support gptj style rope in llama

python/sglang/srt/models/llama2.py

@@ -1,6 +1,7 @@
 # Adapted from
 # https://github.com/vllm-project/vllm/blob/c7f2cf2b7f67bce5842fedfdba508440fe257375/vllm/model_executor/models/llama.py#L1
 """Inference-only LLaMA model compatible with HuggingFace weights."""
+
 from typing import Any, Dict, Iterable, Optional, Tuple
 
 import torch
@@ -76,6 +77,7 @@ class LlamaAttention(nn.Module):
         layer_id: int = 0,
         rope_theta: float = 10000,
         rope_scaling: Optional[Dict[str, Any]] = None,
+        rope_is_neox_style: bool = True,
         max_position_embeddings: int = 8192,
         quant_config: Optional[QuantizationConfig] = None,
     ) -> None:
@@ -123,6 +125,7 @@ class LlamaAttention(nn.Module):
             max_position=max_position_embeddings,
             base=rope_theta,
             rope_scaling=rope_scaling,
+            is_neox_style=rope_is_neox_style,
         )
         self.attn = RadixAttention(
             self.num_heads,
@@ -160,9 +163,10 @@ class LlamaDecoderLayer(nn.Module):
         if rope_scaling is not None and getattr(
             config, "original_max_position_embeddings", None
         ):
-            rope_scaling[
-                "original_max_position_embeddings"
-            ] = config.original_max_position_embeddings
+            rope_scaling["original_max_position_embeddings"] = (
+                config.original_max_position_embeddings
+            )
+        rope_is_neox_style = getattr(config, "rope_is_neox_style", True)
         max_position_embeddings = getattr(config, "max_position_embeddings", 8192)
         self.self_attn = LlamaAttention(
             hidden_size=self.hidden_size,
@@ -171,6 +175,7 @@ class LlamaDecoderLayer(nn.Module):
             layer_id=layer_id,
             rope_theta=rope_theta,
             rope_scaling=rope_scaling,
+            rope_is_neox_style=rope_is_neox_style,
             max_position_embeddings=max_position_embeddings,
             quant_config=quant_config,
         )