Microsoft's Phi3-Vision-128k-Context
Phi-3-Vision is a 4.2B parameter multimodal model that brings together language and vision capabilities.
"""
@spaces.GPU
def bot_streaming(message, history):
print(f'message is - {message}')
print(f'history is - {history}')
if message["files"]:
# message["files"][-1] is a Dict or just a string
if type(message["files"][-1]) == dict:
image = message["files"][-1]["path"]
else:
image = message["files"][-1]
else:
# if there's no image uploaded for this turn, look for images in the past turns
# kept inside tuples, take the last one
for hist in history:
if type(hist[0]) == tuple:
image = hist[0][0]
try:
if image is None:
# Handle the case where image is None
raise gr.Error("You need to upload an image for Phi3-Vision to work. Close the error and try again with an Image.")
except NameError:
# Handle the case where 'image' is not defined at all
raise gr.Error("You need to upload an image for Phi3-Vision to work. Close the error and try again with an Image.")
conversation = []
flag=False
for user, assistant in history:
if assistant is None:
#pass
flag=True
conversation.extend([{"role": "user", "content":""}])
continue
if flag==True:
conversation[0]['content'] = f"<|image_1|>\n{user}"
conversation.extend([{"role": "assistant", "content": assistant}])
flag=False
continue
conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
if len(history) == 0:
conversation.append({"role": "user", "content": f"<|image_1|>\n{message['text']}"})
else:
conversation.append({"role": "user", "content": message['text']})
print(f"prompt is -\n{conversation}")
#prompt = f"""User: