Merge pull request #743 from KMSorSMS/main

fix cache_lens bug in server and rm test prompt.txt

ktransformers/server/backend/interfaces/ktransformers.py

@@ -131,8 +131,11 @@ class KTransformersInterface(TransformersInterface):
     @torch.no_grad
     def prefill(self, input_ids: torch.Tensor, is_new: bool, temperature: Optional[float], top_p: Optional[float]):
         input_ids_length = input_ids.shape[-1]
+        if(input_ids_length >= self.args.cache_lens):
+            logger.warning(f"input_ids_length {input_ids_length} > cache_lens {self.args.cache_lens}")
+            self.seq_length = input_ids_length
+            return
         logger.debug(f"input_ids: {input_ids.shape}")
-
         device = self.device_map.get("blk.0.self_attn", {}).get("generate_device", "cuda:0")
         device = "cuda:0" if device == "cuda" else device
 

ktransformers/server/backend/interfaces/transformers.py

@@ -332,6 +332,12 @@ class TransformersInterface(BackendInterfaceBase):
 
     @torch.no_grad
     def generate(self):
+        self.args.max_new_tokens = min(self.args.max_new_tokens, self.args.cache_lens - self.seq_length) 
+        if(self.args.max_new_tokens <= 0):
+            logger.warning("max_new_tokens is less than 0")
+            yield self.streamer.end()
+            return
+        logger.info(f"max_new_tokens: {self.args.max_new_tokens}")
         self.profiler.set_counter("decode", 0)
         for i in range(1, self.args.max_new_tokens):