OpenAssistant-SFT-7-Llama-30B-GPTQ / quantize_config.json
TheBloke's picture
Upload new GPTQs with varied parameters
48dc04c
raw
history blame
133 Bytes
{
"bits": 4,
"group_size": -1,
"damp_percent": 0.01,
"desc_act": true,
"sym": true,
"true_sequential": true
}