kasrahabib/KM45L6V2OC

This model is a fine-tuned version of sentence-transformers/all-MiniLM-L6-v2, for classifying softwrae requirments into functional (F) and Non-functional (NF) types, on Software Requirements Dataset (SWARD). It achieves the following results on the evaluation set:

• Train Loss: 0.0107
• Validation Loss: 0.0404
• Epoch: 14
• Final Macro F1-score: 0.99

Labels: 0 or F -> Functional; 1 or NF -> Non-functional;

Usage Pipeline

from transformers import pipeline

frame_work = 'tf'
task = 'text-classification'
model_ckpt = 'kasrahabib/KM45L6V2OC'

software_requirment_cls = pipeline(task = task, model = model_ckpt, framework = frame_work)

example_1_f = 'The START NEW PROJECT function shall allow the user to create a new project.'
example_2_nf = 'The email string consists of [email protected] and is less than 31 characters in length and is not empty.'
software_requirment_cls([example_1_f, example_2_nf])

[{'label': 'F', 'score': 0.9998922348022461},
 {'label': 'NF', 'score': 0.999846339225769}]

Model Inference:

import numpy as np
from transformers import AutoTokenizer, TFAutoModelForSequenceClassification

model_ckpt = 'kasrahabib/KM45L6V2OC'
tokenizer = AutoTokenizer.from_pretrained(model_ckpt)
model = TFAutoModelForSequenceClassification.from_pretrained(model_ckpt)

example_1_f = 'The START NEW PROJECT function shall allow the user to create a new project.'
example_2_nf = 'The email string consists of [email protected] and is less than 31 characters in length and is not empty.'
requirements = [example_1_f, example_2_nf]

encoded_requirements = tokenizer(requirements, return_tensors = 'np', padding = 'longest')

y_pred = model(encoded_requirements).logits
classifications = np.argmax(y_pred, axis = 1)

classifications = [model.config.id2label[output] for output in classifications]
print(classifications)

['F', 'NF']

Usage Locally Downloaded (e.g., GitHub):

1 - Clone the repository:

git lfs install
git clone url_of_repo

2 - Locate the path to the downloaded directory
3 - Write the link to the path in the model_ckpt variable

Then modify the code as below:

import numpy as np
from transformers import AutoTokenizer, TFAutoModelForSequenceClassification

model_ckpt = 'rest_of_the_path/KM45L6V2OC'
tokenizer = AutoTokenizer.from_pretrained(model_ckpt)
model = TFAutoModelForSequenceClassification.from_pretrained(model_ckpt)

example_1_f = 'The START NEW PROJECT function shall allow the user to create a new project.'
example_2_nf = 'The email string consists of [email protected] and is less than 31 characters in length and is not empty.'
requirements = [example_1_f, example_2_nf]

encoded_requirements = tokenizer(requirements, return_tensors = 'np', padding = 'longest')

y_pred = model(encoded_requirements).logits
classifications = np.argmax(y_pred, axis = 1)

classifications = [model.config.id2label[output] for output in classifications]
print(classifications)

[{'label': 'F', 'score': 0.9998922348022461},
 {'label': 'NF', 'score': 0.999846339225769}]

Training hyperparameters

The following hyperparameters were used during training:

• optimizer: {'name': 'Adam', 'weight_decay': None, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': True, 'is_legacy_optimizer': False, 'learning_rate': {'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 2e-05, 'decay_steps': 9030, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}}, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-08, 'amsgrad': False}
• training_precision: float32

Framework versions

• Transformers 4.26.1
• TensorFlow 2.11.0
• Datasets 2.10.1
• Tokenizers 0.13.2