empathetic_dialogues_context_classification

This model is a fine-tuned version of albert-base-v2 on empatheticDialogues dataset. It achieves the following results on the evaluation set:

• Loss: 1.6758
• F1: 0.5315
• Accuracy: 0.5315

Access to Code

https://gist.github.com/zolfaShefreie/ae87ac2944e4f7b24609e0c28fde8449

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	F1	Accuracy
2.3973	1.0	558	2.2506	0.4080	0.4080
1.7624	2.0	1116	1.8044	0.4870	0.4870
1.5362	3.0	1674	1.6312	0.5094	0.5094
1.3443	4.0	2232	1.6225	0.5145	0.5145
1.2083	5.0	2790	1.5858	0.5355	0.5355
1.079	6.0	3348	1.5721	0.5409	0.5409
0.9522	7.0	3906	1.5888	0.5308	0.5308
0.8075	8.0	4464	1.6758	0.5315	0.5315

Evaloation Result

on EmpatheticDialogues dataset (test split) and using whole history as input:

on empathetic-dialogues-contexts dataset (test split):

Framework versions

• Transformers 4.44.2
• Pytorch 2.5.0+cu121
• Datasets 3.0.2
• Tokenizers 0.19.1