|
|
|
--- |
|
library_name: peft |
|
tags: |
|
- nlp |
|
- code |
|
- instruct |
|
- llama |
|
datasets: |
|
- HuggingFaceH4/no_robots |
|
base_model: google/gemma-2-2b-it |
|
license: apache-2.0 |
|
--- |
|
|
|
|
|
# monsterapi/gemma-2-2b-norobots |
|
|
|
**Base Model for Fine-tuning:** [google/gemma-2-2b-it](https://huggingface.co/google/gemma-2-2b-it) |
|
**Service Used:** [MonsterAPI](https://monsterapi.ai) |
|
**License:** Apache-2.0 |
|
|
|
## Overview |
|
|
|
`monsterapi/gemma-2-2b-norobots` is a fine-tuned language model designed to improve instruction-following capabilities. The model was trained using the "No Robots" dataset, a high-quality set of 10,000 instructions and demonstrations curated by expert human annotators. This fine-tuning process enhances the base model's performance in understanding and executing single-turn instructions, similar to the goals outlined in OpenAI's InstructGPT. |
|
|
|
### Dataset Details |
|
|
|
**Dataset Summary:** |
|
The "No Robots" dataset is a collection of 10,000 high-quality instructions and demonstrations created by skilled human annotators. The dataset is modeled after the instruction dataset described in OpenAI's InstructGPT paper. It mainly includes single-turn instructions across various categories, aiming to improve the instruction-following capabilities of language models during supervised fine-tuning (SFT). |
|
|
|
## Fine-tuning Details |
|
|
|
**Fine-tuned Model Name:** `monsterapi/gemma-2-2b-norobots` |
|
**Training Time:** 31 minutes |
|
**Cost:** $1.10 |
|
**Epochs:** 1 |
|
**Gradient Accumulation Steps:** 32 |
|
|
|
The model was fine-tuned using MonsterAPI's finetuning service, optimizing the base model `google/gemma-2-2b-it` to perform better on instruction-following tasks. |
|
|
|
## Hyperparameters & Additional Details |
|
|
|
- **Base Model:** `google/gemma-2-2b-it` |
|
- **Dataset:** No Robots (10,000 instructions and demonstrations) |
|
- **Training Duration:** 31 minutes |
|
- **Cost per Epoch:** $1.10 |
|
- **Total Finetuning Cost:** $1.10 |
|
- **Gradient Accumulation Steps:** 32 |
|
|
|
## Use Cases |
|
|
|
This model is well-suited for tasks that require improved instruction-following capabilities, such as: |
|
- Chatbots and virtual assistants |
|
- Content creation tools |
|
- Automated customer support systems |
|
- Task automation in various industries |
|
|
|
## How to Use |
|
|
|
You can load the model directly using the Hugging Face Transformers library: |
|
|
|
```python |
|
from transformers import AutoModelForCausalLM, AutoTokenizer |
|
|
|
model_name = "monsterapi/gemma-2-2b-norobots" |
|
tokenizer = AutoTokenizer.from_pretrained(model_name) |
|
model = AutoModelForCausalLM.from_pretrained(model_name) |
|
|
|
# Example usage |
|
input_text = "Explain the concept of supervised fine-tuning in simple terms." |
|
inputs = tokenizer(input_text, return_tensors="pt") |
|
outputs = model.generate(**inputs) |
|
print(tokenizer.decode(outputs[0], skip_special_tokens=True)) |
|
``` |
|
|
|
## Acknowledgements |
|
|
|
The fine-tuning process was carried out using MonsterAPI's finetuning service, which offers a seamless experience for optimizing large language models. |
|
|
|
## Contact |
|
|
|
For further details or queries, please contact [MonsterAPI](https://monsterapi.ai) or visit the official documentation. |
|
|
