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Llama2-7B/llama2_lora_sft_1_single.yaml | tee results/lora_sft_2/Llama2-7B/llama2_lora_sft_1_single_log.txt CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Llama2-7B/llama2_lora_sft_1_single.yaml | tee results/lora_sft_2/Llama2-7B/llama2_lora_sft_1_single_step500_log.txt
CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Llama2-7B/llama2_lora_sft_2_single.yaml | tee results/lora_sft_2/Llama2-7B/llama2_lora_sft_2_single_log.txt CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Llama2-7B/llama2_lora_sft_2_single.yaml | tee results/lora_sft_2/Llama2-7B/llama2_lora_sft_2_single_step500_log.txt
CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Llama2-7B/llama2_lora_sft_3_single.yaml | tee results/lora_sft_2/Llama2-7B/llama2_lora_sft_3_single_log.txt CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Llama2-7B/llama2_lora_sft_3_single.yaml | tee results/lora_sft_2/Llama2-7B/llama2_lora_sft_3_single_step500_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Llama2-7B/llama2_predict_1.yaml | tee results/inference/Llama2-7B/llama2_predict_1_log.txt FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1.yaml | tee results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1_step500_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Llama2-7B/llama2_predict_2.yaml | tee results/inference/Llama2-7B/llama2_predict_2_log.txt FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_2.yaml | tee results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_2_step500_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Llama2-7B/llama2_predict_3.yaml | tee results/inference/Llama2-7B/llama2_predict_3_log.txt FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_3.yaml | tee results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_3_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Llama2-7B/llama2_predict_1_single.yaml | tee results/inference/Llama2-7B/llama2_predict_1_single_log.txt CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1_single.yaml | tee results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1_single_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Llama2-7B/llama2_predict_2_single.yaml | tee results/inference/Llama2-7B/llama2_predict_2_single_log.txt CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_2_single.yaml | tee results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_2_single_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Llama2-7B/llama2_predict_3_single.yaml | tee results/inference/Llama2-7B/llama2_predict_3_single_log.txt CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_3_single.yaml | tee results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_3_single_step500_log.txt
FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1.yaml | tee results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1_log.txt FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_1.yaml | tee results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_1_step500_log.txt
FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_2.yaml | tee results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_2_log.txt FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_2.yaml | tee results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_2_step500_log.txt
FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_3.yaml | tee results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_3_log.txt FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_3.yaml | tee results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_3_step500_log.txt
CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1_single.yaml | tee results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1_single_log.txt CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_1_single.yaml | tee results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_1_single_step500_log.txt
CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_2_single.yaml | tee results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_2_single_log.txt CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_2_single.yaml | tee results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_2_single_step500_log.txt
CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_3_single.yaml | tee results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_3_single_log.txt CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_3_single.yaml | tee results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_3_single_step500_log.txt
FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_1.yaml | tee results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_1_log.txt FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Qwen-7B/Qwen_lora_sft_1.yaml | tee results/lora_sft_2/Qwen-7B/Qwen_lora_sft_1_step500_log.txt
FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_2.yaml | tee results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_2_log.txt FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Qwen-7B/Qwen_lora_sft_2.yaml | tee results/lora_sft_2/Qwen-7B/Qwen_lora_sft_2_step500_log.txt
FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_3.yaml | tee results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_3_log.txt FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Qwen-7B/Qwen_lora_sft_3.yaml | tee results/lora_sft_2/Qwen-7B/Qwen_lora_sft_3_step500_log.txt
CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_1_single.yaml | tee results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_1_single_log.txt CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Qwen-7B/Qwen_lora_sft_1_single.yaml | tee results/lora_sft_2/Qwen-7B/Qwen_lora_sft_1_single_step500_log.txt
CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_2_single.yaml | tee results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_2_single_log.txt CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Qwen-7B/Qwen_lora_sft_2_single.yaml | tee results/lora_sft_2/Qwen-7B/Qwen_lora_sft_2_single_step500_log.txt
CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_3_single.yaml | tee results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_3_single_log.txt CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Qwen-7B/Qwen_lora_sft_3_single.yaml | tee results/lora_sft_2/Qwen-7B/Qwen_lora_sft_3_single_step500_log.txt
FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Qwen-7B/Qwen_lora_sft_1.yaml | tee results/lora_sft_2/Qwen-7B/Qwen_lora_sft_1_log.txt
FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Qwen-7B/Qwen_lora_sft_2.yaml | tee results/lora_sft_2/Qwen-7B/Qwen_lora_sft_2_log.txt
FORCE_TORCHRUN=1 llamafactory-cli train results/lora_sft_2/Qwen-7B/Qwen_lora_sft_3.yaml | tee results/lora_sft_2/Qwen-7B/Qwen_lora_sft_3_log.txt
CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Qwen-7B/Qwen_lora_sft_1_single.yaml | tee results/lora_sft_2/Qwen-7B/Qwen_lora_sft_1_single_log.txt
CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Qwen-7B/Qwen_lora_sft_2_single.yaml | tee results/lora_sft_2/Qwen-7B/Qwen_lora_sft_2_single_log.txt
CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/lora_sft_2/Qwen-7B/Qwen_lora_sft_3_single.yaml | tee results/lora_sft_2/Qwen-7B/Qwen_lora_sft_3_single_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Qwen-7B/Qwen_predict_1.yaml | tee results/inference/Qwen-7B/Qwen_predict_1_log.txt # FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Llama2-7B/llama2_predict_1.yaml | tee results/inference/Llama2-7B/llama2_predict_1_step500_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Qwen-7B/Qwen_predict_2.yaml | tee results/inference/Qwen-7B/Qwen_predict_2_log.txt # FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Llama2-7B/llama2_predict_2.yaml | tee results/inference/Llama2-7B/llama2_predict_2_step500_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Qwen-7B/Qwen_predict_3.yaml | tee results/inference/Qwen-7B/Qwen_predict_3_log.txt # FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Llama2-7B/llama2_predict_3.yaml | tee results/inference/Llama2-7B/llama2_predict_3_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Qwen-7B/Qwen_predict_1_single.yaml | tee results/inference/Qwen-7B/Qwen_predict_1_single_log.txt # CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Llama2-7B/llama2_predict_1_single.yaml | tee results/inference/Llama2-7B/llama2_predict_1_single_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Qwen-7B/Qwen_predict_2_single.yaml | tee results/inference/Qwen-7B/Qwen_predict_2_single_log.txt # CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Llama2-7B/llama2_predict_2_single.yaml | tee results/inference/Llama2-7B/llama2_predict_2_single_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Qwen-7B/Qwen_predict_3_single.yaml | tee results/inference/Qwen-7B/Qwen_predict_3_single_log.txt # CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Llama2-7B/llama2_predict_3_single.yaml | tee results/inference/Llama2-7B/llama2_predict_3_single_step500_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/ChatGLM2-6B/ChatGLM2_predict_1.yaml | tee results/inference/ChatGLM2-6B/ChatGLM2_predict_1_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/ChatGLM2-6B/ChatGLM2_predict_2.yaml | tee results/inference/ChatGLM2-6B/ChatGLM2_predict_2_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/ChatGLM2-6B/ChatGLM2_predict_3.yaml | tee results/inference/ChatGLM2-6B/ChatGLM2_predict_3_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/ChatGLM2-6B/ChatGLM2_predict_1_single.yaml | tee results/inference/ChatGLM2-6B/ChatGLM2_predict_1_single_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/ChatGLM2-6B/ChatGLM2_predict_2_single.yaml | tee results/inference/ChatGLM2-6B/ChatGLM2_predict_2_single_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/ChatGLM2-6B/ChatGLM2_predict_3_single.yaml | tee results/inference/ChatGLM2-6B/ChatGLM2_predict_3_single_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Baichuan2-7B/Baichuan2_predict_1.yaml | tee results/inference/Baichuan2-7B/Baichuan2_predict_1_log.txt # FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Qwen-7B/Qwen_predict_1.yaml | tee results/inference/Qwen-7B/Qwen_predict_1_step500_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Baichuan2-7B/Baichuan2_predict_2.yaml | tee results/inference/Baichuan2-7B/Baichuan2_predict_2_log.txt # FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Qwen-7B/Qwen_predict_2.yaml | tee results/inference/Qwen-7B/Qwen_predict_2_step500_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Baichuan2-7B/Baichuan2_predict_3.yaml | tee results/inference/Baichuan2-7B/Baichuan2_predict_3_log.txt # FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Qwen-7B/Qwen_predict_3.yaml | tee results/inference/Qwen-7B/Qwen_predict_3_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Baichuan2-7B/Baichuan2_predict_1_single.yaml | tee results/inference/Baichuan2-7B/Baichuan2_predict_1_single_log.txt # CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Qwen-7B/Qwen_predict_1_single.yaml | tee results/inference/Qwen-7B/Qwen_predict_1_single_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Baichuan2-7B/Baichuan2_predict_2_single.yaml | tee results/inference/Baichuan2-7B/Baichuan2_predict_2_single_log.txt # CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Qwen-7B/Qwen_predict_2_single.yaml | tee results/inference/Qwen-7B/Qwen_predict_2_single_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Baichuan2-7B/Baichuan2_predict_3_single.yaml | tee results/inference/Baichuan2-7B/Baichuan2_predict_3_single_log.txt # CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Qwen-7B/Qwen_predict_3_single.yaml | tee results/inference/Qwen-7B/Qwen_predict_3_single_step500_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/ChatGLM2-6B/ChatGLM2_predict_1.yaml | tee results/inference/ChatGLM2-6B/ChatGLM2_predict_1_step500_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/ChatGLM2-6B/ChatGLM2_predict_2.yaml | tee results/inference/ChatGLM2-6B/ChatGLM2_predict_2_step500_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/ChatGLM2-6B/ChatGLM2_predict_3.yaml | tee results/inference/ChatGLM2-6B/ChatGLM2_predict_3_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/ChatGLM2-6B/ChatGLM2_predict_1_single.yaml | tee results/inference/ChatGLM2-6B/ChatGLM2_predict_1_single_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/ChatGLM2-6B/ChatGLM2_predict_2_single.yaml | tee results/inference/ChatGLM2-6B/ChatGLM2_predict_2_single_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/ChatGLM2-6B/ChatGLM2_predict_3_single.yaml | tee results/inference/ChatGLM2-6B/ChatGLM2_predict_3_single_step500_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Baichuan2-7B/Baichuan2_predict_1.yaml | tee results/inference/Baichuan2-7B/Baichuan2_predict_1_step500_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Baichuan2-7B/Baichuan2_predict_2.yaml | tee results/inference/Baichuan2-7B/Baichuan2_predict_2_step500_log.txt
# FORCE_TORCHRUN=1 llamafactory-cli train results/inference/Baichuan2-7B/Baichuan2_predict_3.yaml | tee results/inference/Baichuan2-7B/Baichuan2_predict_3_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Baichuan2-7B/Baichuan2_predict_1_single.yaml | tee results/inference/Baichuan2-7B/Baichuan2_predict_1_single_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Baichuan2-7B/Baichuan2_predict_2_single.yaml | tee results/inference/Baichuan2-7B/Baichuan2_predict_2_single_step500_log.txt
# CUDA_VISIBLE_DEVICES=0 llamafactory-cli train results/inference/Baichuan2-7B/Baichuan2_predict_3_single.yaml | tee results/inference/Baichuan2-7B/Baichuan2_predict_3_single_step500_log.txt

4
results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1 output_dir: ./results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1_single.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1_single output_dir: ./results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_1_single_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_2.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_2 output_dir: ./results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_2_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_2_single.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_2_single output_dir: ./results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_2_single_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_3.