deepseek-ocr-gptq-4bit / quantization_config.json
SamMikaelson's picture
Upload GPTQ 4-bit packed model (75.0% size reduction, 4.0x compression)
42a1727 verified
raw
history blame contribute delete
371 Bytes
{
"bits": 4,
"group_size": 128,
"damp_percent": 0.01,
"method": "gptq",
"desc_act": false,
"sym": true,
"true_sequential": true,
"packed": true,
"model_name_or_path": "deepseek-ai/DeepSeek-OCR",
"quantization_method": "gptq",
"original_size_gb": 6.21398688107729,
"quantized_size_gb": 1.5534955263137817,
"compression_ratio": 4.000003074242624
}