---
datasets:
- liuhaotian/LLaVA-Pretrain
- liuhaotian/LLaVA-Instruct-150K
language:
- en
tags:
- llava
- phi
license: mit
library_name: transformers
base_model: visheratin/MC-LLaVA-3b
widget:
- text: "What animal is it?"
  src: "https://huggingface.co/datasets/mishig/sample_images/resolve/main/tiger.jpg"
- text: "Where is it?"
  src: "https://huggingface.co/datasets/mishig/sample_images/resolve/main/palace.jpg"
---

# Multi-crop LLaVA-3b

<a target="_blank" href="https://colab.research.google.com/drive/1W7JQrFXwFunAY1XvS31mwC7mrXBgGD_M">
  <img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/>
</a>

## Model details

The fundamental concept behind HelpingAI-Vision is to generate one token embedding per N parts of an image, as opposed to producing N visual token embeddings for the entire image. This approach, based on the Dolphin 2.6 Phi model and incorporating the LLaVA adapter, aims to enhance scene understanding by capturing more detailed information.

For every crop of the image, an embedding is generated using the full SigLIP encoder (size [1, 1152]). Subsequently, all N embeddings undergo processing through the LLaVA adapter, resulting in a token embedding of size [N, 2560]. Currently, these tokens lack explicit information about their position in the original image, with plans to incorporate positional information in a later update.

HelpingAI-Vision was fine-tuned from Dolphin 2.6 Phi, leveraging the vision tower from SigLIP 400M. The training process had a context length of 1200 tokens, determined by the limitations of the L4 GPUs used.

The model adopts the ChatML prompt format, suggesting its potential application in chat-based scenarios. If you have specific queries or would like further details, feel free 
```
<|im_start|>system
You are Dolphin, a helpful AI assistant.<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant
```

## How to use

**Install dependencies**

```bash
!pip install -q open_clip_torch timm einops
```

**Download modeling files**

```python
from huggingface_hub import hf_hub_download

hf_hub_download(repo_id="OEvortex/HelpingAI-Vision", filename="configuration_llava.py", local_dir="./", force_download=True)
hf_hub_download(repo_id="OEvortex/HelpingAI-Vision", filename="configuration_phi.py", local_dir="./", force_download=True)
hf_hub_download(repo_id="OEvortex/HelpingAI-Vision", filename="modeling_llava.py", local_dir="./", force_download=True)
hf_hub_download(repo_id="OEvortex/HelpingAI-Vision", filename="modeling_phi.py", local_dir="./", force_download=True)
hf_hub_download(repo_id="OEvortex/HelpingAI-Vision", filename="processing_llava.py", local_dir="./", force_download=True)
```

**Create a model**

```python
from modeling_llava import LlavaForConditionalGeneration
import torch

model = LlavaForConditionalGeneration.from_pretrained("OEvortex/HelpingAI-Vision", torch_dtype=torch.float16)
model = model.to("cuda")
```

**Create processors**

```python
from transformers import AutoTokenizer
from processing_llava import LlavaProcessor, OpenCLIPImageProcessor

tokenizer = AutoTokenizer.from_pretrained("OEvortex/HelpingAI-Vision")
image_processor = OpenCLIPImageProcessor(model.config.preprocess_config)
processor = LlavaProcessor(image_processor, tokenizer)
```

**Set image and text**

```python
from PIL import Image
import requests

image_file = "https://images.unsplash.com/photo-1439246854758-f686a415d9da"
raw_image = Image.open(requests.get(image_file, stream=True).raw)

prompt = """<|im_start|>system
A chat between a curious human and an artificial intelligence assistant.
The assistant gives helpful, detailed, and polite answers to the human's questions.
The assistant does not hallucinate and pays very close attention to the details.<|im_end|>
<|im_start|>user
<image>
Describe the image.<|im_end|>
<|im_start|>assistant
"""
```

**Process inputs**

```python
with torch.inference_mode():
  inputs = processor(prompt, raw_image, model, return_tensors='pt')

inputs['input_ids'] = inputs['input_ids'].to(model.device)
inputs['attention_mask'] = inputs['attention_mask'].to(model.device)

from transformers import TextStreamer

streamer = TextStreamer(tokenizer)
```

**Generate the data**

```python
%%time
with torch.inference_mode():
  output = model.generate(**inputs, max_new_tokens=200, do_sample=True, top_p=0.9, temperature=1.2, eos_token_id=tokenizer.eos_token_id, streamer=streamer)
print(tokenizer.decode(output[0]).replace(prompt, "").replace("<|im_end|>", ""))
```