xinhe commited on
Commit
6556958
1 Parent(s): 3d3263c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +0 -8
README.md CHANGED
@@ -37,12 +37,8 @@ The calibration dataloader is the train dataloader. The default calibration samp
37
 
38
  ### Test result
39
 
40
- - Batch size = 8
41
- - [Amazon Web Services](https://aws.amazon.com/) c6i.xlarge (Intel ICE Lake: 4 vCPUs, 8g Memory) instance.
42
-
43
  | |INT8|FP32|
44
  |---|:---:|:---:|
45
- | **Throughput (samples/sec)** |12.506|9.436|
46
  | **Accuracy (eval-f1)** |0.8893|0.8897|
47
  | **Model size (MB)** |215|448|
48
 
@@ -54,7 +50,3 @@ int8_model = OptimizedModel.from_pretrained(
54
  'Intel/xlnet-base-cased-mrpc-int8-static',
55
  )
56
  ```
57
-
58
- Notes:
59
- - The INT8 model has better performance than the FP32 model when the CPU is fully occupied. Otherwise, there will be the illusion that INT8 is inferior to FP32.
60
-
 
37
 
38
  ### Test result
39
 
 
 
 
40
  | |INT8|FP32|
41
  |---|:---:|:---:|
 
42
  | **Accuracy (eval-f1)** |0.8893|0.8897|
43
  | **Model size (MB)** |215|448|
44
 
 
50
  'Intel/xlnet-base-cased-mrpc-int8-static',
51
  )
52
  ```