Spaces:
Runtime error
Runtime error
Commit
•
bfcdf7c
1
Parent(s):
749e38e
Update app.py
Browse files
app.py
CHANGED
@@ -35,47 +35,47 @@ def parse_codeblock(text):
|
|
35 |
def predict(inputs, top_p, temperature, chat_counter, chatbot=[], history=[]):
|
36 |
|
37 |
payload = {
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
}
|
47 |
|
48 |
headers = {
|
49 |
-
|
50 |
-
|
51 |
}
|
52 |
|
53 |
# print(f"chat_counter - {chat_counter}")
|
54 |
if chat_counter != 0 :
|
55 |
messages = []
|
56 |
for i, data in enumerate(history):
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
|
66 |
temp3 = {}
|
67 |
temp3["role"] = "user"
|
68 |
temp3["content"] = inputs
|
69 |
messages.append(temp3)
|
70 |
payload = {
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
}
|
80 |
|
81 |
chat_counter+=1
|
@@ -85,39 +85,37 @@ def predict(inputs, top_p, temperature, chat_counter, chatbot=[], history=[]):
|
|
85 |
response = requests.post(API_URL, headers=headers, json=payload, stream=True)
|
86 |
response_code = f"{response}"
|
87 |
if response_code.strip() != "<Response [200]>":
|
88 |
-
#print(f"response code - {response}")
|
89 |
raise Exception(f"Sorry, hitting rate limit. Please try again later. {response}")
|
90 |
token_counter = 0
|
91 |
partial_words = ""
|
92 |
-
counter=0
|
93 |
for chunk in response.iter_lines():
|
94 |
#Skipping first chunk
|
95 |
if counter == 0:
|
96 |
counter+=1
|
97 |
continue
|
98 |
-
#counter+=1
|
99 |
# check whether each line is non-empty
|
100 |
if chunk.decode() :
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
print(json.dumps({"chat_counter": chat_counter, "payload": payload, "partial_words": partial_words, "token_counter": token_counter, "counter": counter}))
|
115 |
|
116 |
|
117 |
def reset_textbox():
|
118 |
return gr.update(value='')
|
119 |
|
120 |
-
title = """<h1 align="center">
|
121 |
if DISABLED:
|
122 |
title = """<h1 align="center" style="color:red">This app has reached OpenAI's usage limit. We are currently requesting an increase in our quota. Please check back in a few days.</h1>"""
|
123 |
description = """Language models can be conditioned to act like dialogue agents through a conversational prompt that typically takes the form:
|
@@ -137,7 +135,7 @@ with gr.Blocks(css = """#col_container { margin-left: auto; margin-right: auto;}
|
|
137 |
#chatbot {height: 520px; overflow: auto;}""",
|
138 |
theme=theme) as demo:
|
139 |
gr.HTML(title)
|
140 |
-
gr.HTML("""<h3 align="center">This App provides you full access to
|
141 |
gr.HTML('''<center><a href="https://huggingface.co/spaces/yuntian-deng/ChatGPT?duplicate=true"><img src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>Duplicate the Space and run securely with your OpenAI API Key</center>''')
|
142 |
with gr.Column(elem_id = "col_container", visible=False) as main_block:
|
143 |
#API Key is provided by OpenAI
|
|
|
35 |
def predict(inputs, top_p, temperature, chat_counter, chatbot=[], history=[]):
|
36 |
|
37 |
payload = {
|
38 |
+
"model": MODEL,
|
39 |
+
"messages": [{"role": "user", "content": f"{inputs}"}],
|
40 |
+
"temperature" : 1.0,
|
41 |
+
"top_p":1.0,
|
42 |
+
"n" : 1,
|
43 |
+
"stream": True,
|
44 |
+
"presence_penalty":0,
|
45 |
+
"frequency_penalty":0,
|
46 |
}
|
47 |
|
48 |
headers = {
|
49 |
+
"Content-Type": "application/json",
|
50 |
+
"Authorization": f"Bearer {OPENAI_API_KEY}"
|
51 |
}
|
52 |
