#problem related to the inability of GPT Chat to work for an extended period of time in one session

1 messages · Page 1 of 1 (latest)

lusty prairie
#

I have found a serious problem related to the inability of GPT Chat to work for an extended period of time in one session as various errors constantly occur. I want to finish writing my book, but there is not enough time for it and I want to use GPT Chat. Allow me to use GPT Chat with the function of remembering previous messages in the session even if it has been reloaded, I agree to anything for this, you can do whatever you want with the received data, because it will solve all the problems that I have encountered while using GPT Chat. Perhaps the obtained data will help you in working on improving GPT Chat.

lunar cliff
lusty prairie
lunar cliff
lusty prairie
lusty prairie
#

yep

lunar cliff
#

never heard of it. The error happensin the code geerated by chatgpt, or in the webpage?

lusty prairie
#

Therefore, the easiest way is to allow my chat to see the correspondence, I caught two completely different code errors while I was trying to work with it, and after that it went completely crazy

#

Errors are associated with the transfer of a large amount of data, after which this or another error pops up and you have to reload the page

lunar cliff
lusty prairie
#

?

#

allredy

lunar cliff
# lusty prairie

oh well I missunderstood that you were using the data analysis version.

lunar cliff
lusty prairie
#

To catch the first error, I copied about half and one a page of Word, for the second I copied one page of Word more than 4 times in a row.

lunar cliff
# lusty prairie To catch the first error, I copied about half and one a page of Word, for the se...

ok. The issue is actually very simple. You simply cannot. GPT-4 cannot handle that much data. I'ts just a fact. The total number of tokens should not exceed 8192 or it will start forgetting things. I suggest you to get chatgpt to write a chapter, then a summary with important informations of the following of the story, then edit message with summary, plus aks to write next chapter. To avoid having to give it too much text.

Text to tokens here: https://platform.openai.com/tokenizer

Explore developer resources, tutorials, API docs, and dynamic examples to get the most out of OpenAI's platform.

lusty prairie
#

That is, give a little bit of information at a time?

#

any ways to work with much data?

#

i agree on all ahah

lunar cliff
# lusty prairie any ways to work with much data?

no, except if you get access to paid and expensive gpt4-32k model api (4x more data). You can also use gpt-3-16k (double amount of data), but it is worse than gpt.4, in the playground, also paid as you use.

lusty prairie
#

In general, the gpt chat is useless in this regard, so I’ll wait for something similar from others, for now I’ll just put the idea aside