Thomas Wang issues

Repositories
Issues
Comments

Results 28 issues of


                                            Thomas Wang

Share a list of weight attributes instead of a single one in TiedLayerSpec API

Attempt to make unigram faster 2.

[WIP] tr13 evaluation: T0 eval + XNLI

[WIP] OPT experimentations

Some update to tr10 config

This PR is for sorting out the tr10-104B config.

Add tr6g for loss reweighting according to positional loss frequency

[Tensorboard] Log text prediction in evaluation

A very useful tool in order to understand model performance beyond obtaining loss: Actually show what are the predictions. It'd be very useful to be able to "see" the output...

Good First Issue

[WIP] Hack my way to get OPT running

[MLM] Train script for non causal decoder

WIP: Shared t5 code

requires: https://github.com/microsoft/DeepSpeed/pull/2035 TODO: - [x] Make sure we can run shared enc/dec with MLM - [x] Add test making sure that it runs. with MLM - [ ] Make sure...