hailin
/
jd_train
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

.

This commit is contained in:
hailin 2025-08-28 19:02:07 +08:00
parent 484f5d869c
commit 156805e04d
3 changed files with 6 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
mm-zero3.sh
View File

@ -1,9 +1,9 @@
deepspeed --hostfile hostfile \
--num_nodes 6 --num_gpus 4 \
/home/test/jd_train/train_sft_ds.py \
--model_name_or_path /home/test/Qwen3-1.7B \
--model_name_or_path /home/test/Qwen3-8B \
--data_glob "/home/test/datasets/my_corpus/train.jsonl" \
--output_dir /home/test/checkpoints/q3-1_7b-ds4 \
--output_dir /home/test/checkpoints/q3-8b-ds4 \
--seq_len 512 \
--per_device_train_batch_size 1 \
--gradient_accumulation_steps 1 \

4
ss-zero3.sh
View File

@ -3,9 +3,9 @@ export PYTORCH_CUDA_ALLOC_CONF="max_split_size_mb:128"
export NCCL_DEBUG=INFO
torchrun --nproc_per_node 4 /home/test/jd_train/train_sft_ds.py \
--model_name_or_path /home/test/Qwen3-1.7B \
--model_name_or_path /home/test/Qwen3-8B \
--data_glob "/home/test/datasets/my_corpus/train.jsonl" \
--output_dir /home/test/checkpoints/q3-1_7b-ds4 \
--output_dir /home/test/checkpoints/q3-8b-ds4 \
--seq_len 512 \
--per_device_train_batch_size 1 \
--gradient_accumulation_steps 1 \

4
ss.sh
View File

@ -4,9 +4,9 @@ export NCCL_DEBUG=INFO
export CUDA_VISIBLE_DEVICES=0
torchrun --nproc_per_node 1 /home/test/jd_train/train_sft_ds.py \
--model_name_or_path /home/test/Qwen3-1.7B \
--model_name_or_path /home/test/Qwen3-8B \
--data_glob "/home/test/datasets/my_corpus/train.jsonl" \
--output_dir /home/test/checkpoints/smoke-q3-1_7b-ds \
--output_dir /home/test/checkpoints/smoke-q3-8b-ds \
--seq_len 512 \
--per_device_train_batch_size 1 \
--gradient_accumulation_steps 1 \