# Config for single device QLoRA with lora_finetune_distributed.py
# using a Llama2 7B model
#
# This config assumes that you've run the following command before launching
# this run:
#   tune download meta-llama/Llama-2-7b-hf --output-dir /tmp/Llama-2-7b-hf --ignore-patterns "*.bin" --hf-token <HF_TOKEN>
#
# To launch on 2 devices, run the following command from root:
#   tune run --nnodes 1 --nproc_per_node 2 lora_finetune_distributed --config llama2/7B_qlora
#
# You can add specific overrides through the command line. For example
# to override the checkpointer directory while launching training
# you can run:
#   tune run --nnodes 1 --nproc_per_node 2 lora_finetune_distributed --config llama2/7B_qlora checkpointer.checkpoint_dir=<YOUR_CHECKPOINT_DIR>
#
# This config works best when the model is being fine-tuned on 2+ GPUs.
# For single device LoRA finetuning please use 7B_lora_single_device.yaml
# or 7B_qlora_single_device.yaml

output_dir: /tmp/torchtune/llama2_7B/qlora # /tmp may be deleted by your system. Change it to your preference.

# Model Arguments
model:
  _component_: torchtune.models.llama2.qlora_llama2_7b
  lora_attn_modules: ['q_proj', 'v_proj', 'output_proj']
  apply_lora_to_mlp: True
  apply_lora_to_output: False
  lora_rank: 8  # higher increases accuracy and memory
  lora_alpha: 16  # usually alpha=2*rank
  lora_dropout: 0.0

tokenizer:
  _component_: torchtune.models.llama2.llama2_tokenizer
  path: /tmp/Llama-2-7b-hf/tokenizer.model
  max_seq_len: null

checkpointer:
  _component_: torchtune.training.FullModelHFCheckpointer
  checkpoint_dir: /tmp/Llama-2-7b-hf
  checkpoint_files: [
    model-00001-of-00002.safetensors,
    model-00002-of-00002.safetensors
  ]
  adapter_checkpoint: null
  recipe_checkpoint: null
  output_dir: ${output_dir}
  model_type: LLAMA2
resume_from_checkpoint: False
save_adapter_weights_only: False

# Dataset and Sampler
dataset:
  _component_: torchtune.datasets.alpaca_cleaned_dataset
  packed: False  # True increases speed
  split: train[:95%]
seed: null
shuffle: True

# Validation
run_val_every_n_steps: null  # Change to an integer to enable validation every N steps
dataset_val:
  _component_: torchtune.datasets.alpaca_cleaned_dataset
  split: train[95%:]
batch_size_val: ${batch_size}

# Optimizer and Scheduler
optimizer:
  _component_: torch.optim.AdamW
  fused: True
  weight_decay: 0.01
  lr: 3e-4
lr_scheduler:
  _component_: torchtune.training.lr_schedulers.get_cosine_schedule_with_warmup
  num_warmup_steps: 100

loss:
  _component_: torchtune.modules.loss.LinearCrossEntropyLoss

fsdp:
  cpu_offload: False

# Training
epochs: 1
batch_size: 2
max_steps_per_epoch: null
gradient_accumulation_steps: 8  # Use to increase effective batch size
clip_grad_norm: null
compile: False  # torch.compile the model + loss, True increases speed + decreases memory

# Logging
metric_logger:
  _component_: torchtune.training.metric_logging.DiskLogger
  log_dir: ${output_dir}/logs
log_every_n_steps: 1
log_peak_memory_stats: True
log_level: INFO  # DEBUG, WARN, etc.

# Environment
device: cuda
dtype: bf16
enable_activation_checkpointing: True  # True reduces memory
enable_activation_offloading: False  # True reduces memory


# Profiler (disabled)
profiler:
  _component_: torchtune.training.setup_torch_profiler
  enabled: False

  #Output directory of trace artifacts
  output_dir: ${output_dir}/profiling_outputs

  #`torch.profiler.ProfilerActivity` types to trace
  cpu: True
  cuda: True

  #trace options passed to `torch.profiler.profile`
  profile_memory: False
  with_stack: False
  record_shapes: True
  with_flops: False

  # `torch.profiler.schedule` options:
  # wait_steps -> wait, warmup_steps -> warmup, active_steps -> active, num_cycles -> repeat
  wait_steps: 5
  warmup_steps: 3
  active_steps: 2
  num_cycles: 1
