scripts/knowledge_distillation_qwen2.yaml

# Model Arguments
model:
  _component_: torchtune.models.qwen2.lora_qwen2_0_5b
  lora_attn_modules: ['q_proj', 'k_proj', 'v_proj']
  apply_lora_to_mlp: False
  lora_rank: 32
  lora_alpha: 64

teacher_model:
  _component_: torchtune.models.qwen2.qwen2_1_5b

tokenizer:
  _component_: torchtune.models.qwen2.qwen2_tokenizer
  path: {{student_model_dir}}/vocab.json
  merges_file: {{student_model_dir}}/merges.txt
  max_seq_len: null

checkpointer:
  _component_: torchtune.training.FullModelHFCheckpointer
  checkpoint_dir: {{student_model_dir}}
  checkpoint_files: [
    model.safetensors
  ]
  recipe_checkpoint: null
  output_dir: {{model_output_dir}}
  model_type: QWEN2

teacher_checkpointer:
  _component_: torchtune.training.FullModelHFCheckpointer
  checkpoint_dir: {{teacher_model_dir}}
  checkpoint_files: [
    model.safetensors
  ]
  recipe_checkpoint: null
  output_dir: teacher-tmp
  model_type: QWEN2

resume_from_checkpoint: False

# Dataset and Sampler
dataset:
    _component_: torchtune.datasets.instruct_dataset
    source: json
    data_files: {{train_path}}
    column_map:
        input: instruction
        output: output
    train_on_input: False
    packed: False
    split: train
seed: null
shuffle: True

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

loss:
  _component_: torchtune.modules.loss.CEWithChunkedOutputLoss

kd_loss:
  _component_: torchtune.modules.loss.ForwardKLWithChunkedOutputLoss
kd_ratio: 0.5

# Training
epochs: 1
max_steps_per_epoch: null
batch_size: 4
gradient_accumulation_steps: 4
compile: False

# Logging
output_dir: {{log_dir}}/kd_output
metric_logger:
  _component_: torchtune.training.metric_logging.{{metric_logger}}
  log_dir: {{log_dir}}/training_logs
log_every_n_steps: 1
log_peak_memory_stats: True

# Environment
device: cuda
dtype: bf16
enable_activation_checkpointing: False