LLaDA
LLaDA copied to clipboard
Could you provide the loss curve of sft?
Hello,
Thank you for your excellent work. Could you provide the loss curve of sft? Because when I tried it myself, I found that the loss curve would fluctuate greatly, occasionally changing to 1 or 2, and sometimes to 0.0000.
+1
Thank you all for your attention! I'm truly sorry that the loss curve of LLaDA SFT has been cleared because the company's cluster regularly cleans up the logs.
Has it been resolved? My loss is also like this, it has never converged.