Was the training done with FP8 or BF16?
#14
by
mindkrypted
- opened
As the title says, if the training was done in BF16, could we expect the release of those weights for getting better results while doing quantization?
Thanks,
M2 was trained with FP8