Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -18,84 +18,23 @@ PLACEHOLDER = """
|
|
18 |
</div>
|
19 |
"""
|
20 |
|
21 |
-
model_id_llama3 = "xtuner/llava-llama-3-8b-v1_1-transformers"
|
22 |
-
model_id_phi3 = "xtuner/llava-phi-3-mini-hf"
|
23 |
|
24 |
-
|
25 |
-
processor = AutoProcessor.from_pretrained(model_id_phi3)
|
26 |
|
27 |
-
|
28 |
-
model_id_llama3,
|
29 |
-
torch_dtype=torch.float16,
|
30 |
-
low_cpu_mem_usage=True,
|
31 |
-
)
|
32 |
-
model_llama3.to("cuda:0")
|
33 |
-
model_llama3.generation_config.eos_token_id = 128009
|
34 |
|
35 |
-
|
36 |
-
|
37 |
torch_dtype=torch.float16,
|
38 |
low_cpu_mem_usage=True,
|
39 |
)
|
40 |
-
model_phi3.to("cuda:0")
|
41 |
-
model_phi3.generation_config.eos_token_id = 128009
|
42 |
|
43 |
-
|
44 |
-
|
45 |
-
def bot_streaming_llama3(message, history):
|
46 |
-
print(message)
|
47 |
-
if message["files"]:
|
48 |
-
# message["files"][-1] is a Dict or just a string
|
49 |
-
if type(message["files"][-1]) == dict:
|
50 |
-
image = message["files"][-1]["path"]
|
51 |
-
else:
|
52 |
-
image = message["files"][-1]
|
53 |
-
else:
|
54 |
-
# if there's no image uploaded for this turn, look for images in the past turns
|
55 |
-
# kept inside tuples, take the last one
|
56 |
-
for hist in history:
|
57 |
-
if type(hist[0]) == tuple:
|
58 |
-
image = hist[0][0]
|
59 |
-
try:
|
60 |
-
if image is None:
|
61 |
-
# Handle the case where image is None
|
62 |
-
gr.Error("You need to upload an image for LLaVA to work.")
|
63 |
-
except NameError:
|
64 |
-
# Handle the case where 'image' is not defined at all
|
65 |
-
gr.Error("You need to upload an image for LLaVA to work.")
|
66 |
-
|
67 |
-
prompt = f"<|start_header_id|>user<|end_header_id|>\n\n<image>\n{message['text']}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"
|
68 |
-
# print(f"prompt: {prompt}")
|
69 |
-
image = Image.open(image)
|
70 |
-
inputs = processor(prompt, image, return_tensors='pt').to(0, torch.float16)
|
71 |
-
|
72 |
-
streamer = TextIteratorStreamer(processor, **{"skip_special_tokens": False, "skip_prompt": True})
|
73 |
-
generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=1024, do_sample=False)
|
74 |
-
|
75 |
-
thread = Thread(target=model_llama3.generate, kwargs=generation_kwargs)
|
76 |
-
thread.start()
|
77 |
-
|
78 |
-
text_prompt = f"<|start_header_id|>user<|end_header_id|>\n\n{message['text']}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"
|
79 |
-
# print(f"text_prompt: {text_prompt}")
|
80 |
-
|
81 |
-
buffer = ""
|
82 |
-
time.sleep(0.5)
|
83 |
-
for new_text in streamer:
|
84 |
-
# find <|eot_id|> and remove it from the new_text
|
85 |
-
if "<|eot_id|>" in new_text:
|
86 |
-
new_text = new_text.split("<|eot_id|>")[0]
|
87 |
-
buffer += new_text
|
88 |
-
|
89 |
-
# generated_text_without_prompt = buffer[len(text_prompt):]
|
90 |
-
generated_text_without_prompt = buffer
|
91 |
-
# print(generated_text_without_prompt)
|
92 |
-
time.sleep(0.06)
|
93 |
-
# print(f"new_text: {generated_text_without_prompt}")
|
94 |
-
yield generated_text_without_prompt
|
95 |
|
96 |
|
97 |
@spaces.GPU
|
98 |
-
def
|
99 |
print(message)
|
100 |
if message["files"]:
|
101 |
# message["files"][-1] is a Dict or just a string
|
@@ -125,7 +64,7 @@ def bot_streaming_phi3(message, history):
|
|
125 |
streamer = TextIteratorStreamer(processor, **{"skip_special_tokens": False, "skip_prompt": True})
|
126 |
generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=1024, do_sample=False)
|
127 |
|
128 |
-
thread = Thread(target=
|
129 |
thread.start()
|
130 |
|
131 |
text_prompt = f"<|start_header_id|>user<|end_header_id|>\n\n{message['text']}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"
|
@@ -147,57 +86,20 @@ def bot_streaming_phi3(message, history):
|
|
147 |
yield generated_text_without_prompt
|
148 |
|
149 |
|
150 |
-
|
151 |
-
|
152 |
-
#chatbot=gr.Chatbot(placeholder=PLACEHOLDER,scale=1)
|
153 |
-
#chat_input = gr.MultimodalTextbox(interactive=True, file_types=["image"], placeholder="Enter message or upload file...", show_label=False)
|
154 |
-
|
155 |
with gr.Blocks(fill_height=True, ) as demo:
|
156 |
-
|
157 |
-
|
158 |
-
|
159 |
-
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
|
166 |
-
bubble_full_width=False,
|
167 |
-
label='LLaVa-Phi3'
|
168 |
)
|
169 |
-
|
170 |
-
