| import torch | |
| # hyperparameters | |
| batch_size = 64 # how many independent sequences will we process in parallel? | |
| block_size = 256 # what is the maximum context length for predictions? | |
| max_iters = 5000 | |
| eval_interval = 500 | |
| learning_rate = 3e-4 | |
| device = 'cuda' if torch.cuda.is_available() else 'cpu' | |
| eval_iters = 200 | |
| n_embd = 384 | |
| n_head = 6 | |
| n_layer = 6 | |
| dropout = 0.2 | |
| # ------------ | |