curriculum learning - llama3-8b-instruct finetuned on easy/hard stages of clem games, with hard stage being the (game,exp,task_id) tuples the base model failed at.
this is the model trained on only the easy stage (successful task_ids of base model)
batch_size: 1
grad_accum_steps: 4
learning_rate: 0.0002
num_train_epochs: 1
lora_r: 8
lora_alpha: 16
lora_dropout: 0.05
