Vedanuj Goswami comments

Results 9 comments of


                                            Vedanuj Goswami

NLLB inference time model loading fails due to inconsistent vocabulary size

There are three additional tokens in the vocabulary that we add during training. Here is a response related to this : https://github.com/huggingface/transformers/issues/18043#issuecomment-1179317930 More specifically here are the additional fine-grained data...

Translation Quality of NLLB-200(Dense, 3.3B) model is worse as compared to all other models for Japanese & English. Can anyone suggest why?

Which models have you compared NLLB-200 3.3B dense with?

Temperature scaled sampling for multi dataset

@shinobi-AI Since, @parthduggal is already working on this can you check any other issue?

errors while running pytests .\test

Hi .. It seems your installation is missing omegaconf as shown in the errors : ``` ModuleNotFoundError: No module named 'omegaconf' ``` Can you try installing the package using pip...

TextVQA

This branch contains an example notebook that shows how to add a dataset : https://github.com/facebookresearch/mmf/blob/notebooks/notebooks/kdd_tutorial.ipynb Let us know if that helps or you need more information.

Cannot use mmf/utils/inference.py with existing configs

cc @brettallenyo Can you please take a look at this?

Absolute paths in VisualBERT configuration file

Hi @Yulmart .. can you share the command you used to run?

[fix] CircleCI GPU tests; sharded DDP tests

GPU tests are still failing it seems.

Replacing nn.Linear w/ te.Linear FP8 convergence issue

@ptrendx Any updates on this issue? We are observing similar loss differences across different model scales for llama.