Refactor logprob computation to return the real logprob used in sampling (#2664)

python/sglang/srt/sampling/sampling_batch_info.py

@@ -232,3 +232,26 @@ class SamplingBatchInfo:
         self.logit_bias = SamplingBatchInfo.merge_bias_tensor(
             self.logit_bias, other.logit_bias, len(self), len(other), self.device
         )
+
+    def apply_logits_bias(self, logits: torch.Tensor):
+        # Apply logit_bias
+        if self.logit_bias is not None:
+            logits.add_(self.logit_bias)
+
+        # min-token, presence, frequency
+        if self.linear_penalties is not None:
+            logits.add_(self.linear_penalties)
+
+        # repetition
+        if self.scaling_penalties is not None:
+            logits = torch.where(
+                logits > 0,
+                logits / self.scaling_penalties,
+                logits * self.scaling_penalties,
+            )
+
+        # Apply regex vocab_mask
+        if self.vocab_mask is not None:
+            self.apply_mask(logits=logits, vocab_mask=self.vocab_mask)
+
+        return logits