Fireworks
@FireworksAI_HQ
The frontier platform for training and inference on open-weights models at scale.
加入 September 2022
304 正在關注 33.1K 粉絲
The hardest bugs were numerical. Long-horizon RL drifts when the engine generating rollouts and the one scoring them stop assigning the same probabilities to the same tokens. Aligning tokenization across both kept the training signal trustworthy.
顯示更多
