Update tfm_alive_fine-tuning.yaml

tfm_alive_fine-tuning.yaml

@@ -23,145 +23,8 @@ model:
   alive_hidden_dim: 512  # Hidden dimension for alive prediction head
   alive_hidden_layers: 1     # Number of hidden layers in alive prediction head
 
-# Training configuration
-training:
-  batch_size: 16
-  grad_accum_steps: 4
-  learning_rate_prediction_head: 0.00005
-  learning_rate_embedder: 0
-  learning_rate_processor: 0.00001
-  weight_decay: 0.01
-  num_epochs: 30
-  warmup_steps: 1000
-  max_grad_norm: 1.0
-  
-  scheduler: 'cosine'  # 'cosine' or 'linear'
-  
-  base_model_path: '/share/guwanjun-local/cs2-demo-analytics/checkpoints_pretraining_v2/final.pth'
-  checkpoint_dir: 'checkpoints_alive_fine-tuning_v2'
-
-  from_scratch: False  # Whether to train from scratch or fine-tune from a pre-trained model
-
-  use_lora: False # Whether to use LoRA for fine-tuning
-  lora_r: 8
-  lora_alpha: 16
-  lora_dropout: 0.1
-
 # Data configuration
 data:
-  train_data_path: 
-    - archive_1.pt
-    - new_archive_1.pt
-    - archive_2.pt
-    - new_archive_2.pt
-    - archive_3.pt
-    - new_archive_3.pt
-    - archive_4.pt
-    - new_archive_4.pt
-    - new_archive_5.pt
-    - archive_5.pt
-    - archive_6.pt
-    - new_archive_6.pt
-    - archive_7.pt
-    - new_archive_7.pt
-    - archive_8.pt
-    - new_archive_8.pt
-    - archive_9.pt
-    - new_archive_9.pt
-    - archive_10.pt
-    - new_archive_10.pt
-    - archive_11.pt
-    - new_archive_11.pt
-    - archive_12.pt
-    - new_archive_12.pt
-    - archive_13.pt
-    - new_archive_13.pt
-    - archive_14.pt
-    - new_archive_14.pt
-    - archive_15.pt
-    - new_archive_15.pt
-    - archive_16.pt
-    - new_archive_16.pt
-    - archive_17.pt
-    - new_archive_17.pt
-    - archive_18.pt
-    - new_archive_18.pt
-    - archive_19.pt
-    - new_archive_19.pt
-    - archive_20.pt
-    - new_archive_20.pt
-    - archive_21.pt
-    - new_archive_21.pt
-    - archive_22.pt
-    - new_archive_22.pt
-    - archive_23.pt
-    - new_archive_23.pt
-    - archive_24.pt
-    - new_archive_24.pt
-    - archive_25.pt
-    - new_archive_25.pt
-    - archive_26.pt
-    - new_archive_26.pt
-    - archive_27.pt
-    - new_archive_27.pt
-    - archive_28.pt
-    - new_archive_28.pt
-    - archive_29.pt
-    - new_archive_29.pt
-    - archive_30.pt
-    - new_archive_30.pt
-    - archive_31.pt
-    - new_archive_31.pt
-    - archive_32.pt
-    - new_archive_32.pt
-    - archive_33.pt
-    - new_archive_33.pt
-    - archive_34.pt
-    - new_archive_34.pt
-    - archive_35.pt
-    - new_archive_35.pt
-    - archive_36.pt
-    - new_archive_36.pt
-    - archive_37.pt
-    - new_archive_37.pt
-    - archive_38.pt
-    - new_archive_38.pt
-    - archive_39.pt
-    - new_archive_39.pt
-    - archive_40.pt
-    - new_archive_40.pt
-    - archive_41.pt
-    - new_archive_41.pt
-    - archive_42.pt
-    - new_archive_42.pt
-    - archive_43.pt
-    - new_archive_43.pt
-    - archive_44.pt
-    - new_archive_44.pt
-    - archive_45.pt
-    - archive_46.pt
-    - archive_47.pt
-    - archive_48.pt
-    - archive_49.pt
-    - archive_50.pt
-    - archive_51.pt
-    - archive_52.pt
-    - archive_53.pt
-    - archive_54.pt
-    - archive_55.pt
-    - archive_56.pt
-    - archive_57.pt
-    - archive_58.pt
-    - archive_59.pt
-    - archive_60.pt
-    - archive_61.pt
-    - archive_106.pt
-    - archive_107.pt
-  val_data_path: 
-    - archive_108.pt
-    - archive_109.pt
-  num_workers: 4
-  
   # Data dimensions (must match model)
   ticks_per_sample: 64  # Number of ticks in each training sample
   seq_len: 512          # Must match model.seq_len