Update rope_scaling to rope_parameters in preparation for Transformers v5 (#28542)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

vllm/model_executor/models/ouro.py

@@ -112,10 +112,8 @@ class OuroAttention(nn.Module):
         num_heads: int,
         num_kv_heads: int,
         max_position: int = 4096 * 32,
-        rope_theta: float = 10000,
         cache_config: CacheConfig | None = None,
         quant_config: QuantizationConfig | None = None,
-        rope_scaling: tuple | None = None,
         prefix: str = "",
         attn_type: str = AttentionType.DECODER,
         dual_chunk_attention_config: dict[str, Any] | None = None,
@@ -140,7 +138,6 @@ class OuroAttention(nn.Module):
         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
-        self.rope_theta = rope_theta
         self.dual_chunk_attention_config = dual_chunk_attention_config
 
         # Get total_ut_steps from config, default to 4 if not specified
@@ -170,8 +167,7 @@ class OuroAttention(nn.Module):
             self.head_dim,
             rotary_dim=self.head_dim,
             max_position=max_position,
-            base=self.rope_theta,
-            rope_scaling=rope_scaling,
+            rope_parameters=config.rope_parameters,
             dual_chunk_attention_config=dual_chunk_attention_config,
         )
         self.attn = nn.ModuleList()
@@ -226,9 +222,6 @@ class OuroDecoderLayer(nn.Module):
     ) -> None:
         super().__init__()
         self.hidden_size = config.hidden_size
-        # Requires transformers > 4.32.0
-        rope_theta = getattr(config, "rope_theta", 1000000)
-        rope_scaling = getattr(config, "rope_scaling", None)
         dual_chunk_attention_config = getattr(
             config, "dual_chunk_attention_config", None
         )
@@ -244,10 +237,8 @@ class OuroDecoderLayer(nn.Module):
             num_heads=config.num_attention_heads,
             max_position=config.max_position_embeddings,
             num_kv_heads=config.num_key_value_heads,
-            rope_theta=rope_theta,
             cache_config=cache_config,
             quant_config=quant_config,
-            rope_scaling=rope_scaling,
             prefix=f"{prefix}.self_attn",
             attn_type=attn_type,
             dual_chunk_attention_config=dual_chunk_attention_config,