Supra-Router-51M-ONNX / quantize_config.json
Sharjeelbaig's picture
Add verified INT8 ONNX model with KV cache
03081fc verified
Raw
History Blame Contribute Delete
169 Bytes
{
"modes": ["int8", "q8"],
"per_channel": true,
"reduce_range": false,
"block_size": null,
"is_symmetric": true,
"accuracy_level": null,
"quant_type": 1
}