AutoML tranlation model training error (3 times straight)

[mj_kim]

Hello.
I have been using AutoML tranlation model training for several years
and I have trained much more than 50 models,
but recently I faced an error 3 times straight.

When I click "Train" usually it doesn't take more than 48 hours.
But last 3 times it doesn't finish it task(if I train the new model with previous dataset) at all and doesn't even show what error.
So I tried to train the new model with other language, but the result was the same. 

1. EN-ID model with previous dataset (40979 sentence pairs) - The training never ends... I canceled it after about 40 hours.
2. EN-ID model with new dataset (31147 sentence pairs) - Error: Task [2853426188125208577] missed its deadline: 1717920695166
3. EN-ES model with new dataset (10377 sentence pairs)- Passed 24 hours but still running... and the result is again: Task  missed its deadline: 1718001884197

Definitely there is the problem in training... Really need a help...
Thank you.

[NathanKr]

Same problem here, training last around 24h and fail with the error: Task [-6125625568944717823] missed its deadline: 1718203999250

I have around 1800 sentence pairs. 

[NathanKr]

Same problem here

EN-FR, 1823 sentences pairs. Fails after 24h with error Task [-6125625568944717823] missed its deadline: 1718203999250

[Edith_GR]

I also had the same problem (missed its deadline), but with a much larger data set (2 000 000+). I used custom Tuning and Testing data sets. I was wondering if that could have been the cause. Can one of you tell me if they used custom Tuning and Testing sets too?