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

.

This commit is contained in:
hailin 2025-09-05 18:55:38 +08:00
parent b31ca389f2
commit a08adeb9e6
2 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
.deepspeed_env
View File

@ -1,7 +1,7 @@
WANDB_BASE_URL=https://wandb.szaiai.com WANDB_BASE_URL=https://wandb.szaiai.com
WANDB_API_KEY=local-701636f51b4741d3862007df5cf7f12cca53d8d1 WANDB_API_KEY=local-701636f51b4741d3862007df5cf7f12cca53d8d1
WANDB_PROJECT=ds-qwen3 WANDB_PROJECT=ds-qwen3
WANDB_GROUP=q3-32b-ds4-2025-09-04 WANDB_GROUP=q3-1.7b-ds4-2025-09-05
WANDB_RUN_ID=q3-32b-lr2e-5-train1 WANDB_RUN_ID=q3-31.7b-lr2e-5-train1
WANDB_RESUME=allow WANDB_RESUME=allow
export WANDB_DIR=/tmp/$USER/wandb export WANDB_DIR=/tmp/$USER/wandb

4
mm-zero3.sh
View File

@ -28,9 +28,9 @@ export OPENBLAS_NUM_THREADS=8
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/Qwen3-32B \ --model_name_or_path /home/test/Qwen3-1.7B \
--data_glob "/home/test/datasets/my_corpus/train.jsonl" \ --data_glob "/home/test/datasets/my_corpus/train.jsonl" \
--output_dir /home/test/checkpoints/q3-32b-ds4 \ --output_dir /home/test/checkpoints/q3-1.7b-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 \