File size: 331 Bytes
cb2428f |
1 2 3 4 5 6 7 8 9 |
# If it's full parameter training, use `--model xxx` instead of `--adapters xxx`.
# If you are using the validation set for inference, add the parameter `--load_data_args true`.
CUDA_VISIBLE_DEVICES=0 \
swift infer \
--adapters output/vx-xxx/checkpoint-xxx \
--stream true \
--temperature 0 \
--max_new_tokens 2048
|