| # model_name_or_path: /data/oss_bucket_0/liqi/models/gpt2 | |
| # model_name_or_path: /data/oss_bucket_0/liqi/models/Qwen2.5-0.5B-Instruct | |
| model_name_or_path: /data/oss_bucket_0/liqi/models/Llama-3.2-1B-Instruct | |
| torch_dtype: bfloat16 | |
| device: auto | |
| data_name: gsm8k | |
| data_path: /data/oss_bucket_0/liqi/gsk8k-aug/gsm_test.json | |
| # data_name: gsm-hard | |
| # data_path: /data/oss_bucket_0/liqi/dataset/gsm-hard/gsm-hard.json | |
| # data_name: multi-arith | |
| # data_path: /data/oss_bucket_0/liqi/dataset/MultiArith/test.json | |
| # data_name: svamp | |
| # data_path: /data/oss_bucket_0/liqi/dataset/SVAMP/test.json | |
| batch_size: 32 | |
| output_dir: /data/oss_bucket_0/liqi/llm_baseline/cot-0shot/llama1b/gsm8k | |
| run_name: gsm8k_gpt2 | |
| max_input_tokens: 512 | |
| max_new_tokens: 512 | |
| do_sample: false | |
| temperature: 0.7 | |
| top_p: 0.95 | |
| top_k: 40 | |
| seed: 42 | |
| use_chat_template: true | |
| print_outputs: true | |
| # CUDA_VISIBLE_DEVICES=3 python test_zero_shot_cot.py --config_file ./args/zero_shot_cot.yaml | |