README.md

---
base_model: meta-llama/Llama-3.1-8B-Instruct
library_name: transformers
model_name: Llama-3.1-8B-KAM
tags:
- trl
- Llama
- sft
- generated_from_trainer
licence: license
---

<!-- 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. -->

# Model Card for Llama-3.1-8B-KAM

This model is a fine-tuned version of [meta-llama/Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct) on the None dataset.

## Model description

More information needed

## 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="MaRyAm1295/Llama-3.1-8B-KAM", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])
```

## Training procedure

This model was trained with SFT.

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 0.0002
- train_batch_size: 1
- eval_batch_size: 8
- seed: 3407
- gradient_accumulation_steps: 16
- total_train_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 20
- training_steps: 500
- mixed_precision_training: Native AMP

### Training results
#### Step &nbsp;&nbsp;	Training Loss
  - 50	 &nbsp;&nbsp;&nbsp;   2.158200
  - 100	 &nbsp;&nbsp;&nbsp;   1.845900
  - 150	 &nbsp;&nbsp;&nbsp;   1.832200
  - 200	 &nbsp;&nbsp;&nbsp;   1.805300
  - 250	 &nbsp;&nbsp;&nbsp;   1.783800
  - 300	 &nbsp;&nbsp;&nbsp;   1.767500
  - 350	 &nbsp;&nbsp;&nbsp;   1.744800
  - 400	 &nbsp;&nbsp;&nbsp;   1.745600
  - 450	 &nbsp;&nbsp;&nbsp;   1.749500
  - 500	 &nbsp;&nbsp;&nbsp;   1.756100

### Framework versions

- TRL: 0.12.0
- Transformers: 4.46.2
- Pytorch: 2.4.0
- Datasets: 3.0.1
- Tokenizers: 0.20.0