update int8 onnx model and readme

Signed-off-by: dujun <jun.du@intel.com>

README.md

@@ -63,8 +63,8 @@ The original fp32 model comes from the fine-tuned model [bart-large-mrpc](https:
 
 |   |INT8|FP32|
 |---|:---:|:---:|
-| **Accuracy (eval-f1)** |0.9134|0.9120|
-| **Model size (MB)**  |395|1555|
+| **Accuracy (eval-f1)** |0.9236|0.9120|
+| **Model size (MB)**  |764|1555|
 
 
 #### Load ONNX model:

model.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c6f32f2e86da30d0634b16a761c6b435a2a3bf93d060e55f0bdac19df085f11c
-size 413817970
+oid sha256:e3c41e82c6f0f620869a5dc6bb6c4869bcaddf7032f1e8304630739891803329
+size 800224809