Spaces:
Running
on
Zero
Running
on
Zero
Create enhance.py
Browse files- enhance.py +82 -0
enhance.py
ADDED
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import time
|
2 |
+
import requests
|
3 |
+
import json
|
4 |
+
|
5 |
+
def generate(message, max_new_tokens=256, temperature=0.9, top_p=0.95, repetition_penalty=1.0):
|
6 |
+
"""
|
7 |
+
Generates an enhanced prompt using the streaming inference mechanism from a Hugging Face API endpoint.
|
8 |
+
This function formats the prompt with a system instruction, sends a streaming request to the API,
|
9 |
+
and yields the accumulated text as tokens are received.
|
10 |
+
|
11 |
+
Parameters:
|
12 |
+
message (str): The user's input prompt.
|
13 |
+
max_new_tokens (int): The maximum number of tokens to generate.
|
14 |
+
temperature (float): Sampling temperature.
|
15 |
+
top_p (float): Nucleus sampling parameter.
|
16 |
+
repetition_penalty (float): Penalty factor for repetition (not used in the payload but kept for API consistency).
|
17 |
+
|
18 |
+
Yields:
|
19 |
+
str: The accumulated generated text as it streams in.
|
20 |
+
"""
|
21 |
+
# Define the system prompt.
|
22 |
+
SYSTEM_PROMPT = (
|
23 |
+
"You are a prompt enhancer and your work is to enhance the given prompt under 100 words "
|
24 |
+
"without changing the essence, only write the enhanced prompt and nothing else."
|
25 |
+
)
|
26 |
+
# Format the prompt with a timestamp for uniqueness.
|
27 |
+
timestamp = time.time()
|
28 |
+
formatted_prompt = (
|
29 |
+
f"<s>[INST] SYSTEM: {SYSTEM_PROMPT} [/INST]"
|
30 |
+
f"[INST] {message} {timestamp} [/INST]"
|
31 |
+
)
|
32 |
+
|
33 |
+
# Define the API endpoint and headers.
|
34 |
+
api_url = "https://ruslanmv-hf-llm-api.hf.space/api/v1/chat/completions"
|
35 |
+
headers = {"Content-Type": "application/json"}
|
36 |
+
|
37 |
+
# Build the payload for the inference request.
|
38 |
+
payload = {
|
39 |
+
"model": "mixtral-8x7b",
|
40 |
+
"messages": [{"role": "user", "content": formatted_prompt}],
|
41 |
+
"temperature": temperature,
|
42 |
+
"top_p": top_p,
|
43 |
+
"max_tokens": max_new_tokens,
|
44 |
+
"use_cache": False,
|
45 |
+
"stream": True
|
46 |
+
}
|
47 |
+
|
48 |
+
try:
|
49 |
+
response = requests.post(api_url, headers=headers, json=payload, stream=True)
|
50 |
+
response.raise_for_status()
|
51 |
+
full_output = ""
|
52 |
+
|
53 |
+
# Process the streaming response line by line.
|
54 |
+
for line in response.iter_lines():
|
55 |
+
if not line:
|
56 |
+
continue
|
57 |
+
|
58 |
+
decoded_line = line.decode("utf-8").strip()
|
59 |
+
# Remove the "data:" prefix if present.
|
60 |
+
if decoded_line.startswith("data:"):
|
61 |
+
decoded_line = decoded_line[len("data:"):].strip()
|
62 |
+
|
63 |
+
# Check if the stream is finished.
|
64 |
+
if decoded_line == "[DONE]":
|
65 |
+
break
|
66 |
+
|
67 |
+
try:
|
68 |
+
json_data = json.loads(decoded_line)
|
69 |
+
for choice in json_data.get("choices", []):
|
70 |
+
delta = choice.get("delta", {})
|
71 |
+
content = delta.get("content", "")
|
72 |
+
full_output += content
|
73 |
+
yield full_output # Yield the accumulated text so far.
|
74 |
+
|
75 |
+
# If the finish reason is provided, stop further streaming.
|
76 |
+
if choice.get("finish_reason") == "stop":
|
77 |
+
return
|
78 |
+
except json.JSONDecodeError:
|
79 |
+
# If a line is not valid JSON, skip it.
|
80 |
+
continue
|
81 |
+
except requests.exceptions.RequestException as e:
|
82 |
+
yield f"Error during generation: {str(e)}"
|