config: | |
(): custom_colbert.utils.train_custom_colbert_models.ColModelTrainingConfig | |
output_dir: !path ../../../models/without_tabfquad_no_pairwise/train_colpali-3b-mix-448 | |
processor: | |
() : custom_colbert.utils.wrapper.AutoProcessorWrapper | |
pretrained_model_name_or_path: "./models/paligemma-3b-mix-448" | |
max_length: 50 | |
model: | |
(): custom_colbert.utils.wrapper.AutoColModelWrapper | |
pretrained_model_name_or_path: "./models/paligemma-3b-mix-448" | |
training_objective: "colbertv1" | |
# attn_implementation: "eager" | |
torch_dtype: !ext torch.bfloat16 | |
# 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 custom_colbert.utils.dataset_transformation.load_train_set | |
eval_dataset_loader: !import ../data/test_data.yaml | |
max_length: 50 | |
run_eval: true | |
add_suffix: true | |
loss_func: | |
(): custom_colbert.loss.colbert_loss.ColbertLoss | |
tr_args: !import ../tr_args/default_tr_args.yaml | |
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: '(.*(language_model).*(down_proj|gate_proj|up_proj|k_proj|q_proj|v_proj|o_proj).*$|.*(custom_text_proj).*$)' | |
# target_modules: '(.*(language_model).*(down_proj|gate_proj|up_proj|k_proj|q_proj|v_proj|o_proj).*$|.*(custom_text_proj).*$)' | |