|
--- |
|
language: tr |
|
license: mit |
|
--- |
|
|
|
# Turkish Named Entity Recognition (NER) Quantized Model |
|
|
|
This model is the dynamically quantized version of the model |
|
(https://akdeniz27/bert-base-turkish-cased-ner) |
|
|
|
# How to use: |
|
``` |
|
# First install "optimum[onnxruntime]": |
|
!pip install "optimum[onnxruntime]" |
|
|
|
# and import "ORTModelForTokenClassification": |
|
from transformers import AutoTokenizer, pipeline |
|
from optimum.onnxruntime import ORTModelForTokenClassification |
|
|
|
model = ORTModelForTokenClassification.from_pretrained("akdeniz27/bert-base-turkish-cased-ner-quantized", file_name="model_quantized.onnx") |
|
tokenizer = AutoTokenizer.from_pretrained("akdeniz27/bert-base-turkish-cased-ner-quantized") |
|
ner = pipeline('ner', model=model, tokenizer=tokenizer, aggregation_strategy="first") |
|
ner("your text here") |
|
``` |
|
Pls refer (https://github.com/akdeniz27/dynamic_quantization) for details of quantization. |
