falcon-7b-finetuned-mental-health-conversational
This model is a fine-tuned version of ybelkada/falcon-7b-sharded-bf16 on the custom heliosbrahma/mental_health_conversational_dataset dataset.
Model description
This model is fine-tuned on custom mental health conversational dataset. The rationale behind this is to answer mental health related queries that can be factually verified without responding gibberish words.
Intended uses & limitations
The model was trained on the dataset which may contain sensitive information related to mental health. It is important to note that while mental health chatbots built using this model can be helpful, they are not a replacement for professional mental health care.
Training and evaluation data
Model was trained on custom heliosbrahma/mental_health_conversational_dataset dataset which 154 rows of conversational pair of questions and answers.
Training procedure
Model was trained using QLoRA technique to fine-tune on a custom dataset on free-tier GPU available in Google Colab.
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0002
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.03
- training_steps: 180
Framework versions
- Transformers 4.31.0
- Pytorch 2.0.1+cu118
- Datasets 2.13.1
- Tokenizers 0.13.3