quantization:
  initial_format: modelopt-nvfp4
  release_format: compressed-tensors-nvfp4-pack-quantized
  scheme: NVFP4 W4A4
  group_size: 16
calibration:
  dataset: abisee/cnn_dailymail
  config: 3.0.0
  samples: 20
  sequence_length: 8192
preserve_bf16:
  - lm_head
  - model.language_model.embed_tokens
  - re:.*visual.*
  - re:.*conv1d.*
  - re:^mtp.*
mtp:
  tensors: 15
  source: same Huihui checkpoint
validation:
  gpu: RTX 5090 32GB
  power_limit_w: 450
  text: passed
  mtp_n3: passed