Chakshu/conversation_terminator_classifier

This model is a fine-tuned version of google/mobilebert-uncased on an unknown dataset. It achieves the following results on the evaluation set:

• Train Loss: 0.0364
• Train Binary Accuracy: 0.9915
• Epoch: 8

Example Usage

from transformers import AutoTokenizer, TFBertForSequenceClassification, BertTokenizer
import tensorflow as tf

model_name = 'Chakshu/conversation_terminator_classifier' 

tokenizer = BertTokenizer.from_pretrained(model_name)
model = TFBertForSequenceClassification.from_pretrained(model_name)
inputs = tokenizer("I will talk to you later", return_tensors="np", padding=True)
outputs = model(inputs.input_ids, inputs.attention_mask)
probabilities = tf.nn.sigmoid(outputs.logits)

# Round the probabilities to the nearest integer to get the class prediction
predicted_class = tf.round(probabilities)
print("The last message by the user indicates that the conversation has", "'ENDED'" if int(predicted_class.numpy()) == 1 else "'NOT ENDED'")

Model description

Classifies if the user is ending the conversation or wanting to continue it.

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

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': False, 'is_legacy_optimizer': False, 'learning_rate': 2e-05, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False}
• training_precision: float32

Training results

Train Loss	Train Binary Accuracy	Epoch
0.2552	0.9444	0
0.1295	0.9872	1
0.0707	0.9872	2
0.0859	0.9829	3
0.0484	0.9872	4
0.0363	0.9957	5
0.0209	1.0	6
0.0268	0.9957	7
0.0364	0.9915	8

Framework versions

• Transformers 4.28.0
• TensorFlow 2.12.0
• Datasets 2.12.0
• Tokenizers 0.13.3