Manh Dat Tran

> @yeshenpy > > Yes, your description is correct. I tried to use word embeddings and their average as word/doc node features, but this performs worse than using one-hot features...

> Can you check if this PR is included in your commit [fairinternal/wav2letter@e4f6d1d](https://github.com/fairinternal/wav2letter/commit/e4f6d1d236e653257c0377794f251c6810b4b2e6) i wonder do you check this or forgoten?

same problem when run [fl_asr_tutorial_finetune_ctc](https://github.com/flashlight/flashlight/blob/master/flashlight/app/asr/tutorial/README.md#finetuning-with-a-pretrained-ctc-model) with TPU colab

Could you provide to me too? My email: [[email protected]](mailto:[email protected])