This commit is contained in:
parent
7184cb67f3
commit
56a250bc66
|
|
@ -7,9 +7,9 @@ export PYTORCH_CUDA_ALLOC_CONF="max_split_size_mb:128,expandable_segments:True,g
|
||||||
deepspeed --hostfile hostfile \
|
deepspeed --hostfile hostfile \
|
||||||
--num_nodes 6 --num_gpus 4 \
|
--num_nodes 6 --num_gpus 4 \
|
||||||
/home/test/jd_train/train_sft_ds.py \
|
/home/test/jd_train/train_sft_ds.py \
|
||||||
--model_name_or_path /home/test/Qwen3-14B \
|
--model_name_or_path /home/test/Qwen3-8B \
|
||||||
--data_glob "/home/test/datasets/my_corpus/train.jsonl" \
|
--data_glob "/home/test/datasets/my_corpus/train.jsonl" \
|
||||||
--output_dir /home/test/checkpoints/q3-14b-ds4 \
|
--output_dir /home/test/checkpoints/q3-8b-ds4 \
|
||||||
--seq_len 512 \
|
--seq_len 512 \
|
||||||
--per_device_train_batch_size 1 \
|
--per_device_train_batch_size 1 \
|
||||||
--gradient_accumulation_steps 1 \
|
--gradient_accumulation_steps 1 \
|
||||||
|
|
|
||||||
Loading…
Reference in New Issue