Parallel GPT Feedback #1421
world33
started this conversation in
Idea / Feature Request
Parallel GPT Feedback
#1421
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hello I have some queries and some suggestions about the Parallel GPT tool:
1) QUALITY OF RESPONSES
the text quality outcome using the rowy Parallel GPT tool is much inferior and less comprehensive to what you get using the UI at chatgpt.
For example for this prompt:
1) create a html formatted summary of a typical Master of Vegetal Biology program which includes:<h3>Program Name</h3> <h3>Program or Degree abbreviation</h3> <h3>Duration range</h3> <h3>Tuition range</h3> <h3>Overview</h3> <h3>Curriculum Overview by year</h3> <h3>Key Components</h3> <h3>Career Prospects</h3> <h3>Salary Expectations</h3> <h3>Conclusions</h3> 2) please specify in the conclusions that program duration, tuition fees, curriculum, key components, career prospects and salary expectations can vary: a) by the chosen country or location where to study the program b) by the chosen university where to study the program 3) please do not use <h1>, <h2>, <html>, <head>, <body> or <div> tagsI get the following result from Parallel GPT with many blank fields and general "Varies" answers:
`
Program Name
Master of Vegetal BiologyProgram or Degree abbreviation
MVBDuration range
VariesTuition range
VariesOverview
Curriculum Overview by year
Key Components
Career Prospects
Salary Expectations
Conclusions
Program duration, tuition fees, curriculum, key components, career prospects, and salary expectations can vary by the chosen country or location where to study the program and by the chosen university where to study the program.``
Program Name:
Master of Vegetal Biology
Program or Degree Abbreviation:
MVB
Duration Range:
1.5 to 2 years
Tuition Range:
Varies by country and university
Overview:
The Master of Vegetal Biology program is designed for students passionate about plant biology and ecology. This program offers a comprehensive understanding of plant life, from molecular processes to ecosystem dynamics.
Curriculum Overview by Year:
Year 1: Core courses in plant physiology, genetics, and ecology. Research methodology and lab work.
Year 2: Specialization options such as plant breeding, conservation, or biotechnology. Thesis research.
Key Components:
Career Prospects:
Graduates of the Master of Vegetal Biology program can pursue various career paths, including:
Salary Expectations:
Salaries in the field of Vegetal Biology can vary widely depending on factors such as location, level of experience, and job role. On average, graduates can expect competitive salaries that increase with experience.
Conclusions:
It's important to note that the duration, tuition fees, curriculum, key components, career prospects, and salary expectations of the Master of Vegetal Biology program can vary significantly based on the country or location where you choose to study and the specific university you select. Prospective students should carefully research and compare programs to find the best fit for their academic and career goals.
` I set the temperature of the Config file of the parallel GPT tool to 0.3 and made no other modification to the config file. **My question is: are the standard settings of the config file exactly the same as the ones in the online ChatGPT UI? Is there anyway I can get the same quality and comprehensiveness of results?**2) QUANTITY OF RESPONSES
I uploaded around 6,000 of similar prompts shown above and I only got 1,578 valid responses all of inferior quality. The rest were all error responses such as the following logs show:
`2023-09-22 21:29:53:53.732 +10
derivative-function
backend-function
chatGptResponse
{"error":"Error: 429 Rate limit reached for default-gpt-3.5-turbo in organization org-q9styve3MVCMTdXpE8U7nFkS on tokens per min. Limit: 90000 / min. Current: 89887 / min. Contact us through our help center at help.openai.com if you continue to have issues."}
2023-09-22 21:29:53:53.723 +10
derivative-function
backend-function
chatGptResponse
{"error":"Error: 429 Rate limit reached for default-gpt-3.5-turbo in organization org-q9styve3MVCMTdXpE8U7nFkS on requests per min. Limit: 3500 / min. Please try again in 17ms. Contact us through our help center at help.openai.com if you continue to have issues."}
2023-09-22 21:29:53:53.712 +10
derivative-function
backend-function
chatGptResponse
{"error":"Error: 429 Rate limit reached for default-gpt-3.5-turbo in organization org-q9styve3MVCMTdXpE8U7nFkS on tokens per min. Limit: 90000 / min. Current: 89909 / min. Contact us through our help center at help.openai.com if you continue to have issues."}
`
My question is: is there any way to change the config file or the tool in a way that it does not go over the 90000 (sometime it shows 3500?) limit? Maybe by introducing a delay per request or offering an option to submit the prompts one at the time? There is no point to have such a great bulk tool that gives so many errors. It defeats the purpose.
3) TOOL FEATURES - BULK DELETE EMPTY/ERROR RESPONSES
Once received the lower number of inferior results and exported to csv I could not find a way to select them in bulk and retry with the blank rows that previously gave an error. It seems it is only possible to select and delete one row at the time. That means that I had to reupload a new csv file with the blank response prompts.
Also when trying to sort the CHATGP response column it automatically filter the rows with non blank (valid) responses.
Would it be possible to add a drag select option to delete only some rows/entries?
Alternatively would it be possible to add a filter rows option that allows to select blank fields or non blank fields and give the option to delete them in bulk?
Obiously these two features would not be necessary if the tool would not encounter so many limit errors and offer a way to comply with the rate limits set by Chat GPT (90000 / min or 3500 / min shown in the error logs)
This is a fantastic tool for Bulk prompting ChatGPT but needs a bit of tweaking in my opinion to make it really work for large datasets.
Thanks!
Beta Was this translation helpful? Give feedback.
All reactions