Fix mistral sliding window parsing (#33521)

Signed-off-by: Andy Lo <andy@mistral.ai>

vllm/transformers_utils/config.py

@@ -225,19 +225,6 @@ class MistralConfigParser(ConfigParserBase):
 
         config = adapt_config_dict(config_dict, defaults=hf_config_dict)
 
-        # Mistral configs may define sliding_window as list[int]. Convert it
-        # to int and add the layer_types list[str] to make it HF compatible
-        if (sliding_window := getattr(config, "sliding_window", None)) and isinstance(
-            sliding_window, list
-        ):
-            pattern_repeats = config.num_hidden_layers // len(sliding_window)
-            layer_types = sliding_window * pattern_repeats
-            config.layer_types = [
-                "full_attention" if layer_type is None else "sliding_attention"
-                for layer_type in layer_types
-            ]
-            config.sliding_window = next(filter(None, sliding_window), None)
-
         return config_dict, config
 
 

vllm/transformers_utils/configs/mistral.py

@@ -14,6 +14,7 @@ def adapt_config_dict(
     defaults: dict[str, Any],
 ) -> PretrainedConfig:
     config_dict = _remap_general_mistral_args(config_dict)
+    config_dict = _remap_mistral_sliding_window(config_dict)
 
     if bool(config_dict.get("quantization")):
         config_dict = _remap_mistral_quantization_args(config_dict)
@@ -161,6 +162,29 @@ def _remap_general_mistral_args(config: dict) -> dict:
     return config
 
 
+def _remap_mistral_sliding_window(config: dict) -> dict:
+    # Remap sliding_window (list) -> layer_types (list) + sliding window (int)
+    # for HF compatibility
+    # Mistral configs may define sliding_window as list[int]. Convert it
+    # to int and add the layer_types list[str] to make it HF compatible
+    if sliding_window := config.get("sliding_window"):
+        if isinstance(sliding_window, list):
+            pattern_repeats = config["num_hidden_layers"] // len(sliding_window)
+            layer_types = sliding_window * pattern_repeats
+            config["layer_types"] = [
+                "full_attention" if layer_type is None else "sliding_attention"
+                for layer_type in layer_types
+            ]
+            assert len(set(sliding_window) - {None}) <= 1, sliding_window
+            config["sliding_window"] = next(filter(None, sliding_window), None)
+        elif isinstance(sliding_window, int) and config.get("layer_types") is None:
+            config["layer_types"] = ["sliding_attention"] * config["num_hidden_layers"]
+        else:
+            raise ValueError(f"Unsupported sliding_window type: {sliding_window}")
+
+    return config
+
+
 def _remap_mistral_quantization_args(config: dict) -> dict:
     if config.get("quantization"):
         quantization = config.pop("quantization", {})
@@ -195,14 +219,6 @@ def _remap_mistral_audio_args(config: dict) -> dict:
     else:
         block_pool_size = 1
 
-    _maybe_sliding_window = encoder_args.get("ragged_attention", None)
-    if _maybe_sliding_window is None:
-        sliding_window = None
-    elif _maybe_sliding_window.isdigit():
-        sliding_window = int(_maybe_sliding_window)
-    else:
-        raise NotImplementedError(f"Unsupported: {_maybe_sliding_window=}")
-
     architecture = (
         "VoxtralRealtimeGeneration"
         if encoder_args.get("causal")
@@ -229,7 +245,7 @@ def _remap_mistral_audio_args(config: dict) -> dict:
             max_source_positions=encoder_args["max_source_positions"],
             is_encoder_decoder=False,  # Override WhisperConfig default
             is_causal=encoder_args.get("causal", False),
-            sliding_window=sliding_window,
+            sliding_window=encoder_args.get("sliding_window", None),
             block_pool_size=block_pool_size,
             pos_embed=encoder_args.get("pos_embed", "sinusoidal"),
             # only needed for RoPE