83 lines
3.0 KiB
Markdown
83 lines
3.0 KiB
Markdown
---
|
|
license: mit
|
|
datasets:
|
|
- lgaalves/camel-ai-physics
|
|
language:
|
|
- en
|
|
pipeline_tag: text-generation
|
|
---
|
|
|
|
|
|
|
|
# gpt2-xl-camel-ai-physics (1.5B)
|
|
|
|
**lgaalves/gpt2-xl_camel-ai-physics** is an instruction fine-tuned model based on the GPT-2 transformer architecture.
|
|
|
|
|
|
### Benchmark Metrics
|
|
|
|
| Metric |lgaalves/gpt2-xl_camel-ai-physics |gpt2-xl (base) |
|
|
|-----------------------|-------|-------|
|
|
| Avg. | 36.51 | **36.66** |
|
|
| ARC (25-shot) | 29.52 | **30.29** |
|
|
| HellaSwag (10-shot) | 50.62 | **51.38** |
|
|
| MMLU (5-shot) | **26.79** | 26.43 |
|
|
| TruthfulQA (0-shot) | **39.12** | 38.54 |
|
|
|
|
|
|
We use state-of-the-art [Language Model Evaluation Harness](https://github.com/EleutherAI/lm-evaluation-harness) to run the benchmark tests above, using the same version as the HuggingFace LLM Leaderboard. Please see below for detailed instructions on reproducing benchmark results.
|
|
|
|
### Model Details
|
|
|
|
* **Trained by**: Luiz G A Alves
|
|
* **Model type:** **lgaalves/gpt2-xl_camel-ai-physics** is an auto-regressive language model based on the GPT-2 transformer architecture.
|
|
* **Language(s)**: English
|
|
|
|
### How to use:
|
|
|
|
```python
|
|
# Use a pipeline as a high-level helper
|
|
>>> from transformers import pipeline
|
|
>>> pipe = pipeline("text-generation", model="lgaalves/gpt2-xl_camel-ai-physics")
|
|
>>> question = "What is a large language model?"
|
|
>>> answer = pipe(question)
|
|
>>> print(answer[0]['generated_text'])
|
|
|
|
```
|
|
|
|
or, you can load the model direclty using:
|
|
|
|
```python
|
|
# Load model directly
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained("lgaalves/gpt2-xl_camel-ai-physics")
|
|
model = AutoModelForCausalLM.from_pretrained("lgaalves/gpt2-xl_camel-ai-physics")
|
|
```
|
|
|
|
### Training Dataset
|
|
|
|
`lgaalves/gpt2-xl_camel-ai-physics` trained on the GPT4 generated dataset [lgaalves/camel-physics](https://huggingface.co/datasets/lgaalves/camel-physics).
|
|
|
|
### Training Procedure
|
|
|
|
`lgaalves/gpt2-xl_camel-ai-physics` was instruction fine-tuned using LoRA on 1 Tesla V100-SXM2-16GB. It took about 3 hours to train it.
|
|
|
|
|
|
# Intended uses, limitations & biases
|
|
|
|
You can use the raw model for text generation or fine-tune it to a downstream task. The model was not extensively tested and may produce false information. It contains a lot of unfiltered content from the internet, which is far from neutral.
|
|
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
|
|
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_lgaalves__gpt-2-xl_camel-ai-physics)
|
|
|
|
| Metric | Value |
|
|
|-----------------------|---------------------------|
|
|
| Avg. | 29.9 |
|
|
| ARC (25-shot) | 29.52 |
|
|
| HellaSwag (10-shot) | 50.62 |
|
|
| MMLU (5-shot) | 26.79 |
|
|
| TruthfulQA (0-shot) | 39.12 |
|
|
| Winogrande (5-shot) | 57.54 |
|
|
| GSM8K (5-shot) | 0.15 |
|
|
| DROP (3-shot) | 5.57 |
|