Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

预测阶段:poetry run sh ./dbgpt_hub/scripts/predict_sft.sh,Killed #269

Open
GuokaiLiu opened this issue Jun 16, 2024 · 2 comments
Open

Comments

@GuokaiLiu
Copy link

GuokaiLiu commented Jun 16, 2024

  • 环境:

    • WLS-2,Ubuntu22.04, 4090 GPU x1
  • train_sft.sh

CUDA_VISIBLE_DEVICES=0 python dbgpt_hub/train/sft_train.py \
    --model_name_or_path $model_name_or_path \
    --quantization_bit 4 \
    --do_train \
    --dataset $dataset \
    --max_source_length 2048 \
    --max_target_length 512 \
    --finetuning_type lora \
    --lora_target q_proj,v_proj \
    --template llama2 \
    --lora_rank 32 \
    --lora_alpha 32 \
    --output_dir $output_dir \
    --overwrite_cache \
    --overwrite_output_dir \
    --per_device_train_batch_size 1 \
    --gradient_accumulation_steps 16 \
    --lr_scheduler_type cosine_with_restarts \
    --logging_steps 50 \
    --save_steps 2000 \
    --learning_rate 2e-4 \
    --num_train_epochs 8 \
    --plot_loss >> ${train_log}
    # \
    # --bf16  >> ${train_log}
    # --bf16#v100不支持bf16
  • predict_sft.sh
CUDA_VISIBLE_DEVICES=0  python dbgpt_hub/predict/predict.py \
    --model_name_or_path /home/lgk/Downloads/CodeLlama-7b-Instruct-hf \
    --template llama2 \
    --finetuning_type lora \
    --predicted_input_filename dbgpt_hub/data/example_text2sql_dev.json \
    --checkpoint_dir dbgpt_hub/output/adapter/CodeLlama-7b-sql-lora \
    --predicted_out_filename dbgpt_hub/output/pred/pred_codellama7b.sql >> ${pred_log}   
  • poetry run sh ./dbgpt_hub/scripts/predict_sft.sh报错Killed
(dbgpt_hub) lgk@WIN-20240401VAM:~/Projects/DB-GPT-Hub$ poetry run sh ./dbgpt_hub/scripts/predict_sft.sh
Warning: Found deprecated priority 'default' for source 'mirrors' in pyproject.toml. You can achieve the same effect by changing the priority to 'primary' and putting the source first.
/home/lgk/.conda/envs/dbgpt_hub/lib/python3.10/site-packages/transformers/deepspeed.py:23: FutureWarning: transformers.deepspeed module is deprecated and will be removed in a future version. Please import deepspeed modules directly from transformers.integrations
  warnings.warn(
Loading checkpoint shards:   0%|                                                                                     | 0/2 [00:00<?, ?it/s]

Killed
@xhh315
Copy link

xhh315 commented Oct 14, 2024

你好想问一下,你的loss是正常的吗,我的一开始就很小

@Oops322
Copy link

Oops322 commented Oct 15, 2024 via email

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants