I tried to train with `google/flan-t5-large`
It took ~30 minutes per epoch on an A100.
The default timeout on replicate is 30 minutes.
The fine-tuning canceled at 99%
doh
29.1.2023 15:58I tried to train with `google/flan-t5-large`It took ~30 minutes per epoch on an A100. The default timeout on replicate is 30 minutes.The...yay, it seems to work - lots of rough edges on making it work on replicate
28.1.2023 18:12yay, it seems to work - lots of rough edges on making it work on replicateit stops between each epoch to evaluate
Not sure if evaluation is necessary if running in a non-interactive system
28.1.2023 17:45it stops between each epoch to evaluateNot sure if evaluation is necessary if running in a non-interactive systemRunning on a T4 GPU, I'm getting 1.25it/s
Running on an A100 on https://replicate.com, I'm getting 3.0it/s
On an A100 that means ~10min per epoch on the sample training. The walkthrough used 5 epochs. . . ~50 min to fine-tune Flan-T5-Base?
28.1.2023 17:41Running on a T4 GPU, I'm getting 1.25it/sRunning on an A100 on https://replicate.com, I'm getting 3.0it/sOn an A100 that means...After spending the last few weeks exploring how #finetuning #stablediffuson works in #dreambooth at https://replicate.com, I am excited to experiment with fine-tuning an LLM using #flan T5.
I found this walkthrough useful to get started: https://www.philschmid.de/fine-tune-flan-t5
28.1.2023 17:26After spending the last few weeks exploring how #finetuning #stablediffuson works in #dreambooth at https://replicate.com, I am excited to...$ git push origin add_jesse
It’s official. I’ve joined replicate!
13.1.2023 20:31$ git push origin add_jesseIt’s official. I’ve joined replicate!https://replicate.com/aboutAIs are lazy coders too!
My #gpt chat bot has a command "JS" that asks it to generate & run javascript code.
For instance if I say ask it to generate a function to reverse a string, it will create function that reverses the letters one at a time.
Additionally I exposed GPT to itself. Giving examples of how to prompt itself.
Asked to code a javascript graphics SDK, rather than coding it, it decides to describe the goals and call out to another AI instance to do the actual work
😂
3.12.2022 19:46AIs are lazy coders too!My #gpt chat bot has a command "JS" that asks it to generate & run javascript code.For instance if I...The story seemed exciting until the castle scene....
me: summarize the lord of the rings trilogy using only emoji
#chatgpt 🗡️💍🌋🌋🌋👦🏻👩🏻🧝🏻♂️🧝🏻♀️🧝🏻♂️🧝🏻♂️🧝🏻♂️🧝🏻♂️🧝🏻♂️👦🏻🗡️💥💥💥🌋🌋🌋🌋🌋🌋🌋👑💥💥💥👑🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰🏰
2.12.2022 04:08The story seemed exciting until the castle scene....me: summarize the lord of the rings trilogy using only emoji#chatgpt ...this #covid exhaustion is intense
Even with a full night sleep, I need a nap in car between driving the kids to school and driving back home
Then when I get home I need a nap until at least noon
Then I can make it until 9p or so - but everything is swimmy
First symptom on Nov 18th. Tested positive Nov 23rd.
30.11.2022 17:15this #covid exhaustion is intenseEven with a full night sleep, I need a nap in car between driving the kids to school and driving back...