scripts/quant_phi4.yaml

# Model arguments
model:
  _component_: torchtune.models.phi4.lora_phi4_14b

checkpointer:
  _component_: torchtune.training.FullModelHFCheckpointer
  checkpoint_dir: {{model_output_dir}}
  checkpoint_files: [
    ft-model-00001-of-00006.safetensors,
    ft-model-00002-of-00006.safetensors,
    ft-model-00003-of-00006.safetensors,
    ft-model-00004-of-00006.safetensors,
    ft-model-00005-of-00006.safetensors,
    ft-model-00006-of-00006.safetensors
  ]
  recipe_checkpoint: null
  #output_dir: {{model_output_dir}}/quant
  output_dir: quant  
  model_type: PHI3_MINI

device: cuda
dtype: bf16
seed: 1234

quantizer:
  _component_: torchtune.training.quantization.Int8DynActInt4WeightQuantizer
  groupsize: 256