quant_meta.json

{
  "model_id": "OpenGVLab/InternVL2_5-4B",
  "family": "internvl25",
  "method": "awq_internvl_merge",
  "w_bit": 4,
  "group_size": 128,
  "quant_time_s": 410.5,
  "n_calib": 64,
  "calib_seq_len": 256,
  "notes": "Inner InternLM2 AWQ-quantized, then re-attached to full InternVLChatModel. Vision encoder + mlp1 connector remain FP16."
}