LLM_RECIPES_DIR=/code/llm-recipes
source $LLM_RECIPES_DIR/scripts/wmt2024/tokens.sh

python /code/llm-recipes/tools/hf_inference.py \
       --model /work/models/translation_finetuned_hf/mistral-llm-recipes-en-ja-continuous-pretrained-v1-dev-finetune-chunked-docs-all-averaged-71-75 \
       -i /work/wmt2024_test/LLM/wmttest2024.src.sentence_splited.with_template.en-ja.en.jsonl  \
       -o /work/translation/wmt24_test/en-ja/mistral-greedy \
       -g 0 \
       -b 4096 \
       --dynamic_max_new_token_ratio 3.0

echo "Done!"