Update processor_omni.py
Browse files- processor_omni.py +4 -4
processor_omni.py
CHANGED
@@ -618,7 +618,7 @@ class OmniMMProcessor(object):
|
|
618 |
# 本地路径
|
619 |
video_path = video_info['local']
|
620 |
# 帧保存本地路径
|
621 |
-
frame_path = video_path.
|
622 |
mm_obj_byte = self._get_vision_obj_byte('local', video_path)
|
623 |
elif 'base64' in video_info.keys():
|
624 |
md5 = hashlib.md5(video_info['base64'].encode('utf-8')).hexdigest()
|
@@ -626,7 +626,7 @@ class OmniMMProcessor(object):
|
|
626 |
video_path = os.path.join(self.relative_path, md5)
|
627 |
else:
|
628 |
video_path = os.path.join(os.getcwd(), md5)
|
629 |
-
frame_path =
|
630 |
mm_obj_byte = self._get_vision_obj_byte('base64', video_info['base64'])
|
631 |
elif 'url' in video_info.keys():
|
632 |
md5 = hashlib.md5(video_info['url'].encode('utf-8')).hexdigest()
|
@@ -634,7 +634,7 @@ class OmniMMProcessor(object):
|
|
634 |
video_path = os.path.join(self.relative_path, md5)
|
635 |
else:
|
636 |
video_path = os.path.join(os.getcwd(), md5)
|
637 |
-
frame_path =
|
638 |
mm_obj_byte = self._get_vision_obj_byte('url', video_info['url'])
|
639 |
else:
|
640 |
raise ValueError('unvalid video server !!!')
|
@@ -778,7 +778,7 @@ class OmniMMProcessor(object):
|
|
778 |
if len(ret.input_ids) > self.tokenizer.model_max_length-1: # 过滤长文本
|
779 |
raise ValueError(f"Text too long, please check text length! 【{text[:5]+'...'*6+text[-5:]}】")
|
780 |
else:
|
781 |
-
raise ValueError(f"mm_label not supportted! must in ['audio', 'image', 'text'] but get {mm_label}")
|
782 |
return ret
|
783 |
|
784 |
def process_one(self, text, index=0, raw_only=False):
|
|
|
618 |
# 本地路径
|
619 |
video_path = video_info['local']
|
620 |
# 帧保存本地路径
|
621 |
+
frame_path = video_path[:video_path.rfind('.')] + frame_suffix
|
622 |
mm_obj_byte = self._get_vision_obj_byte('local', video_path)
|
623 |
elif 'base64' in video_info.keys():
|
624 |
md5 = hashlib.md5(video_info['base64'].encode('utf-8')).hexdigest()
|
|
|
626 |
video_path = os.path.join(self.relative_path, md5)
|
627 |
else:
|
628 |
video_path = os.path.join(os.getcwd(), md5)
|
629 |
+
frame_path = video_path + frame_suffix
|
630 |
mm_obj_byte = self._get_vision_obj_byte('base64', video_info['base64'])
|
631 |
elif 'url' in video_info.keys():
|
632 |
md5 = hashlib.md5(video_info['url'].encode('utf-8')).hexdigest()
|
|
|
634 |
video_path = os.path.join(self.relative_path, md5)
|
635 |
else:
|
636 |
video_path = os.path.join(os.getcwd(), md5)
|
637 |
+
frame_path = video_path + frame_suffix
|
638 |
mm_obj_byte = self._get_vision_obj_byte('url', video_info['url'])
|
639 |
else:
|
640 |
raise ValueError('unvalid video server !!!')
|
|
|
778 |
if len(ret.input_ids) > self.tokenizer.model_max_length-1: # 过滤长文本
|
779 |
raise ValueError(f"Text too long, please check text length! 【{text[:5]+'...'*6+text[-5:]}】")
|
780 |
else:
|
781 |
+
raise ValueError(f"mm_label not supportted! must in ['audio', 'audiogen', 'image', 'video', 'text'] but get {mm_label}")
|
782 |
return ret
|
783 |
|
784 |
def process_one(self, text, index=0, raw_only=False):
|