---
license: mit
datasets:
- liuhaotian/LLaVA-Pretrain
- liuhaotian/LLaVA-Instruct-150K
language:
- en
- zh
library_name: transformers
pipeline_tag: visual-question-answering
---

# WORK IN PROGRESS

## Model type
TinyLLaVA, a tiny model (1.4B) trained using the exact training recipe of [LLaVA-1.5](https://github.com/haotian-liu/LLaVA).
We trained our TinyLLaVA using [TinyLlama](https://huggingface.co/PY007/TinyLlama-1.1B-Chat-v0.3) as our LLM backbone, and [clip-vit-large-patch14-336](https://huggingface.co/openai/clip-vit-large-patch14-336) as our vision backbone.

## Model Performance
We have evaluated TinyLLaVA on [GQA](https://cs.stanford.edu/people/dorarad/gqa/about.html), [VizWiz](https://www.vizwiz.com/), [VQAv2](https://visualqa.org/), [TextVQA](https://textvqa.org/) and [SQA](https://github.com/lupantech/ScienceQA).

|   Model   |     VQAv2      |      GQA       |       SQA      |      TextVQA   |      VizWiz    |
| -------------------- | :------------: | :------------: | :------------: | :------------: | :------------: |
|   TinyLLaVA-v1-1.4B  |      73.41     |     57.54      |     59.40      |     46.37      |      49.56      |
|   BLIP-2             |      41.00     |     41.00      |     61.00      |     42.50      |      19.60      |
|   LLaVA-v1.5-7B      |      78.50     |     62.00      |     66.80      |     61.3      |      50      |
|   LLaVA-v1.5-13B     |      80.00     |     63.30      |     71.60      |     61.3      |      53.6      |
|   Qwen-VL-7B         |      78.80     |     59.30      |     67.10      |     63.8      |      35.2      |
|   Qwen-VL-13B        |      78.20     |     57.50      |     68.20      |     61.5      |      38.9      |



More evaluations are ongoing.

## Model use
The weights have been converted to hf format.

## How to use the model

First, make sure to have `transformers >= 4.35.3`.
The model supports multi-image and multi-prompt generation. Meaning that you can pass multiple images in your prompt. Make sure also to follow the correct prompt template (`USER: xxx\nASSISTANT:`) and add the token `<image>` to the location where you want to query images:

### Using `pipeline`:

Below we used [`"bczhou/tiny-llava-v1-hf"`](https://huggingface.co/bczhou/tiny-llava-v1-hf) checkpoint.

```python
from transformers import pipeline
from PIL import Image
import requests
model_id = "bczhou/tiny-llava-v1-hf"
pipe = pipeline("image-to-text", model=model_id)
url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/ai2d-demo.jpg"
image = Image.open(requests.get(url, stream=True).raw)
prompt = "USER: <image>\nWhat does the label 15 represent? (1) lava (2) core (3) tunnel (4) ash cloud\nASSISTANT:"
outputs = pipe(image, prompt=prompt, generate_kwargs={"max_new_tokens": 200})
print(outputs[0])
>>> {"generated_text': 'USER:  \nWhat does the label 15 represent? (1) lava (2) core (3) tunnel (4) ash cloud\nASSISTANT: The label 15 represents lava, which is a type of volcanic rock."}
```

### Using pure `transformers`:

Below is an example script to run generation in `float16` precision on a GPU device:

```python
import requests
from PIL import Image
import torch
from transformers import AutoProcessor, LlavaForConditionalGeneration
model_id = "bczhou/tiny-llava-v1-hf"
prompt = "USER: <image>\nWhat are these?\nASSISTANT:"
image_file = "http://images.cocodataset.org/val2017/000000039769.jpg"
model = LlavaForConditionalGeneration.from_pretrained(
model_id,
torch_dtype=torch.float16,
low_cpu_mem_usage=True,
).to(0)
processor = AutoProcessor.from_pretrained(model_id)
raw_image = Image.open(requests.get(image_file, stream=True).raw)
inputs = processor(prompt, raw_image, return_tensors='pt').to(0, torch.float16)
output = model.generate(**inputs, max_new_tokens=200, do_sample=False)
print(processor.decode(output[0][2:], skip_special_tokens=True))
```