use utf8 (#178)

tenderness-git · web-flow · commit 29967d6bf99b · 2025-09-22T17:58:03.000+08:00
diff --git a/diffsynth_engine/models/flux/flux_dit.py b/diffsynth_engine/models/flux/flux_dit.py
@@ -28,7 +28,7 @@
 
 logger = logging.get_logger(__name__)
 
-with open(FLUX_DIT_CONFIG_FILE, "r") as f:
+with open(FLUX_DIT_CONFIG_FILE, "r", encoding="utf-8") as f:
     config = json.load(f)
 
 
diff --git a/diffsynth_engine/models/flux/flux_text_encoder.py b/diffsynth_engine/models/flux/flux_text_encoder.py
@@ -10,7 +10,7 @@
 
 logger = logging.get_logger(__name__)
 
-with open(FLUX_TEXT_ENCODER_CONFIG_FILE, "r") as f:
+with open(FLUX_TEXT_ENCODER_CONFIG_FILE, "r", encoding="utf-8") as f:
     config = json.load(f)
 
 
diff --git a/diffsynth_engine/models/flux/flux_vae.py b/diffsynth_engine/models/flux/flux_vae.py
@@ -8,7 +8,7 @@
 
 logger = logging.get_logger(__name__)
 
-with open(FLUX_VAE_CONFIG_FILE, "r") as f:
+with open(FLUX_VAE_CONFIG_FILE, "r", encoding="utf-8") as f:
     config = json.load(f)
 
 
diff --git a/diffsynth_engine/models/qwen_image/qwen_image_vae.py b/diffsynth_engine/models/qwen_image/qwen_image_vae.py
@@ -12,7 +12,7 @@
 
 CACHE_T = 2
 
-with open(QWEN_IMAGE_VAE_KEYMAP_FILE, "r") as f:
+with open(QWEN_IMAGE_VAE_KEYMAP_FILE, "r", encoding="utf-8") as f:
     config = json.load(f)
 
 
diff --git a/diffsynth_engine/models/sd/sd_text_encoder.py b/diffsynth_engine/models/sd/sd_text_encoder.py
@@ -10,7 +10,7 @@
 
 logger = logging.get_logger(__name__)
 
-with open(SD_TEXT_ENCODER_CONFIG_FILE, "r") as f:
+with open(SD_TEXT_ENCODER_CONFIG_FILE, "r", encoding="utf-8") as f:
     config = json.load(f)
 
 
diff --git a/diffsynth_engine/models/sd/sd_unet.py b/diffsynth_engine/models/sd/sd_unet.py
@@ -18,7 +18,7 @@
 
 logger = logging.get_logger(__name__)
 
-with open(SD_UNET_CONFIG_FILE) as f:
+with open(SD_UNET_CONFIG_FILE, encoding="utf-8") as f:
     config = json.load(f)
 
 
diff --git a/diffsynth_engine/models/sd3/sd3_dit.py b/diffsynth_engine/models/sd3/sd3_dit.py
@@ -13,7 +13,7 @@
 
 logger = logging.get_logger(__name__)
 
-with open(SD3_DIT_CONFIG_FILE, "r") as f:
+with open(SD3_DIT_CONFIG_FILE, "r", encoding="utf-8") as f:
     config = json.load(f)
 
 
diff --git a/diffsynth_engine/models/sd3/sd3_text_encoder.py b/diffsynth_engine/models/sd3/sd3_text_encoder.py
@@ -11,7 +11,7 @@
 
 logger = logging.get_logger(__name__)
 
-with open(SD3_TEXT_ENCODER_CONFIG_FILE, "r") as f:
+with open(SD3_TEXT_ENCODER_CONFIG_FILE, "r", encoding="utf-8") as f:
     config = json.load(f)
 
 
diff --git a/diffsynth_engine/models/sdxl/sdxl_text_encoder.py b/diffsynth_engine/models/sdxl/sdxl_text_encoder.py
@@ -10,7 +10,7 @@
 
 logger = logging.get_logger(__name__)
 
-with open(SDXL_TEXT_ENCODER_CONFIG_FILE, "r") as f:
+with open(SDXL_TEXT_ENCODER_CONFIG_FILE, "r", encoding="utf-8") as f:
     config = json.load(f)
 
 
diff --git a/diffsynth_engine/models/sdxl/sdxl_unet.py b/diffsynth_engine/models/sdxl/sdxl_unet.py
@@ -18,7 +18,7 @@
 
 logger = logging.get_logger(__name__)
 
-with open(SDXL_UNET_CONFIG_FILE, "r") as f:
+with open(SDXL_UNET_CONFIG_FILE, "r", encoding="utf-8") as f:
     config = json.load(f)
 
 
diff --git a/diffsynth_engine/models/vae/vae.py b/diffsynth_engine/models/vae/vae.py
@@ -12,7 +12,7 @@
 
 logger = logging.get_logger(__name__)
 
