Upload training_config.yml with huggingface_hub

training_config.yml

@@ -21,12 +21,12 @@ checkpointer:
   recipe_checkpoint: null
   output_dir: output_checkpoints/experiment_1
   model_type: LLAMA3
-resume_from_checkpoint: true
+resume_from_checkpoint: false
 interim_checkpoint_steps: 5000
 interim_gen_steps: null
 max_new_tokens: 100
-temperature: 0.6
-top_k: 300
+temperature: 0.8
+top_k: 333
 dataset:
   _component_: ds.EvenBatcher
   buffer_size: 36
@@ -46,7 +46,7 @@ dataset:
       train_on_input: false
 seed: null
 shuffle: true
-batch_size: 4
+batch_size: 6
 optimizer:
   _component_: torch.optim.AdamW
   weight_decay: 0.01
@@ -77,8 +77,8 @@ inference:
     {video}
 
     Caption the previous video.'
-  max_new_tokens: 300
-  temperature: 0.6
-  top_k: 300
+  max_new_tokens: 333
+  temperature: 0.8
+  top_k: 333
   quantizer: null
 gradient-accumulation-steps: 32