Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,7 @@
|
|
1 |
---
|
2 |
license: mit
|
3 |
---
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: mit
|
3 |
---
|
4 |
+
converted to onnx from here https://huggingface.co/microsoft/phi-2
|
5 |
+
fp16 with weights block quantized to int4 whenever visible.
|
6 |
+
Last 3 layers are kept in fp32 to avoid fp16 overflow.
|
7 |
+
Some outputs are casted to fp32 to make it friendly for ort-web.
|