[Dashboard] Wrong sequence lengths in HF suite

[ngimel]

Megatron generally follows training procedure for Albert and therefore has to have sequence length of 512 (currently has 128), Albert paper https://arxiv.org/pdf/1909.11942.pdf, see section 4.1
Roberta paper states sequence length of 512 (currently has 128), see section 3.1 of the paper https://arxiv.org/pdf/1907.11692.pdf
Bert original paper
Bert paper states sequence length of 512 tokens https://arxiv.org/pdf/1810.04805.pdf, see section A.2, currently has 128.
I didn't check all the models, so this list is not exhaustive.
cc @anijain2305