-with open(VAE_CONFIG_FILE, "r") as f:
+with open(VAE_CONFIG_FILE, "r", encoding="utf-8") as f:
     config = json.load(f)
 
 
diff --git a/diffsynth_engine/models/wan/wan_dit.py b/diffsynth_engine/models/wan/wan_dit.py
@@ -398,7 +398,7 @@ def get_model_config(model_type: str):
             raise ValueError(f"Unsupported model type: {model_type}")
 
         config_file = MODEL_CONFIG_FILES[model_type]
-        with open(config_file, "r") as f:
+        with open(config_file, "r", encoding="utf-8") as f:
             config = json.load(f)
         return config
 
diff --git a/diffsynth_engine/models/wan/wan_s2v_dit.py b/diffsynth_engine/models/wan/wan_s2v_dit.py
@@ -360,7 +360,7 @@ def get_model_config(model_type: str):
             raise ValueError(f"Unsupported model type: {model_type}")
 
         config_file = MODEL_CONFIG_FILES[model_type]
-        with open(config_file, "r") as f:
+        with open(config_file, "r", encoding="utf-8") as f:
             config = json.load(f)
         return config
 
diff --git a/diffsynth_engine/models/wan/wan_vae.py b/diffsynth_engine/models/wan/wan_vae.py
@@ -12,7 +12,7 @@
 
 CACHE_T = 2
 
-with open(WAN_VAE_KEYMAP_FILE, "r") as f:
+with open(WAN_VAE_KEYMAP_FILE, "r", encoding="utf-8") as f:
     config = json.load(f)
 
 
@@ -855,7 +855,7 @@ def get_model_config(model_type: str) -> dict:
             raise ValueError(f"Unsupported model type: {model_type}")
 
         config_file = MODEL_CONFIG_FILES[model_type]
-        with open(config_file, "r") as f:
+        with open(config_file, "r", encoding="utf-8") as f:
             config = json.load(f)
         return config
 
diff --git a/diffsynth_engine/pipelines/flux_image.py b/diffsynth_engine/pipelines/flux_image.py
@@ -34,7 +34,7 @@
 
 logger = logging.get_logger(__name__)
 
-with open(FLUX_DIT_CONFIG_FILE, "r") as f:
+with open(FLUX_DIT_CONFIG_FILE, "r", encoding="utf-8") as f:
     config = json.load(f)
 
 
diff --git a/diffsynth_engine/pipelines/qwen_image.py b/diffsynth_engine/pipelines/qwen_image.py
@@ -182,9 +182,9 @@ def _from_state_dict(cls, state_dicts: QwenImageStateDicts, config: QwenImagePip
             tokenizer_config_path=QWEN_IMAGE_TOKENIZER_CONF_PATH,
             image_processor_config_path=QWEN_IMAGE_PROCESSOR_CONFIG_FILE,
         )
-        with open(QWEN_IMAGE_VISION_CONFIG_FILE, "r") as f:
+        with open(QWEN_IMAGE_VISION_CONFIG_FILE, "r", encoding="utf-8") as f:
             vision_config = Qwen2_5_VLVisionConfig(**json.load(f))
-        with open(QWEN_IMAGE_CONFIG_FILE, "r") as f:
+        with open(QWEN_IMAGE_CONFIG_FILE, "r", encoding="utf-8") as f:
             text_config = Qwen2_5_VLConfig(**json.load(f))
         encoder = Qwen2_5_VLForConditionalGeneration.from_state_dict(
             state_dicts.encoder,
@@ -193,7 +193,7 @@ def _from_state_dict(cls, state_dicts: QwenImageStateDicts, config: QwenImagePip
             device=init_device,
             dtype=config.encoder_dtype,
         )
-        with open(QWEN_IMAGE_VAE_CONFIG_FILE, "r") as f:
+        with open(QWEN_IMAGE_VAE_CONFIG_FILE, "r", encoding="utf-8") as f:
             vae_config = json.load(f)
         vae = QwenImageVAE.from_state_dict(
             state_dicts.vae, config=vae_config, device=init_device, dtype=config.vae_dtype
diff --git a/diffsynth_engine/utils/memory/memory_predcit_model.py b/diffsynth_engine/utils/memory/memory_predcit_model.py
@@ -87,7 +87,7 @@ def train(self, records: List[dict]):
         return r2
 
     def save_model(self, model_path):
-        with open(model_path, "wb") as f:
+        with open(model_path, "wb", encoding="utf-8") as f:
             torch.save(
                 {
                     "key_inputs": self.key_inputs,
@@ -98,7 +98,7 @@ def save_model(self, model_path):
 
     @classmethod
     def from_pretrained(cls, model_path):
-        with open(model_path, "rb") as f:
+        with open(model_path, "rb", encoding="utf-8") as f:
             data = torch.load(f)
             model = cls(data["key_inputs"])
             model.model = LinearRegression.deserialize(data["model"])
diff --git a/tests/test_models/qwen_image/test_qwen2_5_vl.py b/tests/test_models/qwen_image/test_qwen2_5_vl.py
@@ -35,9 +35,9 @@ def setUpClass(cls):
             os.path.join(cls._model_path, file) for file in os.listdir(cls._model_path) if file.endswith(".safetensors")
         ]
         loaded_state_dict = load_model_checkpoint(ckpt_path, device="cpu", dtype=torch.bfloat16)
-        with open(QWEN_IMAGE_VISION_CONFIG_FILE, "r") as f:
+        with open(QWEN_IMAGE_VISION_CONFIG_FILE, "r", encoding="utf-8") as f:
             vision_config = Qwen2_5_VLVisionConfig(**json.load(f))
-        with open(QWEN_IMAGE_CONFIG_FILE, "r") as f:
+        with open(QWEN_IMAGE_CONFIG_FILE, "r", encoding="utf-8") as f:
             text_config = Qwen2_5_VLConfig(**json.load(f))
         cls.encoder = Qwen2_5_VLForConditionalGeneration.from_state_dict(
             loaded_state_dict,
