Spaces:
Runtime error
Runtime error
mateoluksenberg
commited on
Commit
•
b250460
1
Parent(s):
46a2ec1
Update app.py
Browse files
app.py
CHANGED
@@ -36,9 +36,12 @@ h1 {
|
|
36 |
|
37 |
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
|
38 |
|
|
|
|
|
39 |
def extract_text(path):
|
40 |
return open(path, 'r').read()
|
41 |
|
|
|
42 |
def extract_pdf(path):
|
43 |
doc = pymupdf.open(path)
|
44 |
text = ""
|
@@ -46,6 +49,7 @@ def extract_pdf(path):
|
|
46 |
text += page.get_text()
|
47 |
return text
|
48 |
|
|
|
49 |
def extract_docx(path):
|
50 |
doc = docx.Document(path)
|
51 |
data = []
|
@@ -54,6 +58,7 @@ def extract_docx(path):
|
|
54 |
content = '\n\n'.join(data)
|
55 |
return content
|
56 |
|
|
|
57 |
def extract_pptx(path):
|
58 |
prs = Presentation(path)
|
59 |
text = ""
|
@@ -63,6 +68,7 @@ def extract_pptx(path):
|
|
63 |
text += shape.text + "\n"
|
64 |
return text
|
65 |
|
|
|
66 |
def mode_load(path):
|
67 |
choice = ""
|
68 |
file_type = path.split(".")[-1]
|
@@ -79,13 +85,17 @@ def mode_load(path):
|
|
79 |
choice = "doc"
|
80 |
print(content[:100])
|
81 |
return choice, content[:5000]
|
|
|
|
|
82 |
elif file_type in ["png", "jpg", "jpeg", "bmp", "tiff", "webp"]:
|
83 |
content = Image.open(path).convert('RGB')
|
84 |
choice = "image"
|
85 |
return choice, content
|
|
|
86 |
else:
|
87 |
raise gr.Error("Oops, unsupported files.")
|
88 |
|
|
|
89 |
@spaces.GPU()
|
90 |
def stream_chat(message, history: list, temperature: float, max_length: int, top_p: float, top_k: int, penalty: float):
|
91 |
|
@@ -100,9 +110,8 @@ def stream_chat(message, history: list, temperature: float, max_length: int, top
|
|
100 |
print(f'history is - {history}')
|
101 |
conversation = []
|
102 |
prompt_files = []
|
103 |
-
|
104 |
-
|
105 |
-
choice, contents = mode_load(message["files"][-1].path)
|
106 |
if choice == "image":
|
107 |
conversation.append({"role": "user", "image": contents, "content": message['text']})
|
108 |
elif choice == "doc":
|
@@ -110,9 +119,11 @@ def stream_chat(message, history: list, temperature: float, max_length: int, top
|
|
110 |
conversation.append({"role": "user", "content": format_msg})
|
111 |
else:
|
112 |
if len(history) == 0:
|
|
|
113 |
contents = None
|
114 |
conversation.append({"role": "user", "content": message['text']})
|
115 |
else:
|
|
|
116 |
for prompt, answer in history:
|
117 |
if answer is None:
|
118 |
prompt_files.append(prompt[0])
|
@@ -120,17 +131,17 @@ def stream_chat(message, history: list, temperature: float, max_length: int, top
|
|
120 |
else:
|
121 |
conversation.extend([{"role": "user", "content": prompt}, {"role": "assistant", "content": answer}])
|
122 |
if len(prompt_files) > 0:
|
123 |
-
choice, contents = mode_load(prompt_files[-1]
|
124 |
else:
|
125 |
choice = ""
|
126 |
conversation.append({"role": "user", "image": "", "content": message['text']})
|
127 |
|
|
|
128 |
if choice == "image":
|
129 |
conversation.append({"role": "user", "image": contents, "content": message['text']})
|
130 |
elif choice == "doc":
|
131 |
format_msg = contents + "\n\n\n" + "{} files uploaded.\n" + message['text']
|
132 |
conversation.append({"role": "user", "content": format_msg})
|
133 |
-
|
134 |
print(f"Conversation is -\n{conversation}")
|
135 |
|
136 |
input_ids = tokenizer.apply_chat_template(conversation, tokenize=True, add_generation_prompt=True,
|
@@ -157,14 +168,19 @@ def stream_chat(message, history: list, temperature: float, max_length: int, top
|
|
157 |
buffer += new_text
|
158 |
yield buffer
|
159 |
|
160 |
-
chatbot = gr.Chatbot()
|
161 |
|
|
|
|
|
|
|
162 |
chat_input = gr.MultimodalTextbox(
|
163 |
interactive=True,
|
164 |
placeholder="Enter message or upload a file ...",
|
165 |
show_label=False,
|
166 |
-
|
|
|
167 |
|
|
|
|
|
168 |
EXAMPLES = [
|
169 |
[{"text": "Write a poem about spring season in French Language", }],
|
170 |
[{"text": "what does this chart mean?", "files": ["sales.png"]}],
|
@@ -178,6 +194,8 @@ with gr.Blocks(css=CSS, theme="soft", fill_height=True) as demo:
|
|
178 |
gr.ChatInterface(
|
179 |
fn=stream_chat,
|
180 |
