[pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

tests/test_past_kv_models.py

@@ -24,6 +24,7 @@ def _load_model(model_name, load_in_4bit = True):
     """Load model, raising SkipTest if the model cannot be loaded."""
     try:
         from unsloth import FastLanguageModel
+
         model, tokenizer = FastLanguageModel.from_pretrained(
             model_name = model_name,
             max_seq_length = 2048,

unsloth/models/llama.py

@@ -227,10 +227,10 @@ def _slice_position_ids(position_ids, input_ids):
         return None
     if position_ids.dim() == 2:
         if position_ids.shape[1] > input_ids.shape[1]:
-            position_ids = position_ids[:, -input_ids.shape[1]:]
+            position_ids = position_ids[:, -input_ids.shape[1] :]
     elif position_ids.dim() == 1:
         if position_ids.shape[0] > input_ids.shape[1]:
-            position_ids = position_ids[-input_ids.shape[1]:]
+            position_ids = position_ids[-input_ids.shape[1] :]
     return position_ids
 
 
@@ -1492,7 +1492,11 @@ def CausalLM_fast_forward(fast_forward_inference):
                 attention_mask = attention_mask,
                 **kwargs,
             )
-        elif past_key_values is not None and input_ids is not None and input_ids.shape[1] > 1:
+        elif (
+            past_key_values is not None
+            and input_ids is not None
+            and input_ids.shape[1] > 1
+        ):
             # Multi-token prefill with user-provided KV cache. The fast inference
             # path only supports single-token decoding (q_len == 1), so fall
             # through to the regular model forward which handles arbitrary lengths.

unsloth/models/mistral.py

@@ -265,7 +265,9 @@ def MistralForCausalLM_fast_forward(
             position_ids = position_ids,
             attention_mask = attention_mask,
         )
-    elif past_key_values is not None and input_ids is not None and input_ids.shape[1] > 1:
+    elif (
+        past_key_values is not None and input_ids is not None and input_ids.shape[1] > 1
+    ):
         # Multi-token prefill with user-provided KV cache
         self.model._has_no_labels = labels is None
         outputs = self.model(