Feature #3641
Updated by Ram Kordale 5 months ago
Let the first cell contain all inputs. Inputs are questions_file, model, index, board, grade, subject, chapter_num, styles (which is an array of strings)
Also, set CONTEXT_SIZE to 40960.
questions_file has the format:
Chapter 1
E
Chapter 2
E
...
where E is of the format:
E1
E2
...
where Ei is of format
#$Exercise: <exercise#> (example: 1.2.5.1)
#$Question: <question>
#$Figure: List of <GCS URLs> with name <exercise #>.1.jpg, <exercise #>.2.jpg...
#$Answer: <answer>
-for every <style> in styles
--template[key, value] = ["stylexyz1abc", <style>]
--num_tokens=0
--for every question in the file
--prompt=<question>
--response,num_tokens=call_openai_api(version=1.0, prompt, template, "temperature": 1, "max_tokens": 4096, "top_p": 1, "frequency_penalty": 0, "presence_penalty": 0, prompt)
---if (num_tokens > CONTEXT_SIZE) num_tokens=0
---styled_question = content between "###---###---###"
---store styled_question and answer with key <board>-<grade>-<subject>-<chapter>-<exercise>-<style>
---if <board>-<grade>-<subject>-<chapter>-<exercise>-standard does not exist
----store <question> and answer with key <board>-<grade>-<subject>-<chapter>-<exercise>-standard.
You can start with a single entry file such as
#$Exercise: 1.2.5.1
#$Question: Maya dyes dresses. She had to dye 30 dresses. She has so far finished 20 dresses. What fraction of dresses has she finished?
#$Answer: 2/3