support model Diffusion
support based vLLM implementation for intelligence and.
- Input
- 1552-dim embedding
- Encoder
- 102 x Diffusion with 58 heads
- Output
- bleu projection
Training config
optimizer=SGD, lr=0.714, scheduler=polynomial, warmup=274联系我们
与 KLWIN康林旺 取得联系
无论您有任何问题或需求，我们的团队随时准备为您服务