chat_input = gr.MultimodalTextbox(interactive=True, file_types=["image"], placeholder="Enter message or upload file...", show_label=False)
|
171 |
-
|
172 |
-
gr.Examples(examples=[{"text": "What is on the flower?", "files": ["./bee.png"]},
|
173 |
-
{"text": "How to make this pastry?", "files": ["./baklava.png"]},],
|
174 |
-
inputs=chat_input)
|
175 |
-
|
176 |
-
#chat_input.submit(lambda: gr.MultimodalTextbox(interactive=False), None, [chat_input]).then(bot_streaming_llama3, [chat_input, chatbot1,], [chatbot1,])
|
177 |
-
|
178 |
-
chat_msg1 = chat_input.submit(bot_streaming_llama3, [chat_input, chatbot1,], [chatbot1,])
|
179 |
-
chat_msg2 = chat_input.submit(bot_streaming_phi3, [chat_input, chatbot2,], [chatbot2,])
|
180 |
-
|
181 |
-
#bot_msg1 = chat_msg1.then(bot, chatbot1, chatbot1, api_name="bot_response1")
|
182 |
-
#chat_msg1.then(lambda: gr.MultimodalTextbox(interactive=True), None, [chat_input])
|
183 |
-
#bot_msg2 = chat_msg2.then(bot, chatbot2, chatbot2, api_name="bot_response2")
|
184 |
-
#bot_msg2.then(lambda: gr.MultimodalTextbox(interactive=True), None, [chat_input])
|
185 |
-
|
186 |
-
chatbot1.like(print_like_dislike, None, None)
|
187 |
-
chatbot2.like(print_like_dislike, None, None)
|
188 |
-
|
189 |
-
|
190 |
-
#gr.ChatInterface(
|
191 |
-
#fn=bot_streaming_llama3,
|
192 |
-
#title="LLaVA Llama-3-8B",
|
193 |
-
#examples=[{"text": "What is on the flower?", "files": ["./bee.jpg"]},
|
194 |
-
# {"text": "How to make this pastry?", "files": ["./baklava.png"]}],
|
195 |
-
#description="Try [LLaVA Llama-3-8B](https://huggingface.co/xtuner/llava-llama-3-8b-v1_1-transformers). Upload an image and start chatting about it, or simply try one of the examples below. If you don't upload an image, you will receive an error.",
|
196 |
-
#stop_btn="Stop Generation",
|
197 |
-
#multimodal=True,
|
198 |
-
#textbox=chat_input,
|
199 |
-
#chatbot=chatbot,
|
200 |
-
#)
|
201 |
|
202 |
demo.queue(api_open=False)
|
203 |
-
demo.launch(show_api=False, share=False)
|
|
|
18 |
</div>
|
19 |
"""
|
20 |
|
|
|
|
|
21 |
|
22 |
+
model_id = "xtuner/llava-llama-3-8b-v1_1-transformers"
|
|
|
23 |
|
24 |
+
processor = AutoProcessor.from_pretrained(model_id)
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
|
26 |
+
model = LlavaForConditionalGeneration.from_pretrained(
|
27 |
+
model_id,
|
28 |
torch_dtype=torch.float16,
|
29 |
low_cpu_mem_usage=True,
|
30 |
)
|
|
|
|
|
31 |
|
32 |
+
model.to("cuda:0")
|
33 |
+
model.generation_config.eos_token_id = 128009
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
34 |
|
35 |
|
36 |
@spaces.GPU
|
37 |
+
def bot_streaming(message, history):
|
38 |
print(message)
|
39 |
if message["files"]:
|
40 |
# message["files"][-1] is a Dict or just a string
|
|
|
64 |
streamer = TextIteratorStreamer(processor, **{"skip_special_tokens": False, "skip_prompt": True})
|
65 |
generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=1024, do_sample=False)
|
66 |
|
67 |
+
thread = Thread(target=model.generate, kwargs=generation_kwargs)
|
68 |
thread.start()
|
69 |
|
70 |
text_prompt = f"<|start_header_id|>user<|end_header_id|>\n\n{message['text']}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"
|
|
|
86 |
yield generated_text_without_prompt
|
87 |
|
88 |
|
89 |
+
chatbot=gr.Chatbot(placeholder=PLACEHOLDER,scale=1)
|
90 |
+
chat_input = gr.MultimodalTextbox(interactive=True, file_types=["image"], placeholder="Enter message or upload file...", show_label=False)
|
|
|
|
|
|
|
91 |
with gr.Blocks(fill_height=True, ) as demo:
|
92 |
+
gr.ChatInterface(
|
93 |
+
fn=bot_streaming,
|
94 |
+
title="LLaVA Llama-3-8B",
|
95 |
+
examples=[{"text": "What is on the flower?", "files": ["./bee.jpg"]},
|
96 |
+
{"text": "How to make this pastry?", "files": ["./baklava.png"]}],
|
97 |
+
description="Try [LLaVA Llama-3-8B](https://huggingface.co/xtuner/llava-llama-3-8b-v1_1-transformers). Upload an image and start chatting about it, or simply try one of the examples below. If you don't upload an image, you will receive an error.",
|
98 |
+
stop_btn="Stop Generation",
|
99 |
+
multimodal=True,
|
100 |
+
textbox=chat_input,
|
101 |
+
chatbot=chatbot,
|
|
|
|
|
102 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
103 |
|
104 |
demo.queue(api_open=False)
|
105 |
+
demo.launch(show_api=False, share=False)
|