Spaces:
Sleeping
Sleeping
Update vlm.py
Browse files
vlm.py
CHANGED
@@ -49,6 +49,14 @@ def encode_image(image_path):
|
|
49 |
#
|
50 |
# Build messages
|
51 |
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
52 |
def build_messages(message: dict, history: list[dict]):
|
53 |
"""Build messages given message & history from a **multimodal** chat interface.
|
54 |
Args:
|
@@ -77,13 +85,16 @@ def build_messages(message: dict, history: list[dict]):
|
|
77 |
}
|
78 |
)
|
79 |
|
80 |
-
#
|
81 |
-
messages = history
|
|
|
|
|
82 |
messages.append({'role': 'user', 'content': user_content})
|
83 |
logger.info(f"{messages=}")
|
84 |
|
85 |
return messages
|
86 |
|
|
|
87 |
#
|
88 |
# stream response
|
89 |
#
|
|
|
49 |
#
|
50 |
# Build messages
|
51 |
#
|
52 |
+
def normalize_message_content(msg: dict) -> dict:
|
53 |
+
if isinstance(msg.get("content"), str):
|
54 |
+
return {
|
55 |
+
"role": msg["role"],
|
56 |
+
"content": [{"type": "text", "text": msg["content"]}]
|
57 |
+
}
|
58 |
+
return msg
|
59 |
+
|
60 |
def build_messages(message: dict, history: list[dict]):
|
61 |
"""Build messages given message & history from a **multimodal** chat interface.
|
62 |
Args:
|
|
|
85 |
}
|
86 |
)
|
87 |
|
88 |
+
# Normalize existing history content
|
89 |
+
messages = [normalize_message_content(msg) for msg in history]
|
90 |
+
|
91 |
+
# Append new user message
|
92 |
messages.append({'role': 'user', 'content': user_content})
|
93 |
logger.info(f"{messages=}")
|
94 |
|
95 |
return messages
|
96 |
|
97 |
+
|
98 |
#
|
99 |
# stream response
|
100 |
#
|