Hassan
Hassan
Greetings. I'll preface my question with a disclaimer that I don't have much experience in ML and I'm still exploring myself, so I apologize if this question may sound silly...
Is there a way to pass attention masks of tokenized queries to the PPOTrainer or the `generate()` function? I see that other classes in the library have padding data collators,...
Running the script in `examples/research_projects/stack_llama/scripts/rl_training.py` with the following command: ``` accelerate launch --multi_gpu --num_machines 1 --num_processes 4 llama_example.py --model_name meta-llama/Llama-2-7b-chat-hf --tokenizer_name meta-llama/Llama-2-7b-chat-hf --log_with wandb --reward_model text-classification:Hello-SimpleAI/chatgpt-detector-roberta ``` produces the following...
Running the docker-compose file to start the react server fails in the react-ui container due to this error: `The engine "node" is incompatible with this module. Expected version ">=16.0.0
* move utility functions in `catanatron_server` to `utils.py` * create `catanatron_server.pickle` module for operations on pickle files * add ability to save game states when using main script to local...
For visiting old game states, it goes without saying that they need to be saved somewhere. However, saving the hundreds of game states of each game is very costly if...