21 lines
730 B
Bash
Executable File
21 lines
730 B
Bash
Executable File
FORCE_COLOR=1 deepspeed --hostfile hostfile \
|
|
--num_nodes 6 --num_gpus 4 \
|
|
train_sft_lora.py \
|
|
--model_name_or_path /home/test/Qwen3-32B \
|
|
--data_glob "/home/test/datasets/my_corpus/train*.jsonl" \
|
|
--output_dir /home/test/checkpoints/q3-32b-lora \
|
|
--seq_len 1024 \
|
|
--bf16 \
|
|
--per_device_train_batch_size 1 \
|
|
--gradient_accumulation_steps 4 \
|
|
--learning_rate 2e-4 \
|
|
--warmup_ratio 0.03 \
|
|
--lora_r 16 --lora_alpha 64 --lora_dropout 0.05 \
|
|
--lora_exclude lm_head \
|
|
--max_steps 3000 \
|
|
--log_interval 10 \
|
|
--eval_steps 50 \
|
|
--gradient_checkpointing \
|
|
--deepspeed /home/test/jd_train/ds_config_zero3_lora_gpu.json \
|
|
--report_to wandb --wandb_project ds-qwen3-lora
|