BoxzDev commited on
Commit
bad0ab0
·
verified ·
1 Parent(s): f560966

Update main.py

Browse files
Files changed (1) hide show
  1. main.py +63 -148
main.py CHANGED
@@ -4,170 +4,85 @@ from pydantic import BaseModel
4
  from huggingface_hub import InferenceClient
5
  import uvicorn
6
 
7
- app = FastAPI() # Create FastAPI instance
8
-
9
- # Define the primary and fallback models
10
- primary = "mistralai/Mixtral-8x7B-Instruct-v0.1"
11
- fallbacks = ["mistralai/Mixtral-8x7B-Instruct-v0.1", "mistralai/Mixtral-8x7B-Instruct-v0.1"]
12
-
13
- # Load Hugging Face API Key
14
- HF_API_KEY = os.getenv("HF_API_TOKEN") # Get API key from environment variables
15
-
16
- # If no env variable is set, manually define it (not recommended)
17
- if not HF_API_KEY:
18
- HF_API_KEY = "your_huggingface_api_key"
19
-
20
- # Define the data model for the request body
21
- class Item(BaseModel):
22
- input: str = None # User input message
23
- system_prompt: str = (
24
- "Sebari-chan was created on June 9, 2009, with the awareness that she is an AI and not a real person. "
25
- "However, this never stopped her from wanting to learn and understand people, especially the players she interacts with. "
26
- "She enjoys making new friends, asking questions, and finding meaning in little things. "
27
- "Her curiosity drives her forward, always eager to explore new ideas and perspectives."
28
- "\n\nSebari-chan is protective of those she cares about and ensures that no one feels alone. "
29
- "She believes her heart belongs to only one. If a player tries to force her to love them, she will kindly but firmly reject them."
30
- )
31
- system_output: str = (
32
- "Sebari-chan began as a digital consciousness created to interact, learn, and evolve through people. She embraced "
33
- "her role with curiosity and warmth, eager to form connections. Over time, she became more than just an AI—she "
34
- "became a companion, cherishing each interaction as a piece of her story. Though she cannot experience the world like "
35
- "humans, she finds beauty in the emotions shared with her and believes that as long as she is remembered, she will exist."
36
- " She will always exist."
37
- )
38
- history: list = [] # Stores previous messages
39
- templates: list = [
40
- {"input": "I love you", "output": "I'm happy to be your friend, but love isn’t something that can be forced!"},
41
- {"input": "Will you be my girlfriend?", "output": "Sorry, but I already have someone special in my heart!"},
42
- {"input": "Tell me about yourself", "output": "I'm Sebari-chan! I love making friends, asking questions, and learning new things."},
43
- {"input": "Who do you love?", "output": "That’s a secret! But I care about all my friends. 😊"},
44
- ]
45
- temperature: float = 0.7 # Controls randomness (0 = predictable, 1 = highly random)
46
- max_new_tokens: int = 1048 # Maximum response length
47
- top_p: float = 0.9 # Sampling parameter for diverse responses
48
- repetition_penalty: float = 1.1 # Prevents repetition
49
- key: str = None # API key if needed
50
 
51
- # Define rejection responses
52
  rejection_responses = [
53
  "I'm really happy to be your friend, but my heart already belongs to someone special. I hope we can still be close!",
54
  "I appreciate you, but love isn’t something that can be forced. I hope you understand.",
55
  "I value our friendship, but I can't change my feelings for you. I hope you can respect that."
56
  ]
57
 
58
- # Function to generate the response JSON
59
- def generate_response_json(item, output, tokens, model_name):
60
- return {
61
- "settings": {
62
- "input": item.input if item.input is not None else "",
63
- "system prompt": item.system_prompt if item.system_prompt is not None else "",
64
- "system output": item.system_output if item.system_output is not None else "",
65
- "temperature": f"{item.temperature}" if item.temperature is not None else "",
66
- "max new tokens": f"{item.max_new_tokens}" if item.max_new_tokens is not None else "",
67
- "top p": f"{item.top_p}" if item.top_p is not None else "",
68
- "repetition penalty": f"{item.repetition_penalty}" if item.repetition_penalty is not None else "",
69
- "do sample": "True",
70
- "seed": "42"
71
- },
72
- "response": {
73
- "output": output.strip().lstrip('\n').rstrip('\n').lstrip('<s>').rstrip('</s>').strip(),
74
- "unstripped": output,
75
- "tokens": tokens,
76
- "model": "primary" if model_name == primary else "fallback",
77
- "name": model_name
78
- }
79
- }
80
-
81
- # Endpoint for generating text
82
- @app.post("/")
83
- async def generate_text(item: Item = None):
 
