Update evals/evals.sh

evals/evals.sh

@@ -1,6 +1,6 @@
-CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 accelerate launch --main_process_port 29500 -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 4 > 1-mmlu.out 2>&1
-CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 accelerate launch --main_process_port 29500 -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 4 > 2-hellaswag.out 2>&1
-CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 accelerate launch --main_process_port 29500 -m lm_eval --model hf --model_args pretrained=../,trust_remote_code=True --tasks winogrande --num_fewshot 5 --apply_chat_template --fewshot_as_multiturn --batch_size 8 > 3-winogrande.out 2>&1
-CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 accelerate launch --main_process_port 29500 -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 4 > 4-gsm8k.out 2>&1
-CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 accelerate launch --main_process_port 29500 -m lm_eval --model hf --model_args pretrained=../,trust_remote_code=True --tasks humaneval  --batch_size 4 --confirm_run_unsafe_code > 5-humaneval.out 2>&1
-CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 accelerate launch --main_process_port 29500 -m lm_eval --model hf --model_args pretrained=../,trust_remote_code=True --tasks coqa --batch_size 4 > 6-coqa.out 2>&1
+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 16 > 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 16 > 2-hellaswag.out 2>&1
+accelerate launch --main_process_port 29501 -m lm_eval --model hf --model_args pretrained=../,trust_remote_code=True --tasks winogrande --num_fewshot 5 --apply_chat_template --fewshot_as_multiturn --batch_size 16 > 3-winogrande.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 16 > 4-gsm8k-ori.out 2>&1
+accelerate launch --main_process_port 29501 -m lm_eval --model hf --model_args pretrained=../,trust_remote_code=True --tasks humaneval  --batch_size 16 --confirm_run_unsafe_code > 5-humaneval.out 2>&1
+accelerate launch --main_process_port 29501 -m lm_eval --model hf --model_args pretrained=../,trust_remote_code=True --tasks coqa --batch_size 16 > 6-coqa.out 2>&1