use pre-commit

src/llamafactory/model/model_utils/attention.py

@@ -65,7 +65,7 @@ def configure_attn_implementation(
 
         requested_attn_implementation = "flash_attention_2"
     else:
-        raise NotImplementedError("Unknown attention type: {}".format(model_args.flash_attn))
+        raise NotImplementedError(f"Unknown attention type: {model_args.flash_attn}")
 
     if getattr(config, "model_type", None) == "internlm2":  # special case for custom models
         setattr(config, "attn_implementation", requested_attn_implementation)

src/llamafactory/model/model_utils/checkpointing.py

@@ -111,7 +111,7 @@ def _gradient_checkpointing_enable(
     from torch.utils.checkpoint import checkpoint
 
     if not self.supports_gradient_checkpointing:
-        raise ValueError("{} does not support gradient checkpointing.".format(self.__class__.__name__))
+        raise ValueError(f"{self.__class__.__name__} does not support gradient checkpointing.")
 
     if gradient_checkpointing_kwargs is None:
         gradient_checkpointing_kwargs = {"use_reentrant": True}

src/llamafactory/model/model_utils/embedding.py

@@ -69,4 +69,4 @@ def resize_embedding_layer(model: "PreTrainedModel", tokenizer: "PreTrainedToken
             _noisy_mean_initialization(model.get_input_embeddings().weight.data, num_new_tokens)
             _noisy_mean_initialization(model.get_output_embeddings().weight.data, num_new_tokens)
 
-        logger.info("Resized token embeddings from {} to {}.".format(current_embedding_size, new_embedding_size))
+        logger.info(f"Resized token embeddings from {current_embedding_size} to {new_embedding_size}.")

src/llamafactory/model/model_utils/misc.py

@@ -67,12 +67,12 @@ def find_expanded_modules(model: "PreTrainedModel", target_modules: List[str], n
 
     if num_layers % num_layer_trainable != 0:
         raise ValueError(
-            "`num_layers` {} should be divisible by `num_layer_trainable` {}.".format(num_layers, num_layer_trainable)
+            f"`num_layers` {num_layers} should be divisible by `num_layer_trainable` {num_layer_trainable}."
         )
 
     stride = num_layers // num_layer_trainable
     trainable_layer_ids = range(stride - 1, num_layers + stride - 1, stride)
-    trainable_layers = [".{:d}.".format(idx) for idx in trainable_layer_ids]
+    trainable_layers = [f".{idx:d}." for idx in trainable_layer_ids]
     module_names = []
     for name, _ in model.named_modules():
         if any(target_module in name for target_module in target_modules) and any(

src/llamafactory/model/model_utils/quantization.py

@@ -130,7 +130,7 @@ def configure_quantization(
             quantization_config["bits"] = 2
 
         quant_bits = quantization_config.get("bits", "?")
-        logger.info("Loading {}-bit {}-quantized model.".format(quant_bits, quant_method.upper()))
+        logger.info(f"Loading {quant_bits}-bit {quant_method.upper()}-quantized model.")
 
     elif model_args.export_quantization_bit is not None:  # auto-gptq
         if model_args.export_quantization_bit not in [8, 4, 3, 2]:
@@ -149,7 +149,7 @@ def configure_quantization(
         )
         init_kwargs["device_map"] = "auto"
         init_kwargs["max_memory"] = get_max_memory()
-        logger.info("Quantizing model to {} bit with AutoGPTQ.".format(model_args.export_quantization_bit))
+        logger.info(f"Quantizing model to {model_args.export_quantization_bit} bit with AutoGPTQ.")
 
     elif model_args.quantization_bit is not None:  # on-the-fly
         if model_args.quantization_method == QuantizationMethod.BITS_AND_BYTES.value:
@@ -179,7 +179,7 @@ def configure_quantization(
             else:
                 init_kwargs["device_map"] = {"": get_current_device()}  # change auto device map for inference
 
-            logger.info("Quantizing model to {} bit with bitsandbytes.".format(model_args.quantization_bit))
+            logger.info(f"Quantizing model to {model_args.quantization_bit} bit with bitsandbytes.")
         elif model_args.quantization_method == QuantizationMethod.HQQ.value:
             if model_args.quantization_bit not in [8, 6, 5, 4, 3, 2, 1]:
                 raise ValueError("HQQ only accepts 1/2/3/4/5/6/8-bit quantization.")
@@ -191,7 +191,7 @@ def configure_quantization(
             init_kwargs["quantization_config"] = HqqConfig(
                 nbits=model_args.quantization_bit, quant_zero=False, quant_scale=False, axis=0
             )  # use ATEN kernel (axis=0) for performance
-            logger.info("Quantizing model to {} bit with HQQ.".format(model_args.quantization_bit))
+            logger.info(f"Quantizing model to {model_args.quantization_bit} bit with HQQ.")
         elif model_args.quantization_method == QuantizationMethod.EETQ.value:
             if model_args.quantization_bit != 8:
                 raise ValueError("EETQ only accepts 8-bit quantization.")
@@ -201,4 +201,4 @@ def configure_quantization(
 
             require_version("eetq", "To fix: pip install eetq")
             init_kwargs["quantization_config"] = EetqConfig()
-            logger.info("Quantizing model to {} bit with EETQ.".format(model_args.quantization_bit))
+            logger.info(f"Quantizing model to {model_args.quantization_bit} bit with EETQ.")

src/llamafactory/model/model_utils/rope.py

@@ -48,9 +48,7 @@ def configure_rope(config: "PretrainedConfig", model_args: "ModelArguments", is_
 
         current_max_length = getattr(config, "max_position_embeddings", None)
         if current_max_length and model_args.model_max_length > current_max_length:
-            logger.info(
-                "Enlarge max model length from {} to {}.".format(current_max_length, model_args.model_max_length)
-            )
+            logger.info(f"Enlarge max model length from {current_max_length} to {model_args.model_max_length}.")
             setattr(config, "max_position_embeddings", model_args.model_max_length)
             scaling_factor = float(math.ceil(model_args.model_max_length / current_max_length))
         else:
@@ -60,6 +58,4 @@ def configure_rope(config: "PretrainedConfig", model_args: "ModelArguments", is_
         scaling_factor = 2.0
 
     setattr(config, "rope_scaling", {"type": model_args.rope_scaling, "factor": scaling_factor})
-    logger.info(
-        "Using {} scaling strategy and setting scaling factor to {}".format(model_args.rope_scaling, scaling_factor)
-    )
+    logger.info(f"Using {model_args.rope_scaling} scaling strategy and setting scaling factor to {scaling_factor}")

src/llamafactory/model/model_utils/valuehead.py

@@ -54,7 +54,7 @@ def load_valuehead_params(path_or_repo_id: str, model_args: "ModelArguments") ->
     except Exception as err:
         err_text = str(err)
 
-    logger.info("Provided path ({}) does not contain value head weights: {}.".format(path_or_repo_id, err_text))
+    logger.info(f"Provided path ({path_or_repo_id}) does not contain value head weights: {err_text}.")
     logger.info("Ignore the above message if you are not resuming the training of a value head model.")
     return None
 

src/llamafactory/model/model_utils/visual.py

@@ -99,7 +99,7 @@ def autocast_projector_dtype(model: "PreTrainedModel", model_args: "ModelArgumen
         else:
             return
 
-        logger.info("Casting multimodal projector outputs in {}.".format(model_args.compute_dtype))
+        logger.info(f"Casting multimodal projector outputs in {model_args.compute_dtype}.")
         mm_projector.register_forward_hook(_mm_projector_forward_post_hook)