base_model: JackFram/llama-68m,

sequence_len: 1024
load_in_8bit: true
adapter: lora
lora_r: 64
lora_alpha: 32
lora_dropout: 0.1
lora_target_linear: true
rl: dpo

datasets:
  - path: arcee-ai/distilabel-intel-orca-dpo-pairs-binarized
    type: chatml.ultra
    split: train

num_epochs: 1
micro_batch_size: 4
gradient_accumulation_steps: 1
output_dir: ./outputs/lora-out
learning_rate: 0.00001
optimizer: paged_adamw_8bit
lr_scheduler: cosine
max_steps": 20
save_steps: 10
warmup_steps: 5
gradient_checkpointing: True
gradient_checkpointing_kwargs:
  use_reentrant: false
special_tokens:
  pad_token: <|end_of_text|>