model:
  _component_: torchtune.models.llama3_1.qlora_llama3_1_8b
  lora_attn_modules:
  - q_proj
  - v_proj
  - k_proj
  - output_proj
  apply_lora_to_mlp: true
  apply_lora_to_output: false
  lora_rank: 8
  lora_alpha: 16
tokenizer:
  _component_: torchtune.models.llama3.llama3_tokenizer
  path: /tmp/Meta-Llama-3.1-8B-Instruct/original/tokenizer.model
  max_seq_len: null
checkpointer:
  _component_: torchtune.training.FullModelHFCheckpointer
  checkpoint_dir: /tmp/Meta-Llama-3.1-8B-Instruct/
  checkpoint_files:
  - model-00001-of-00004.safetensors
  - model-00002-of-00004.safetensors
  - model-00003-of-00004.safetensors
  - model-00004-of-00004.safetensors
  recipe_checkpoint: null
  output_dir: /tmp/Meta-Llama-3.1-8B-Instruct/
  model_type: LLAMA3
resume_from_checkpoint: false
save_adapter_weights_only: false
dataset:
  _component_: torchtune.datasets.alpaca_dataset
  source: b-r-ve/alpaca_fare_rules_shorter_length_1000_aed_2024_all_with_airline_30_08_24
  train_on_input: true
  split: train[:100%]
seed: null
shuffle: true
batch_size: 2
optimizer:
  _component_: torch.optim.AdamW
  weight_decay: 0.01
  lr: 0.0003
lr_scheduler:
  _component_: torchtune.modules.get_cosine_schedule_with_warmup
  num_warmup_steps: 100
loss:
  _component_: torchtune.modules.loss.CEWithChunkedOutputLoss
epochs: 1
max_steps_per_epoch: null
gradient_accumulation_steps: 16
compile: false
output_dir: /tmp/qlora_finetune_output/
metric_logger:
  _component_: torchtune.training.metric_logging.WandBLogger
  log_dir: ${output_dir}
  project: torchtune_llama3_1_8B_qlora_single_device
log_every_n_steps: 1
log_peak_memory_stats: false
device: cuda
dtype: bf16
enable_activation_checkpointing: true
profiler:
  _component_: torchtune.training.setup_torch_profiler
  enabled: false
  output_dir: ${output_dir}/profiling_outputs
  cpu: true
  cuda: true
  profile_memory: false
  with_stack: false
  record_shapes: true
  with_flops: false
  wait_steps: 5
  warmup_steps: 5
  active_steps: 2
  num_cycles: 1