add_ctxemb: 'False'
add_vocab: 'False'
all_gather: 'true'
batch_size_training: '16'
checkpoint_type: StateDictType.SHARDED_STATE_DICT
clipping_norm: '-1.0'
compare: 'False'
cpu_np_head: 'False'
ctx_proj_layer: 'False'
ctx_use_peft: 'False'
dataset: llava_llama3_selfrag_single_dataset_5k
dist_checkpoint_folder: llama3.full.prompt_llava.context_mask.with_context.without_sp.data_5k_epoch2
dist_checkpoint_root_folder: ./checkpoints
enable_fsdp: 'True'
freeze_ctx_encoder: 'False'
freeze_layers: 'False'
freeze_question_encoder: 'False'
from_hf: 'False'
fsdp_activation_checkpointing: 'True'
gamma: '0.85'
load_np_head: 'False'
low_cpu_fsdp: 'True'
lr: 2e-05
memory_bank_length: '0'
micro_batch_size: '4'
mixed_precision: 'True'
model_name: meta-llama/Llama-3.1-8B
model_use_peft: 'False'
natural_form: 'False'
np_weight: '100.0'
num_epochs: '3'
num_freeze_layers: '1'
num_workers_dataloader: '1'
one_gpu: 'False'
optimizer: AdamW
output_dir: peft_checkpoint
peft_method: None
pure_bf16: 'True'
quantization: 'False'
question_proj_layer: 'False'
resume_epoch: '0'
ret_checkpoint_folder: ''
ret_first: 'False'
retriever: ''
run_validation: 'True'
save_model: 'True'
save_optimizer: 'False'
seed: '2'
sharding_strategy: ShardingStrategy.FULL_SHARD
single: 'False'
target_modules: q_proj|k_proj|v_proj|o_proj|gate_proj|up_proj|down_proj
token_name: meta-llama/Llama-3.1-8B-Instruct
train: 'True'
use_fast_kernels: 'False'
use_fp16: 'False'
val_batch_size: '1'
weight_decay: '0.0'