Kap Thang

> Hi @thangk > I'm manually merging the nmt branch to the main branch. I will post my questions here and would appreciate it if you reply here. > >...

> [@thangk](https://github.com/thangk) > > > Yes, nmt handles if I recall correctly (I'll double check). > > Please double check. I need to know if I should connect our embedding...

I'm not sure if this was already fixed because I noticed some changes were made in the GPU mode section. But I ran it with my `compare.py` script, and it...

> Hi [@thangk](https://github.com/thangk) Please see the above two comments. The GPU code fails to generate teamsvecs due to an out-of-memory error! Okay, I'll modify the current logic and then test...

> Can you give me a timeline? We need to fix this and move on asap. Thank you. I was test running it after my first fix attempt, but I...

I've set it to use 90% of available memory for the processing. It's working so far and will report back of errors, if any - currently running the entire imdb...

> thank you. you mean in the codebase, we can set sth, right? Yes, here's the bit about setting how much GPU memory to use. We can make it configurable...

It seems the GPU method we originally had did not actually process the teamsvecs generation. So, I had to make modifications and try different attempts. It appears that it might...

> * the links are based on the specific version. It's better to link it such that always show the latest version (link to the weburl without version info). >...

@hosseinfani , Since we're dropping the index column in the teamsvecs.pkl do we still need the indexes.pkl to keep track of the records of each row in the teamsvecs.pkl? I...