multimodal=True,
|
|
|
|
|
181 |
textbox=chat_input,
|
182 |
chatbot=chatbot,
|
183 |
fill_height=True,
|
@@ -228,4 +246,5 @@ with gr.Blocks(css=CSS, theme="soft", fill_height=True) as demo:
|
|
228 |
gr.Examples(EXAMPLES, [chat_input])
|
229 |
|
230 |
if __name__ == "__main__":
|
231 |
-
|
|
|
|
36 |
|
37 |
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
|
38 |
|
39 |
+
|
40 |
+
|
41 |
def extract_text(path):
|
42 |
return open(path, 'r').read()
|
43 |
|
44 |
+
|
45 |
def extract_pdf(path):
|
46 |
doc = pymupdf.open(path)
|
47 |
text = ""
|
|
|
49 |
text += page.get_text()
|
50 |
return text
|
51 |
|
52 |
+
|
53 |
def extract_docx(path):
|
54 |
doc = docx.Document(path)
|
55 |
data = []
|
|
|
58 |
content = '\n\n'.join(data)
|
59 |
return content
|
60 |
|
61 |
+
|
62 |
def extract_pptx(path):
|
63 |
prs = Presentation(path)
|
64 |
text = ""
|
|
|
68 |
text += shape.text + "\n"
|
69 |
return text
|
70 |
|
71 |
+
|
72 |
def mode_load(path):
|
73 |
choice = ""
|
74 |
file_type = path.split(".")[-1]
|
|
|
85 |
choice = "doc"
|
86 |
print(content[:100])
|
87 |
return choice, content[:5000]
|
88 |
+
|
89 |
+
|
90 |
elif file_type in ["png", "jpg", "jpeg", "bmp", "tiff", "webp"]:
|
91 |
content = Image.open(path).convert('RGB')
|
92 |
choice = "image"
|
93 |
return choice, content
|
94 |
+
|
95 |
else:
|
96 |
raise gr.Error("Oops, unsupported files.")
|
97 |
|
98 |
+
|
99 |
@spaces.GPU()
|
100 |
def stream_chat(message, history: list, temperature: float, max_length: int, top_p: float, top_k: int, penalty: float):
|
101 |
|
|
|
110 |
print(f'history is - {history}')
|
111 |
conversation = []
|
112 |
prompt_files = []
|
113 |
+
if message["files"]:
|
114 |
+
choice, contents = mode_load(message["files"][-1])
|
|
|
115 |
if choice == "image":
|
116 |
conversation.append({"role": "user", "image": contents, "content": message['text']})
|
117 |
elif choice == "doc":
|
|
|
119 |
conversation.append({"role": "user", "content": format_msg})
|
120 |
else:
|
121 |
if len(history) == 0:
|
122 |
+
# raise gr.Error("Please upload an image first.")
|
123 |
contents = None
|
124 |
conversation.append({"role": "user", "content": message['text']})
|
125 |
else:
|
126 |
+
# image = Image.open(history[0][0][0])
|
127 |
for prompt, answer in history:
|
128 |
if answer is None:
|
129 |
prompt_files.append(prompt[0])
|
|
|
131 |
else:
|
132 |
conversation.extend([{"role": "user", "content": prompt}, {"role": "assistant", "content": answer}])
|
133 |
if len(prompt_files) > 0:
|
134 |
+
choice, contents = mode_load(prompt_files[-1])
|
135 |
else:
|
136 |
choice = ""
|
137 |
conversation.append({"role": "user", "image": "", "content": message['text']})
|
138 |
|
139 |
+
|
140 |
if choice == "image":
|
141 |
conversation.append({"role": "user", "image": contents, "content": message['text']})
|
142 |
elif choice == "doc":
|
143 |
format_msg = contents + "\n\n\n" + "{} files uploaded.\n" + message['text']
|
144 |
conversation.append({"role": "user", "content": format_msg})
|
|
|
145 |
print(f"Conversation is -\n{conversation}")
|
146 |
|
147 |
input_ids = tokenizer.apply_chat_template(conversation, tokenize=True, add_generation_prompt=True,
|
|
|
168 |
buffer += new_text
|
169 |
yield buffer
|
170 |
|
|
|
171 |
|
172 |
+
chatbot = gr.Chatbot(
|
173 |
+
#rtl=True,
|
174 |
+
)
|
175 |
chat_input = gr.MultimodalTextbox(
|
176 |
interactive=True,
|
177 |
placeholder="Enter message or upload a file ...",
|
178 |
show_label=False,
|
179 |
+
#rtl=True,
|
180 |
+
|
181 |
|
182 |
+
|
183 |
+
)
|
184 |
EXAMPLES = [
|
185 |
[{"text": "Write a poem about spring season in French Language", }],
|
186 |
[{"text": "what does this chart mean?", "files": ["sales.png"]}],
|
|
|
194 |
gr.ChatInterface(
|
195 |
fn=stream_chat,
|
196 |
multimodal=True,
|
197 |
+
|
198 |
+
|
199 |
textbox=chat_input,
|
200 |
chatbot=chatbot,
|
201 |
fill_height=True,
|
|
|
246 |
gr.Examples(EXAMPLES, [chat_input])
|
247 |
|
248 |
if __name__ == "__main__":
|
249 |
+
|
250 |
+
demo.queue(api_open=False).launch(show_api=False, share=False, )#server_name="0.0.0.0", )
|