Final model — trained on GoEmotions + ISEAR + MentalHealth Reddit

README.md

@@ -0,0 +1,66 @@
+---
+library_name: transformers
+license: apache-2.0
+base_model: distilroberta-base
+tags:
+- generated_from_trainer
+metrics:
+- accuracy
+model-index:
+- name: mindmate-emotion-classifier
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# mindmate-emotion-classifier
+
+This model is a fine-tuned version of [distilroberta-base](https://huggingface.co/distilroberta-base) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.5719
+- Accuracy: 0.7976
+- F1 Weighted: 0.7962
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 2e-05
+- train_batch_size: 32
+- eval_batch_size: 64
+- seed: 42
+- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
+- lr_scheduler_type: linear
+- lr_scheduler_warmup_steps: 0.1
+- num_epochs: 3
+- mixed_precision_training: Native AMP
+
+### Training results
+
+| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 Weighted |
+|:-------------:|:-----:|:----:|:---------------:|:--------:|:-----------:|
+| 0.6452        | 1.0   | 2268 | 0.6177          | 0.7778   | 0.7804      |
+| 0.5152        | 2.0   | 4536 | 0.5764          | 0.7867   | 0.7849      |
+| 0.4566        | 3.0   | 6804 | 0.5719          | 0.7976   | 0.7962      |
+
+
+### Framework versions
+
+- Transformers 5.0.0
+- Pytorch 2.10.0+cu128
+- Datasets 4.0.0
+- Tokenizers 0.22.2