Upload configs.yaml with huggingface_hub

configs.yaml

@@ -9,7 +9,7 @@ eval_strategy: 'no'
 finetuning_type: full
 formatting: sharegpt
 global_batch_size: 96
-gradient_accumulation_steps: 1
+gradient_accumulation_steps: 3
 hub_model_id: mlfoundations-dev/llama3-1_8b_4o_annotated_aops
 include_hp: dcft/train/hp_settings/reasoning.yaml
 learning_rate: 1.0e-05
@@ -21,7 +21,7 @@ model_name_or_path: Qwen/Qwen2.5-7B-Instruct
 num_train_epochs: 3.0
 output_dir: /tmp/dcft_checkpoints/llama3-1_8b_4o_annotated_aops
 overwrite_cache: true
-per_device_train_batch_size: 3
+per_device_train_batch_size: 1
 plot_loss: true
 preprocessing_num_workers: 16
 push_to_db: true