File size: 411 Bytes
7469638 ebf7e08 fe53831 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 |
---
license: apache-2.0
base_model:
- intfloat/multilingual-e5-small
---
the original model is OriginalModel.mlpackage with float32
# the quantization model info :
## quantization and input maxlength
coreML: using linear quantize nbit=8
input max = 128
## note
i tried turn it into float 16, but it changed too much for its prediction. quantization using linear nbit=8, it works almost like the original. |