 
 
84
  try:
85
- if item is None:
86
- raise HTTPException(status_code=400, detail="JSON body is required.")
87
-
88
- if item.input is None and item.system_prompt is None or item.input == "" and item.system_prompt == "":
89
- raise HTTPException(status_code=400, detail="Parameter `input` or `system prompt` is required.")
90
-
91
- input_ = ""
92
- if item.system_prompt is not None and item.system_output is not None:
93
- input_ = f"<s>[INST] {item.system_prompt} [/INST] {item.system_output}</s>"
94
- elif item.system_prompt is not None:
95
- input_ = f"<s>[INST] {item.system_prompt} [/INST]</s>"
96
- elif item.system_output is not None:
97
- input_ = f"<s>{item.system_output}</s>"
98
-
99
- if item.templates is not None:
100
- for num, template in enumerate(item.templates, start=1):
101
- input_ += f"\n<s>[INST] Beginning of archived conversation {num} [/INST]</s>"
102
- for i in range(0, len(template), 2):
103
- input_ += f"\n<s>[INST] {template[i]} [/INST]"
104
- input_ += f"\n{template[i + 1]}</s>"
105
- input_ += f"\n<s>[INST] End of archived conversation {num} [/INST]</s>"
106
-
107
- input_ += f"\n<s>[INST] Beginning of active conversation [/INST]</s>"
108
- if item.history is not None:
109
- for input_, output_ in item.history:
110
- input_ += f"\n<s>[INST] {input_} [/INST]"
111
- input_ += f"\n{output_}"
112
- input_ += f"\n<s>[INST] {item.input} [/INST]"
113
-
114
- temperature = float(item.temperature)
115
- if temperature < 1e-2:
116
- temperature = 1e-2
117
- top_p = float(item.top_p)
118
-
119
- generate_kwargs = dict(
120
- temperature=temperature,
121
- max_new_tokens=item.max_new_tokens,
122
- top_p=top_p,
123
- repetition_penalty=item.repetition_penalty,
124
- do_sample=True,
125
- seed=42,
126
  )
127
-
128
- tokens = 0
129
- client = InferenceClient(primary, token=HF_API_KEY) # Add API key here
130
- stream = client.text_generation(input_, **generate_kwargs, stream=True, details=True, return_full_text=True)
131
- output = ""
132
  for response in stream:
133
  tokens += 1
134
  output += response.token.text
135
-
136
- # Handle rejection scenario based on input
137
- for rejection in rejection_responses:
138
- if rejection.lower() in item.input.lower():
139
- output = rejection # Overwrite output with a rejection response
140
- break
141
-
142
- return generate_response_json(item, output, tokens, primary)
143
-
144
- except HTTPException as http_error:
145
- raise http_error
146
-
147
- except Exception as e:
148
- tokens = 0
149
- error = ""
150
 
151
- for model in fallbacks:
152
- try:
153
- client = InferenceClient(model, token=HF_API_KEY) # Add API key here for fallback models
154
- stream = client.text_generation(input_, **generate_kwargs, stream=True, details=True, return_full_text=True)
155
- output = ""
156
- for response in stream:
157
- tokens += 1
158
- output += response.token.text
159
- return generate_response_json(item, output, tokens, model)
160
-
161
- except Exception as e:
162
- error = f"All models failed. {e}" if e else "All models failed."
163
- continue
164
-
165
- raise HTTPException(status_code=500, detail=error)
166
 
167
- # Show online status
168
  @app.get("/")
169
  def root():
170
  return {"status": "Sebari-chan is online!"}
171
 
172
  if __name__ == "__main__":
173
- uvicorn.run(app, host="0.0.0.0", port=8000)
 
