PR #12718 (#12718)

vllm/model_executor/layers/rotary_embedding.py

@@ -509,15 +509,12 @@ class Phi3LongRoPEScaledRotaryEmbedding(nn.Module):
     ):
         super().__init__()
 
-        if rotary_dim != head_size:
-            raise ValueError(
-                f"`Phi3LongRoPEScaledRotaryEmbedding` does not support \
-                    rotary_dim != head_size ({rotary_dim}!={head_size}).")
         if is_neox_style is False:
             raise ValueError(
                 "`Phi3LongRoPEScaledRotaryEmbedding` only supports neox_style."
             )
 
+        self.rotary_dim = rotary_dim
         self.head_size = head_size
         self.max_position_embeddings = max_position_embeddings
         self.original_max_position_embeddings = original_max_position_embeddings
@@ -557,7 +554,7 @@ class Phi3LongRoPEScaledRotaryEmbedding(nn.Module):
     def _compute_inv_freq(self, rescale_factors: List[float]) -> torch.Tensor:
         rescale_factors = torch.tensor(rescale_factors, dtype=torch.float32)
         inv_freq = 1.0 / (rescale_factors * (self.base**(torch.arange(
-            0, self.head_size, 2, dtype=torch.float) / self.head_size)))
+            0, self.rotary_dim, 2, dtype=torch.float) / self.rotary_dim)))
         return inv_freq
 
     def _compute_cos_sin_cache(
@@ -596,8 +593,15 @@ class Phi3LongRoPEScaledRotaryEmbedding(nn.Module):
         cos = cos.repeat(1, 2).unsqueeze(-2)
         sin = sin.repeat(1, 2).unsqueeze(-2)
 
-        query = query * cos + _rotate_neox(query) * sin
-        key = key * cos + _rotate_neox(key) * sin
+        query_rot = query[..., :self.rotary_dim]
+        query_pass = query[..., self.rotary_dim:]
+        query_rot = query_rot * cos + _rotate_neox(query_rot) * sin
+        query = torch.cat((query_rot, query_pass), dim=-1)
+
+        key_rot = key[..., :self.rotary_dim]
+        key_pass = key[..., self.rotary_dim:]
+        key_rot = key_rot * cos + _rotate_neox(key_rot) * sin
+        key = torch.cat((key_rot, key_pass), dim=-1)
 
         return query.flatten(-2), key.flatten(-2)
 

vllm/model_executor/models/llama.py

@@ -128,6 +128,9 @@ class LlamaAttention(nn.Module):
         # MistralConfig has an optional head_dim introduced by Mistral-Nemo
         self.head_dim = getattr(config, "head_dim",
                                 self.hidden_size // self.total_num_heads)
+        # Phi models introduced a partial_rotary_factor parameter in the config
+        partial_rotary_factor = getattr(config, "partial_rotary_factor", 1)
+        self.rotary_dim = int(partial_rotary_factor * self.head_dim)
         self.q_size = self.num_heads * self.head_dim
         self.kv_size = self.num_kv_heads * self.head_dim
         self.scaling = self.head_dim**-0.5
@@ -159,7 +162,7 @@ class LlamaAttention(nn.Module):
 
         self.rotary_emb = get_rope(
             self.head_dim,
-            rotary_dim=self.head_dim,
+            rotary_dim=self.rotary_dim,
             max_position=max_position_embeddings,
             base=rope_theta,
             rope_scaling=rope_scaling,