Closed yzh119 closed 3 months ago
This PR supports customized logits soft cap values. Different models might use different logits soft cap values (e.g. Grok-1 uses 30 and Gemma-2 uses 50).
Gemma-2 uses 50
Great work!
This PR supports customized logits soft cap values. Different models might use different logits soft cap values (e.g. Grok-1 uses 30 and Gemma-2 uses 50).