---
base_model: HuggingFaceTB/SmolLM2-135M
library_name: transformers
model_name: SmolLM2-FT-MyDataset
tags:
- generated_from_trainer
- smol-course
- module_1
- trl
- sft
licence: license
---

# Model Card for SmolLM2-FT-MyDataset

This model is a fine-tuned version of [HuggingFaceTB/SmolLM2-135M](https://huggingface.co/HuggingFaceTB/SmolLM2-135M).
It has been trained using [TRL](https://github.com/huggingface/trl).

## Quick start

```python
from transformers import pipeline

question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="riswanahamed/SmolLM2-FT-MyDataset", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])
```

## Training procedure
### Training Methods

#### **What I Did**
I fine-tuned a pre-trained language model using the Hugging Face `transformers` library. The base model was adapted to perform better on specific task by training it on a domain-specific dataset.

#### **How I Did It**

**Fine-Tuning Setup**:
   - Configured the model training parameters, including the learning rate, batch size, and number of steps.
   - Used `SFTTrainer` from Hugging Face for seamless training with built-in evaluation capabilities.
   - Trained the model for 1 epoch to prevent overfitting, as the dataset was relatively small and hardware resources were limited.

**Training Environment**:
   - The training was performed in Google Colab using a CPU/GPU environment. 
   - Adjusted batch sizes and learning rates to balance between performance and available resources.

4. **Evaluation**:
   - Monitored training loss and validation loss at regular intervals to ensure the model was learning effectively.
   - Evaluated the model using metrics like [accuracy, F1 score, or other task-specific metrics].

5. **Saving the Model**:
   - The fine-tuned model was saved to a specified output directory for reuse.

#### **What the User Should Do**
1. **Use the Model**: 
   - Load the model using the Hugging Face `transformers` library.
   - Tokenize your inputs and pass them to the model for inference.
   - If your task or domain differs, fine-tune the model further on your dataset.
   - Follow the same process: prepare the dataset, set training configurations, and monitor evaluation metrics.

2. **Experiment with Parameters**:
   - If you have access to better hardware, experiment with larger batch sizes or additional epochs to improve results.
   - Use hyperparameter tuning to find the best configuration for your use case.

### Framework versions

- TRL: 0.13.0
- Transformers: 4.47.1
- Pytorch: 2.5.1+cu121
- Datasets: 3.2.0
- Tokenizers: 0.21.0

## Citations



Cite TRL as:
    
```bibtex
@misc{vonwerra2022trl,
	title        = {{TRL: Transformer Reinforcement Learning}},
	author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
	year         = 2020,
	journal      = {GitHub repository},
	publisher    = {GitHub},
	howpublished = {\url{https://github.com/huggingface/trl}}
}
```