Create evals.sh

evals.sh

@@ -0,0 +1,6 @@
+accelerate launch --main_process_port 29501 -m lm_eval --model hf --model_args pretrained=../,trust_remote_code=True --tasks mmlu       --num_fewshot 5 --apply_chat_template --fewshot_as_multiturn --batch_size 1 > 1-mmlu.out 2>&1
+accelerate launch --main_process_port 29501 -m lm_eval --model hf --model_args pretrained=../,trust_remote_code=True --tasks hellaswag  --num_fewshot 5 --apply_chat_template --fewshot_as_multiturn --batch_size 1 > 2-hellaswag.out 2>&1
+accelerate launch --main_process_port 29501 -m lm_eval --model hf --model_args pretrained=../,trust_remote_code=True --tasks commonsense_qa       --num_fewshot 5 --apply_chat_template --fewshot_as_multiturn --batch_size 1 > 3-commonsenseqa.out 2>&1
+accelerate launch --main_process_port 29501 -m lm_eval --model hf --model_args pretrained=../,trust_remote_code=True --tasks gsm8k      --num_fewshot 5 --apply_chat_template --fewshot_as_multiturn --batch_size 1 > 4-gsm8k.out 2>&1
+accelerate launch --main_process_port 29501 -m lm_eval --model hf --model_args pretrained=../,trust_remote_code=True --tasks mbpp --batch_size 1 --apply_chat_template --fewshot_as_multiturn --confirm_run_unsafe_code > 5-mbpp.out 2>&1
+accelerate launch --main_process_port 29501 -m lm_eval --model hf --model_args pretrained=../,trust_remote_code=True --tasks longbench_multi_news --apply_chat_template  --batch_size 1 > 6-multinews.out 2>&1