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 | |
# ------------ | |