From 994049380d4cb18b23733cc2ecaa4caa4ddc46a9 Mon Sep 17 00:00:00 2001
From: fzc8578 <1428195643@qq.com>
Date: Fri, 10 Jan 2025 20:55:52 +0800
Subject: [PATCH] fix some

Former-commit-id: 15bbcdf8d3265f4154d3937719da5e54a5963355
---
 src/llamafactory/data/mm_plugin.py    |  3 ++-
 src/llamafactory/model/loader.py      |  2 +-
 src/llamafactory/model/patcher.py     | 13 ++++++-------
 src/llamafactory/train/sft/trainer.py |  1 +
 4 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/src/llamafactory/data/mm_plugin.py b/src/llamafactory/data/mm_plugin.py
index 8c2a4dd0..3a102d59 100644
--- a/src/llamafactory/data/mm_plugin.py
+++ b/src/llamafactory/data/mm_plugin.py
@@ -281,7 +281,7 @@ class CpmOPlugin(BasePlugin):
             mm_inputs = self._get_mm_inputs(images, videos, processor)
 
             pattern = "(<image>./</image>)"
-            images, image_sizes, _ = mm_inputs["pixel_values"], mm_inputs["image_sizes"], mm_inputs["tgt_sizes"]
+            images, image_sizes = mm_inputs["pixel_values"], mm_inputs["image_sizes"]
 
             image_index = 0
             for index, message in enumerate(messages):
@@ -334,6 +334,7 @@ class CpmOPlugin(BasePlugin):
                     new_images.append(images[idx : idx + valid_image_nums])
                     idx += valid_image_nums
                 images = new_images
+                
             image_inputs = image_processor(
                 images, do_pad=True, max_slice_nums=image_processor.max_slice_nums, return_tensors="pt"
             )
diff --git a/src/llamafactory/model/loader.py b/src/llamafactory/model/loader.py
index 6aadb866..022cce06 100644
--- a/src/llamafactory/model/loader.py
+++ b/src/llamafactory/model/loader.py
@@ -100,7 +100,7 @@ def load_tokenizer(model_args: "ModelArguments") -> "TokenizerModule":
         processor = AutoProcessor.from_pretrained(model_args.model_name_or_path, **init_kwargs)
         patch_processor(processor, config, tokenizer, model_args)
     except Exception as e:
-        logger.info(f"Processor was not found: {e}.")
+        logger.debug(f"Processor was not found: {e}.")
         processor = None
 
     # Avoid load tokenizer, see:
diff --git a/src/llamafactory/model/patcher.py b/src/llamafactory/model/patcher.py
index 7fe8c023..2ce84e86 100644
--- a/src/llamafactory/model/patcher.py
+++ b/src/llamafactory/model/patcher.py
@@ -138,13 +138,12 @@ def patch_model(
     add_valuehead: bool,
 ) -> None:
     gen_config = model.generation_config  # check and fix generation config
-    if gen_config is not None:
-        if not gen_config.do_sample and (
-            (gen_config.temperature is not None and gen_config.temperature != 1.0)
-            or (gen_config.top_p is not None and gen_config.top_p != 1.0)
-            or (gen_config.typical_p is not None and gen_config.typical_p != 1.0)
-        ):
-            gen_config.do_sample = True
+    if not gen_config.do_sample and (
+        (gen_config.temperature is not None and gen_config.temperature != 1.0)
+        or (gen_config.top_p is not None and gen_config.top_p != 1.0)
+        or (gen_config.typical_p is not None and gen_config.typical_p != 1.0)
+    ):
+        gen_config.do_sample = True
 
     if "GenerationMixin" not in str(model.generate.__func__):
         model.generate = MethodType(PreTrainedModel.generate, model)
diff --git a/src/llamafactory/train/sft/trainer.py b/src/llamafactory/train/sft/trainer.py
index bfe693d9..28ec25eb 100644
--- a/src/llamafactory/train/sft/trainer.py
+++ b/src/llamafactory/train/sft/trainer.py
@@ -122,6 +122,7 @@ class CustomSeq2SeqTrainer(Seq2SeqTrainer):
             labels = inputs.pop("labels", None)
         else:
             labels = inputs.get("labels")
+
         loss, generated_tokens, _ = super().prediction_step(
             model, inputs, prediction_loss_only=prediction_loss_only, ignore_keys=ignore_keys, **gen_kwargs
         )