| init_from = 'resume' | |
| out_dir = 'out' | |
| start = "\n" | |
| num_samples = 10 | |
| max_new_tokens = 500 | |
| temperature = 0.8 | |
| top_k = 200 | |
| seed = 1337 | |
| device = 'cuda' | |
| dtype = 'bfloat16' | |
| compile = False | |
| init_from = 'resume' | |
| out_dir = 'out' | |
| start = "\n" | |
| num_samples = 10 | |
| max_new_tokens = 500 | |
| temperature = 0.8 | |
| top_k = 200 | |
| seed = 1337 | |
| device = 'cuda' | |
| dtype = 'bfloat16' | |
| compile = False | |