|
--- |
|
license: mit |
|
datasets: |
|
- liuhaotian/LLaVA-Pretrain |
|
- liuhaotian/LLaVA-Instruct-150K |
|
language: |
|
- en |
|
metrics: |
|
- accuracy |
|
- precision |
|
- recall |
|
- f1 |
|
base_model: |
|
- apple/aimv2-large-patch14-224 |
|
- apple/OpenELM |
|
pipeline_tag: image-text-to-text |
|
tags: |
|
- cpu |
|
- nano |
|
- small |
|
- tiny |
|
- llava |
|
model_size: 0.6B parameters |
|
--- |
|
|
|
**<center><span style="font-size:2em;">Tiny Llava 4 CPU 🐛</span></center>** |
|
|
|
--- |
|
|
|
### 🚀 **Model Overview** |
|
`tiny-llava-open-elm-aimv2` is a lightweight image-text-to-text model that combines **[OpenELM 270M - INSTRUCT](https://huggingface.co/apple/OpenELM-270M-Instruct)** as the LLM backbone and **[AIMv2-Large-Patch14-224-distilled (309M)](https://huggingface.co/apple/aimv2-large-patch14-224-distilled)** as the vision encoder. The model has been fine-tuned using **LoRA (Low-Rank Adaptation)** for efficient training. It was developed using the **[TinyLLaVA Factory](https://github.com/TinyLLaVA/TinyLLaVA_Factory)** codebase, which provides a modular framework for lightweight multi-modal models. |
|
|
|
The model is designed to run efficiently on **CPU**, making it ideal for resource-constrained environments. It is trained and evaluated on **POPE** and **TextVQA** benchmarks. The total model size is **0.6B parameters**. |
|
|
|
--- |
|
|
|
### Usage |
|
Execute the following test code: |
|
```python |
|
from transformers import AutoTokenizer, AutoModelForCausalLM |
|
hf_path = 'cpu4dream/llava-small-OpenELM-AIMv2-0.6B-auto' |
|
model = AutoModelForCausalLM.from_pretrained(hf_path, trust_remote_code=True) |
|
model.cuda() |
|
config = model.config |
|
tokenizer = AutoTokenizer.from_pretrained(hf_path, use_fast=False, model_max_length = config.tokenizer_model_max_length,padding_side = config.tokenizer_padding_side) |
|
prompt="What are these?" |
|
image_url="http://images.cocodataset.org/test-stuff2017/000000000001.jpg" |
|
output_text, genertaion_time = model.chat(prompt=prompt, image=image_url, tokenizer=tokenizer) |
|
print('model output:', output_text) |
|
print('runing time:', genertaion_time) |
|
``` |
|
|
|
--- |
|
|
|
### 📊 **Performance** |
|
|
|
| Model Name | VQAv2 | GQA | SQA | TextVQA | MM-VET | POPE | MME | MMMU | |
|
|:-----------------------------------------------------------:|:-----:|:-----:|:-----:|:-------:|:------:|:-----:|:------:|:-----:| |
|
| [LLaVA-1.5-7B](https://huggingface.co/llava-hf/llava-1.5-7b-hf) | 78.5 | 62.0 | 66.8 | 58.2 | 30.5 | 85.9 | 1510.7 | - | |
|
| [bczhou/TinyLLaVA-3.1B](https://huggingface.co/bczhou/TinyLLaVA-3.1B) | 79.9 | 62.0 | 69.1 | 59.1 | 32.0 | 86.4 | 1464.9 | - | |
|
| [tinyllava/TinyLLaVA-Gemma-SigLIP-2.4B](https://huggingface.co/tinyllava/TinyLLaVA-Gemma-SigLIP-2.4B) | 78.4 | 61.6 | 64.4 | 53.6 | 26.9 | 86.4 | 1339.0 | 31.7 | |
|
| [tinyllava/TinyLLaVA-Phi-2-SigLIP-3.1B](https://huggingface.co/tinyllava/TinyLLaVA-Phi-2-SigLIP-3.1B) | 80.1 | 62.1 | 73.0 | 60.3 | 37.5 | 87.2 | 1466.4 | 38.4 | |
|
| cpu4dream/llava-small-OpenELM-AIMv2-0.6B | - | - | - | 39.68 | - | 83.93 | - | - | |
|
|
|
--- |
|
|
|
### 🔗 **References** |
|
- [OpenELM](https://huggingface.co/apple/OpenELM) |
|
- [AIMv2-Large-Patch14-224](https://huggingface.co/apple/aimv2-large-patch14-224) |
|
- [TinyLLaVA Factory GitHub](https://github.com/TinyLLaVA/TinyLLaVA_Factory) |
