Commit
·
aa81b56
1
Parent(s):
994a4c4
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,5 +1,12 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
| 4 |
# Quantized BERT-base MNLI model with 90% of usntructured sparsity
|
| 5 |
The pruned and quantized model in the OpenVINO IR. The pruned model was taken from this [source](https://huggingface.co/neuralmagic/oBERT-12-downstream-pruned-unstructured-90-mnli) and quantized with the code below using HF Optimum for OpenVINO:
|
|
@@ -20,7 +27,7 @@ def preprocess_function(examples, tokenizer):
|
|
| 20 |
# Load the default quantization configuration detailing the quantization we wish to apply
|
| 21 |
quantization_config = OVConfig()
|
| 22 |
# Instantiate our OVQuantizer using the desired configuration
|
| 23 |
-
quantizer = OVQuantizer.from_pretrained(model)
|
| 24 |
# Create the calibration dataset used to perform static quantization
|
| 25 |
|
| 26 |
calibration_dataset = quantizer.get_calibration_dataset(
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
datasets:
|
| 4 |
+
- mnli
|
| 5 |
+
metrics:
|
| 6 |
+
- accuracy
|
| 7 |
+
tags:
|
| 8 |
+
- sequence-classification
|
| 9 |
+
- int8
|
| 10 |
---
|
| 11 |
# Quantized BERT-base MNLI model with 90% of usntructured sparsity
|
| 12 |
The pruned and quantized model in the OpenVINO IR. The pruned model was taken from this [source](https://huggingface.co/neuralmagic/oBERT-12-downstream-pruned-unstructured-90-mnli) and quantized with the code below using HF Optimum for OpenVINO:
|
|
|
|
| 27 |
# Load the default quantization configuration detailing the quantization we wish to apply
|
| 28 |
quantization_config = OVConfig()
|
| 29 |
# Instantiate our OVQuantizer using the desired configuration
|
| 30 |
+
quantizer = OVQuantizer.from_pretrained(model, feature="sequence-classification")
|
| 31 |
# Create the calibration dataset used to perform static quantization
|
| 32 |
|
| 33 |
calibration_dataset = quantizer.get_calibration_dataset(
|