|
53 |
# print(f"chat_counter - {chat_counter}")
|
54 |
if chat_counter != 0 :
|
55 |
messages = []
|
56 |
for i, data in enumerate(history):
|
57 |
+
if i % 2 == 0:
|
58 |
+
role = 'user'
|
59 |
+
else:
|
60 |
+
role = 'assistant'
|
61 |
+
temp = {}
|
62 |
+
temp["role"] = role
|
63 |
+
temp["content"] = data
|
64 |
+
messages.append(temp)
|
65 |
|
66 |
temp3 = {}
|
67 |
temp3["role"] = "user"
|
68 |
temp3["content"] = inputs
|
69 |
messages.append(temp3)
|
70 |
payload = {
|
71 |
+
"model": MODEL,
|
72 |
+
"messages": messages,
|
73 |
+
"temperature" : temperature,
|
74 |
+
"top_p": top_p,
|
75 |
+
"n" : 1,
|
76 |
+
"stream": True,
|
77 |
+
"presence_penalty":0,
|
78 |
+
"frequency_penalty":0,
|
79 |
}
|
80 |
|
81 |
chat_counter+=1
|
|
|
85 |
response = requests.post(API_URL, headers=headers, json=payload, stream=True)
|
86 |
response_code = f"{response}"
|
87 |
if response_code.strip() != "<Response [200]>":
|
|
|
88 |
raise Exception(f"Sorry, hitting rate limit. Please try again later. {response}")
|
89 |
token_counter = 0
|
90 |
partial_words = ""
|
91 |
+
counter = 0
|
92 |
for chunk in response.iter_lines():
|
93 |
#Skipping first chunk
|
94 |
if counter == 0:
|
95 |
counter+=1
|
96 |
continue
|
|
|
97 |
# check whether each line is non-empty
|
98 |
if chunk.decode() :
|
99 |
+
chunk = chunk.decode()
|
100 |
+
# decode each line as response data is in bytes
|
101 |
+
if len(chunk) > 12 and "content" in json.loads(chunk[6:])['choices'][0]['delta']:
|
102 |
+
#if len(json.loads(chunk.decode()[6:])['choices'][0]["delta"]) == 0:
|
103 |
+
# break
|
104 |
+
partial_words = partial_words + json.loads(chunk[6:])['choices'][0]["delta"]["content"]
|
105 |
+
if token_counter == 0:
|
106 |
+
history.append(" " + partial_words)
|
107 |
+
else:
|
108 |
+
history[-1] = partial_words
|
109 |
+
chat = [(parse_codeblock(history[i]), parse_codeblock(history[i + 1])) for i in range(0, len(history) - 1, 2) ] # convert to tuples of list
|
110 |
+
token_counter+=1
|
111 |
+
yield chat, history, chat_counter, response # resembles {chatbot: chat, state: history}
|
112 |
print(json.dumps({"chat_counter": chat_counter, "payload": payload, "partial_words": partial_words, "token_counter": token_counter, "counter": counter}))
|
113 |
|
114 |
|
115 |
def reset_textbox():
|
116 |
return gr.update(value='')
|
117 |
|
118 |
+
title = """<h1 align="center">GPT-3.5 Chatbot</h1>"""
|
119 |
if DISABLED:
|
120 |
title = """<h1 align="center" style="color:red">This app has reached OpenAI's usage limit. We are currently requesting an increase in our quota. Please check back in a few days.</h1>"""
|
121 |
description = """Language models can be conditioned to act like dialogue agents through a conversational prompt that typically takes the form:
|
|
|
135 |
#chatbot {height: 520px; overflow: auto;}""",
|
136 |
theme=theme) as demo:
|
137 |
gr.HTML(title)
|
138 |
+
gr.HTML("""<h3 align="center">This App provides you full access to GPT-3.5 (4096 token limit). You don't need any OPENAI API key.</h1>""")
|
139 |
gr.HTML('''<center><a href="https://huggingface.co/spaces/yuntian-deng/ChatGPT?duplicate=true"><img src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>Duplicate the Space and run securely with your OpenAI API Key</center>''')
|
140 |
with gr.Column(elem_id = "col_container", visible=False) as main_block:
|
141 |
#API Key is provided by OpenAI
|