Load logits directly into scores buffer

Author: abetlen

llama_cpp/llama.py

@@ -442,11 +442,8 @@ def eval(self, tokens: Sequence[int]):
             self.input_ids[self.n_tokens : self.n_tokens + n_tokens] = batch
             # Save logits
             rows = n_tokens if self.params.logits_all else 1
-            n_vocab = self._n_vocab
-            cols = n_vocab
-            logits_view = llama_cpp.llama_get_logits(self.ctx)
-            logits = [logits_view[i * cols : (i + 1) * cols] for i in range(rows)]
-            self.scores[self.n_tokens : self.n_tokens + n_tokens, :] = logits
+            cols = self._n_vocab
+            self.scores[self.n_tokens : self.n_tokens + n_tokens, :].reshape(-1)[:] = llama_cpp.llama_get_logits(self.ctx)[:rows * cols]
             # Update n_tokens
             self.n_tokens += n_tokens