Optimize and fix prompt generation #6
Conversation
mzukowski-reef
force-pushed
the
optimize-and-fix-prompt-generation
branch
from
9cb8272 to
645b6d6
Compare
| @@ -4,17 +4,17 @@ | |||
|
|
|||
| class PromptGeneratingPrompt: | |||
| def random_select(self, arr: list[str], num: int = 5) -> str: | |||
| random.shuffle(arr) | |||
| return ", ".join(arr[:num]) + ", etc" | |||
| return random.choices(arr, k=num) | |||
There was a problem hiding this comment.
| return random.choices(arr, k=num) | |
| return random.sample(arr, k=num) |
I think we should not pick the same thing twice.
| # strip quotations | ||
| line = line.strip("\"'") |
There was a problem hiding this comment.
Won't that mess up lines, where a quote mark is at the start/end of the line, and the other quote in the middle of the line? The strip will only remote the quote at the start/end, not the quote in the middle.
There was a problem hiding this comment.
we don't care - we care ONLY about the lines that are single line prompt - and ends with ? - if result is broken or "not perfect" - it will be ignored
| @@ -19,7 +20,10 @@ def generate_prompts( | |||
| max_new_tokens: int = 2000, | |||
| temperature: float = 1.0, | |||
| filepath: str = "prompts.txt", | |||
| leftover_prompts: deque = None, | |||
There was a problem hiding this comment.
| leftover_prompts: deque = None, | |
| leftover_prompts: deque | None = None, |
Please don't hate me for this :D
There was a problem hiding this comment.
I... I... :D
| if len(new_prompts) > total_prompts: | ||
| # Save extra prompts for next batch | ||
| leftover_prompts.extend(new_prompts[total_prompts:]) |
There was a problem hiding this comment.
I think you need the original value of total_prompts that was passed in the function. We are subtracting the number of prompts generated from this variable.
No description provided.