HI tried too tune model with certain data for my organisation and when I try that model in the chat screen it gives me internal error. Please don’t tell me it is an intermittent error I was having this issue from yesterday.
3 Likes
Having exactly the same issue. Created structured prompt with system and 20+ IO rows and it worked with test cases. Once used this as data for model supervised tuning and trained the model - every prompt to the tuned model ends with internal error.
Where can I see any kind of logs for that? What can be the cause?
1 Like
yeah I got the same, I noticed that a100’s aren’t available on colab too. Thinking google’s are diverting compute to training/inference to gemini 2 thinking and new models with deepseek etc… Hopefully not this could be a really useful feature.
Tbh maybe a good reason to finetune models independently of locked in platforms. Not something that’s reliable for real applications/use.