ESFT/scripts/eval.sh

13 lines
382 B
Bash
Raw Normal View History

# first, download adapter models and put them to the corresponding directories
python scripts/eval.py \
--eval_datasets=translation \
--base_model_path=deepseek-ai/ESFT-vanilla-lite \
--adapter_dir=all_models/adapters/token \
--output_dir=results/completions/token \
--max_new_tokens=512 \
--openai_api_key=REPLACE_WITH_YOUR_KEY \
--eval_batch_size=2