Upload training_config.yml with huggingface_hub

training_config.yml

@@ -16,10 +16,10 @@ checkpointer:
   _component_: torchtune.utils.FullModelMetaCheckpointer
   checkpoint_dir: checkpoints/Meta-Llama-3-8B-Instruct/original
   checkpoint_files:
-  - meta_model_0.pt
+  - meta_model_1.pt
   adapter_checkpoint: null
   recipe_checkpoint: null
-  output_dir: output_checkpoints/experiment_5
+  output_dir: output_checkpoints/experiment_6
   model_type: LLAMA3
 resume_from_checkpoint: false
 interim_checkpoint_steps: 15000
@@ -49,16 +49,16 @@ shuffle: true
 batch_size: 6
 optimizer:
   _component_: torch.optim.AdamW
-  weight_decay: 0.008
-  lr: 0.0001
+  weight_decay: 0.99
+  lr: 2.0e-05
 lr_scheduler:
   _component_: torchtune.modules.get_cosine_schedule_with_warmup
-  num_warmup_steps: 100
+  num_warmup_steps: 4
 loss:
   _component_: torch.nn.CrossEntropyLoss
-epochs: 6
+epochs: 60
 max_steps_per_epoch: null
-gradient_accumulation_steps: 16
+gradient_accumulation_steps: 260
 compile: false
 output_dir: /tmp/lora_finetune_output
 metric_logger: