4-bit quant using Intel AutoRound, This is essentially the default W4A16 scheme

quant details

{
  "bits": 4,
  "group_size": 128,
  "sym": true,
  "data_type": "int",
  "batch_size": 1,
  "gradient_accumulate_steps": 8,
  "autoround_version": "0.9.2",
  "quant_method": "auto-round",
  "packing_format": "auto_round:auto_gptq"
}

Downloads last month: 8

Safetensors

Model size

1B params

Tensor type

I32

BF16

F16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for hoborific/BereavedCompound-v1.0-24b-W4A16-AutoRound

Base model

FlareRebellion/BereavedCompound-v1.0-24b

Quantized

(5)

this model