Delete training_config.yml

training_config.yml

@@ -1,70 +0,0 @@
-config:
-  (): colpali_engine.trainer.colmodel_training.ColModelTrainingConfig
-  output_dir: !path ../../../models/colphi3-ba256-ckpt-5e
-  processor:
-    () : colpali_engine.utils.transformers_wrappers.AutoProcessorWrapper
-    class_to_instanciate: !ext colpali_engine.models.ColPhi3_5Processor
-    pretrained_model_name_or_path: "./models/Phi-3.5-vision-instruct"
-    trust_remote_code: True
-  model:
-    (): colpali_engine.utils.transformers_wrappers.AllPurposeWrapper
-    class_to_instanciate: !ext colpali_engine.models.ColPhi3_5
-    pretrained_model_name_or_path: "./models/Phi-3.5-vision-instruct"
-    # attn_implementation: "eager"
-    torch_dtype:  !ext torch.bfloat16
-    trust_remote_code: True
-#    device_map: "auto"
-#    quantization_config:
-#      (): transformers.BitsAndBytesConfig
-#      load_in_4bit: true
-#      bnb_4bit_quant_type: "nf4"
-#      bnb_4bit_compute_dtype:  "bfloat16"
-#      bnb_4bit_use_double_quant: true
-
-  dataset_loading_func: !ext colpali_engine.utils.dataset_transformation.load_train_set
-  eval_dataset_loader: !import ../data/test_data.yaml
-
-  # max_length: 50
-  run_eval: true
-  
-  loss_func:
-    (): colpali_engine.loss.late_interaction_losses.ColbertPairwiseCELoss
-  tr_args:
-    (): transformers.training_args.TrainingArguments
-    output_dir: null
-    overwrite_output_dir: true
-    num_train_epochs: 5
-    per_device_train_batch_size: 32
-    gradient_checkpointing: true
-    gradient_checkpointing_kwargs: { "use_reentrant": false }
-    # gradient_checkpointing: true
-    # 6 x 8 gpus = 48 batch size
-    # gradient_accumulation_steps: 4
-    per_device_eval_batch_size: 8
-    eval_strategy: "steps"
-    dataloader_num_workers: 8
-    # bf16: true
-    save_steps: 500
-    logging_steps: 10
-    eval_steps: 1000
-    warmup_steps: 100
-    learning_rate: 5e-4
-    save_total_limit: 1
-    save_safetensors: False
-    # resume_from_checkpoint: true
-    # optim: "paged_adamw_8bit"
-    # wandb logging
-    # wandb_project: "colqwen2"
-    # run_name: "colqwen2-ba32-nolora"
-    report_to: "wandb"
-  peft_config:
-    (): peft.LoraConfig
-    r: 32
-    lora_alpha: 32
-    lora_dropout: 0.1
-    init_lora_weights: "gaussian"
-    bias: "none"
-    task_type: "FEATURE_EXTRACTION"
-    target_modules: '(.*(model).*(down_proj|gate_up_proj|up_proj|k_proj|q_proj|v_proj|o_proj|qkv_proj|fc1|fc2|lm_head|img_projection.0|img_projection.2).*$)'
-#     target_modules: '(.*(model).*(down_proj|gate_up_proj|up_proj|k_proj|q_proj|v_proj|o_proj|qkv_proj|fc1|fc2).*$)'
-