I observe that the T5 models have different forward outputs given the same input information, even after disabling the dropout layers. I’m curious is there any other randomness in the forward process except dropout? (P.S. The above issue occurs when open the train mode by running model.train(). Ob…

T5 models have non-deterministic outputs even after disabling dropout

mahmutc September 15, 2024, 9:28am 10

indeed even torch.equal returns true. I don’t know how I got two different(but close) results last time, sorry.

Topic		Replies	Views
T5 forward pass versus generate, latter outputs non-sense Beginners	8	2911	March 25, 2021
T5 weird behavior between model.forward() and model.generate 🤗Transformers	0	112	March 31, 2024
T5 randomness in generation Models	4	1245	May 15, 2022
T5 Model Generate and Model Outputs Vastly Different Beginners	2	835	August 19, 2025
The output of T5 is not consistent on multiple sequences 🤗Transformers	1	870	May 11, 2022