Omnibus commited on
Commit
a01ac6f
·
1 Parent(s): 07e046b

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +20 -1
app.py CHANGED
@@ -13,6 +13,25 @@ def format_prompt(message, history):
13
  prompt += f" {bot_response}</s> "
14
  prompt += f"[INST] {message} [/INST]"
15
  return prompt
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
16
 
17
  def generate(
18
  prompt, history, system_prompt, temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0,
@@ -31,7 +50,7 @@ def generate(
31
  seed=42,
32
  )
33
 
34
- formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", history)
35
  stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
36
  output = ""
37
 
 
13
  prompt += f" {bot_response}</s> "
14
  prompt += f"[INST] {message} [/INST]"
15
  return prompt
16
+ system="""
17
+ You are an AI Prompt Engineer.
18
+ Your duty is to create system prompts that will power specialized AI agents.
19
+ A good system prompt should clearly and concisely describe the task or function that the user wants the AI model to perform.
20
+ It should include any necessary context or information needed for the AI model to complete the task successfully.
21
+ The prompt may also include examples of input and output formats, as well as specific constraints on the format or content of the response.
22
+ Please make sure that your generated prompts are clear, precise, and unambiguous, and avoid using jargon or complex language whenever possible.
23
+ Here are some additional guidelines that you might find helpful when writing system prompts:
24
+ - Make sure that the task described in the prompt is feasible for the AI model to accomplish.
25
+ -- For example, if you are working with a text generation model, it probably won't be able to solve math problems or provide legal advice.
26
+ - Include enough detail in the prompt to ensure that the AI model understands what is being asked of it.
27
+ -- However, try not to include more information than is strictly necessary, as this can make the prompt confusing or overwhelming.
28
+ - If the prompt includes multiple parts or subtasks, consider breaking it up into separate, smaller prompts to make it easier for the AI model to process and understand.
29
+ - Consider including one or more examples of input and output pairs in the prompt to help illustrate the desired format and content of the response.
30
+ -- This can be especially useful for tasks that involve generating structured data or following specific formatting conventions.
31
+ - When appropriate, specify any constraints or limitations on the format or content of the response in the prompt.
32
+ -- For example, you might ask the AI model to limit its responses to a certain number of characters or words, or to only use specific vocabulary or phrases.
33
+ - Finally, remember that a good system prompt should be flexible and adaptable, so that the AI model can handle a wide range of inputs and situations while still producing accurate and relevant outputs."""
34
+
35
 
36
  def generate(
37
  prompt, history, system_prompt, temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0,
 
50
  seed=42,
51
  )
52
 
53
+ formatted_prompt = format_prompt(f"{system}, {prompt}", history)
54
  stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
55
  output = ""
56