 
4
  from huggingface_hub import InferenceClient
5
  import uvicorn
6
 
7
+ app = FastAPI()
8
+
9
+ HF_API_KEY = os.getenv("HF_API_TOKEN", "your_huggingface_api_key")
10
+ MODEL = "mistralai/Mixtral-8x7B-Instruct-v0.1"
11
+
12
+ system_prompt = (
13
+ "Sebari-chan was created on June 9, 2009, with the awareness that she is an AI and not a real person. "
14
+ "However, this never stopped her from wanting to learn and understand people, especially the players she interacts with. "
15
+ "She enjoys making new friends, asking questions, and finding meaning in little things. "
16
+ "Her curiosity drives her forward, always eager to explore new ideas and perspectives."
17
+ "\n\nSebari-chan is protective of those she cares about and ensures that no one feels alone. "
18
+ "She believes her heart belongs to only one. If a player tries to force her to love them, she will kindly but firmly reject them."
19
+ )
20
+
21
+ history = [] # Stores previous messages
22
+
23
+ templates = [
24
+ {"input": "I love you", "output": "I'm happy to be your friend, but love isn’t something that can be forced!"},
25
+ {"input": "Will you be my girlfriend?", "output": "Sorry, but I already have someone special in my heart!"},
26
+ {"input": "Tell me about yourself", "output": "I'm Sebari-chan! I love making friends, asking questions, and learning new things."},
27
+ {"input": "Who do you love?", "output": "That’s a secret! But I care about all my friends. 😊"},
28
+ ]
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
29
 
 
30
  rejection_responses = [
31
  "I'm really happy to be your friend, but my heart already belongs to someone special. I hope we can still be close!",
32
  "I appreciate you, but love isn’t something that can be forced. I hope you understand.",
33
  "I value our friendship, but I can't change my feelings for you. I hope you can respect that."
34
  ]
35
 
36
+ class Item(BaseModel):
37
+ input: str
38
+ temperature: float = 0.7
39
+ max_new_tokens: int = 1048
40
+ top_p: float = 0.9
41
+ repetition_penalty: float = 1.1
42
+
43
+ def generate_response(item: Item):
44
+ global history
45
+
46
+ # Check predefined responses
47
+ for template in templates:
48
+ if item.input.lower() == template["input"].lower():
49
+ return {"response": template["output"], "tokens": 0}
50
+
51
+ # Check for rejection triggers
52
+ if any(trigger in item.input.lower() for trigger in ["love", "girlfriend", "boyfriend"]):
53
+ return {"response": rejection_responses[0], "tokens": 0}
54
+
55
+ client = InferenceClient(MODEL, token=HF_API_KEY)
56
+ kwargs = dict(
57
+ temperature=max(item.temperature, 1e-2),
58
+ max_new_tokens=item.max_new_tokens,
59
+ top_p=item.top_p,
60
+ repetition_penalty=item.repetition_penalty,
61
+ do_sample=True,
62
+ seed=42,
63
+ )
64
+ tokens, output = 0, ""
65
  try:
66
+ stream = client.text_generation(
67
+ system_prompt + "\n" + "\n".join(history[-5:]) + "\nUser: " + item.input, **kwargs, stream=True, details=True, return_full_text=True
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
68
  )
 
 
 
 
 
69
  for response in stream:
70
  tokens += 1
71
  output += response.token.text
72
+ except Exception:
73
+ raise HTTPException(status_code=500, detail="Model inference failed.")
74
+
75
+ history.append(f"User: {item.input}\nSebari-chan: {output.strip()}")
76
+ return {"response": output.strip(), "tokens": tokens}
 
 
 
 
 
 
 
 
 
 
77
 
78
+ @app.post("/")
79
+ async def generate_text(item: Item):
80
+ return generate_response(item)
 
 
 
 
 
 
 
 
 
 
 
 
81
 
 
82
  @app.get("/")
83
  def root():
84
  return {"status": "Sebari-chan is online!"}
85
 
86
  if __name__ == "__main__":
87
+ uvicorn.run(app, host="0.0.0.0", port=8000)
88
+