Ganesh Krishnan

It's version 2. The exact bug is in the line above: when params.search.value is integer it doesnt have replace function (like string does) replicate by creating a number tag and...

This was initially from the default model via `ai = aitextgen(tf_gpt2="124M", to_gpu=True)` This was finetuned and the model saved locally as "trained_model/pytorch_model.bin" which I keep overwriting with new inputs. However...

ok it makes sense now. I have to get the 124M model, train it on file, save the model and then reload the model for further training. All of this...

You can run your pgn thru pgn-extractor which writes out the file in the desired format. The main issue with supervised learning is the deliberate exception thrown. If I comment...

thanks @killerducky The training did output files after I removed the exception but I couldn't run the trainer on the data since parse.py file no longer exists and chunkparser.py throws...

@hsntgm You are correct. its now train.py The command is something like this ./train.py --cfg=configs/aihello.yaml --output=/data/aihello.com/training/output

This still has issue parsing pgn files especially if there are multiple games in the pgn. Can we not use a standard pgn parser library similar to python uci-chess?

Supervised Learning is still failing on getline. It reads in junk characters for an ASCII pgn file. Stackoverflow says something about BOM. As a Java programmer I am so shielded...

I could fix it by commenting out init(); method in pgn.cpp. pgn.h has changed and the author forgot to comment the overriden method I guess

Unfortunately I am running a fork of leela chess and I just noticed this master doesn't have the init() method. Sorry I feel stupid now. Past my bedtime! I am...