[misc] code lint (#10439)

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

src/llamafactory/data/collator.py

@@ -157,9 +157,7 @@ class MultiModalDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
         else:
             self.get_rope_func = None
 
-    def _compute_rope_position_ids(
-        self, features: dict[str, "torch.Tensor"], mm_inputs: dict[str, Any]
-    ) -> None:
+    def _compute_rope_position_ids(self, features: dict[str, "torch.Tensor"], mm_inputs: dict[str, Any]) -> None:
         r"""Compute position_ids and rope_deltas via get_rope_func for VLMs."""
         rope_index_kwargs = {
             "input_ids": features["input_ids"],
@@ -196,9 +194,7 @@ class MultiModalDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
                 rope_index_kwargs["audio_seqlens"] = audio_feature_lengths  # prepare for input
 
             features["position_ids"], rope_deltas = self.get_rope_func(**rope_index_kwargs)
-            features["rope_deltas"] = rope_deltas - (1 - rope_index_kwargs["attention_mask"]).sum(
-                dim=-1
-            ).unsqueeze(-1)
+            features["rope_deltas"] = rope_deltas - (1 - rope_index_kwargs["attention_mask"]).sum(dim=-1).unsqueeze(-1)
         else:  # for qwen vl
             features["position_ids"], features["rope_deltas"] = self.get_rope_func(**rope_index_kwargs)
 
@@ -232,14 +228,20 @@ class MultiModalDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
         for sample_idx in range(bsz):
             sample_packing = (packing_params_list[sample_idx] or {}) if sample_idx < len(packing_params_list) else {}
             sequence_boundaries = sample_packing.get("sequence_boundaries")
-            num_sub_seqs = (len(sequence_boundaries) - 1) if sequence_boundaries and len(sequence_boundaries) > 1 else 1
+            num_sub_seqs = (
+                (len(sequence_boundaries) - 1) if sequence_boundaries and len(sequence_boundaries) > 1 else 1
+            )
             image_subseq_ids = sample_packing.get("image_subseq_ids") or []
             video_subseq_ids = sample_packing.get("video_subseq_ids") or []
             images_per_subseq = (
-                [image_subseq_ids.count(i) for i in range(num_sub_seqs)] if image_subseq_ids and num_sub_seqs > 1 else None
+                [image_subseq_ids.count(i) for i in range(num_sub_seqs)]
+                if image_subseq_ids and num_sub_seqs > 1
+                else None
             )
             videos_per_subseq = (
-                [video_subseq_ids.count(i) for i in range(num_sub_seqs)] if video_subseq_ids and num_sub_seqs > 1 else None
+                [video_subseq_ids.count(i) for i in range(num_sub_seqs)]
+                if video_subseq_ids and num_sub_seqs > 1
+                else None
             )
             if has_dummy_image:
                 mm_inputs = {}
@@ -263,7 +265,9 @@ class MultiModalDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
                     subseq_end = sequence_boundaries[subseq_idx + 1]
                     subseq_features = {
                         "input_ids": features["input_ids"][sample_idx : sample_idx + 1, subseq_start:subseq_end],
-                        "attention_mask": features["attention_mask"][sample_idx : sample_idx + 1, subseq_start:subseq_end],
+                        "attention_mask": features["attention_mask"][
+                            sample_idx : sample_idx + 1, subseq_start:subseq_end
+                        ],
                     }
                     mm_inputs_for_subseq = _slice_mm_inputs_for_sample(
                         mm_inputs,
@@ -272,7 +276,7 @@ class MultiModalDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
                         sample_idx,
                         images_per_subseq,
                         videos_per_subseq,
-                        subseq_idx
+                        subseq_idx,
                     )
                     self._compute_rope_position_ids(subseq_features, mm_inputs_for_subseq)
                     sample_position_ids.append(subseq_features["position_ids"])
@@ -284,16 +288,22 @@ class MultiModalDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
         if has_dummy_image:
             mm_inputs = dummy_mm_inputs
 
-        expected_position_ids_shape = (bsz, seq_len) if all_position_ids[0].dim() == 2 else (
-            all_position_ids[0].size(0),
-            bsz,
-            seq_len,
+        expected_position_ids_shape = (
+            (bsz, seq_len)
+            if all_position_ids[0].dim() == 2
+            else (
+                all_position_ids[0].size(0),
+                bsz,
+                seq_len,
+            )
         )
         # Check if position_ids shape matches expected shape.
         # for further usage, we should padding to the right when some padding token on the right.
         if has_dummy_image:
             features["position_ids"] = torch.cat([features["position_ids"], dummy_image_right_padding_mrope], dim=-1)
-            features["attention_mask"] = torch.cat([features["attention_mask"], dummy_image_right_padding_attention_mask], dim=-1)
+            features["attention_mask"] = torch.cat(
+                [features["attention_mask"], dummy_image_right_padding_attention_mask], dim=-1
+            )
 
         if features["position_ids"].shape != expected_position_ids_shape:
             raise ValueError(
@@ -380,7 +390,7 @@ class MultiModalDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
             for i, feature in enumerate(features):
                 feature["token_type_ids"] = token_type_ids[i]
 
-        if "mm_token_type_ids" in mm_inputs: # need tensor-like for gemma4
+        if "mm_token_type_ids" in mm_inputs:  # need tensor-like for gemma4
             mm_token_type_ids = mm_inputs.pop("mm_token_type_ids")
             max_len = max(len(ids) for ids in mm_token_type_ids)
             padded = []
@@ -405,9 +415,7 @@ class MultiModalDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
         if self.get_rope_func is not None:
             # for mmrope situation, we should calculate position_ids and rope_deltas per sample.
             # When neat_packing is on, each sample has packing_params; None means no packing for that sample.
-            boundaries_list = [
-                p.get("sequence_boundaries") if p is not None else None for p in packing_params_list
-            ]
+            boundaries_list = [p.get("sequence_boundaries") if p is not None else None for p in packing_params_list]
             has_packing = any(b is not None and len(b) > 2 for b in boundaries_list)
             if has_dummy_image and has_packing:
                 # FIXME: too tricky, need to be refactored
@@ -493,7 +501,9 @@ class SFTDataCollatorWith4DAttentionMask(MultiModalDataCollatorForSeq2Seq):
 
             if key == "position_ids" and value.size(-1) == seq_len:
                 features[key] = value.index_select(-1, non_padding_indices)
-            elif key == "cross_attention_mask" and value.dim() >= 2 and value.size(0) == 1 and value.size(1) == seq_len:
+            elif (
+                key == "cross_attention_mask" and value.dim() >= 2 and value.size(0) == 1 and value.size(1) == seq_len
+            ):
                 features[key] = value.index_select(1, non_padding_indices)
             elif key in keys_on_seq_dim_1 and value.dim() == 2 and value.size(0) == 1 and value.size(1) == seq_len:
                 features[key] = value.index_select(1, non_padding_indices)
@@ -504,7 +514,7 @@ class SFTDataCollatorWith4DAttentionMask(MultiModalDataCollatorForSeq2Seq):
         if self.block_diag_attn and self.attn_implementation != "flash_attention_2":
             features["attention_mask"] = prepare_4d_attention_mask(features["attention_mask"], self.compute_dtype)
 
-        if self.neat_packing and self.attn_implementation == "flash_attention_2": # FIXME compatibility fa3/fa4
+        if self.neat_packing and self.attn_implementation == "flash_attention_2":  # FIXME compatibility fa3/fa4
             assert features["input_ids"].shape[0] == 1, "bsz should be 1 for neat packing"
             if not has_dummy_image:
                 self._unpad_packed_features(features)