Alexia Jolicoeur-Martineau

Adding FID statistics calculation as an option (can now do "train", "eval", or "fid_stats")

8

With these small changes, you can get the fid statistics by running with --mode "fid_stats". It loops through the dataset for 1 epoch and extract the FID statistics. That makes...

GBDTMO crash in Google colab (see simple example)

6

https://colab.research.google.com/drive/1m_QXVbHbeoho5tF9TUYq6R-PSz3536EU?usp=sharing

Confidence interval weirdness and lack of match with tmle using dbart

2

I am almost always getting non-significant results with _bartCause_ and almost always getting very significant results with the _tmle_ package (when only using dbarts in tmle). I have no idea...

ignoring second gpu

8

Wether I'm using mpi or not, if I have 2 gpus with one node, it will always ignore the second gpu. I am unable to use multiple gpus. How do...

cannot replicate DPO results of zephyr

5

I cannot replicate the DPO results for zephyr. I use a modified version of config_full.yaml with the only difference being that I set gradient_accumulation_steps: 4 instead of 2 because I...

Major bug: Chat template is not actually applied in run_sft.py and run_dpo.py

7

In run_sft.py and run_dpo.py, it says that it applies the chat template. But this is not actually done. In the code below, column_names contains all the names of the columns,...