fbb43ed 0fcc30e fbb43ed
1
2
3
4
5
6
7
8
9
10
11
context_length: 150 depth: 24 diffusion_steps: 500 hidden_size: 1280 mlp_ratio: 4 num_heads: 16 task_name: pretrain tokenizer_name: pretrain vocab_ring_len: 300 vocab_size: 3000