wehub-resource-sync
10 lines
273 B
Bash
10 lines
273 B
Bash
CUDA_VISIBLE_DEVICES=0,1 \
|
|
swift infer \
|
|
--model Qwen/Qwen3-8B \
|
|
--infer_backend sglang \
|
|
--val_dataset AI-ModelScope/alpaca-gpt4-data-zh#2000 \
|
|
--max_new_tokens 2048 \
|
|
--sglang_context_length 8192 \
|
|
--sglang_tp_size 2 \
|
|
--write_batch_size 1000
|