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_3 output_dir: ./results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_3_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_3_single.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_3_single output_dir: ./results/lora_sft_2/Baichuan2-7B/Baichuan2_lora_sft_3_single_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_1.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_1 output_dir: ./results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_1_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_1_single.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_1_single output_dir: ./results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_1_single_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_2.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_2 output_dir: ./results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_2_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_2_single.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_2_single output_dir: ./results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_2_single_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_3.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_3 output_dir: ./results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_3_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_3_single.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_3_single output_dir: ./results/lora_sft_2/ChatGLM2-6B/ChatGLM2_lora_sft_3_single_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Llama2-7B/llama2_lora_sft_1.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Llama2-7B/llama2_lora_sft_1 output_dir: ./results/lora_sft_2/Llama2-7B/llama2_lora_sft_1_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Llama2-7B/llama2_lora_sft_1_single.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Llama2-7B/llama2_lora_sft_1_single output_dir: ./results/lora_sft_2/Llama2-7B/llama2_lora_sft_1_single_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Llama2-7B/llama2_lora_sft_2.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Llama2-7B/llama2_lora_sft_2 output_dir: ./results/lora_sft_2/Llama2-7B/llama2_lora_sft_2_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Llama2-7B/llama2_lora_sft_2_single.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Llama2-7B/llama2_lora_sft_2_single output_dir: ./results/lora_sft_2/Llama2-7B/llama2_lora_sft_2_single_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Llama2-7B/llama2_lora_sft_3.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Llama2-7B/llama2_lora_sft_3 output_dir: ./results/lora_sft_2/Llama2-7B/llama2_lora_sft_3_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Llama2-7B/llama2_lora_sft_3_single.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Llama2-7B/llama2_lora_sft_3_single output_dir: ./results/lora_sft_2/Llama2-7B/llama2_lora_sft_3_single_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Qwen-7B/Qwen_lora_sft_1.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Qwen-7B/Qwen_lora_sft_1 output_dir: ./results/lora_sft_2/Qwen-7B/Qwen_lora_sft_1_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Qwen-7B/Qwen_lora_sft_1_single.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Qwen-7B/Qwen_lora_sft_1_single output_dir: ./results/lora_sft_2/Qwen-7B/Qwen_lora_sft_1_single_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Qwen-7B/Qwen_lora_sft_2.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Qwen-7B/Qwen_lora_sft_2 output_dir: ./results/lora_sft_2/Qwen-7B/Qwen_lora_sft_2_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Qwen-7B/Qwen_lora_sft_2_single.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Qwen-7B/Qwen_lora_sft_2_single output_dir: ./results/lora_sft_2/Qwen-7B/Qwen_lora_sft_2_single_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Qwen-7B/Qwen_lora_sft_3.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Qwen-7B/Qwen_lora_sft_3 output_dir: ./results/lora_sft_2/Qwen-7B/Qwen_lora_sft_3_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true

4
results/lora_sft_2/Qwen-7B/Qwen_lora_sft_3_single.yaml
View File

@ -16,7 +16,7 @@ overwrite_cache: true
preprocessing_num_workers: 16 preprocessing_num_workers: 16
### output ### output
output_dir: ./results/lora_sft_2/Qwen-7B/Qwen_lora_sft_3_single output_dir: ./results/lora_sft_2/Qwen-7B/Qwen_lora_sft_3_single_step500
logging_steps: 3 logging_steps: 3
save_steps: 500 save_steps: 500
plot_loss: true plot_loss: true
@ -31,7 +31,7 @@ lr_scheduler_type: cosine
warmup_ratio: 0.1 warmup_ratio: 0.1
bf16: true bf16: true
ddp_timeout: 180000000 ddp_timeout: 180000000
max_steps: 1000 max_steps: 500
include_num_input_tokens_seen: true include_num_input_tokens_seen: true
include_tokens_per_second: true include_tokens_per_second: true