HF-hub

13:23 · Aug 8, 2024 · Thu

Model tree for hugging-quants/Meta-Llama-3.1-405B-BNB-NF4-BF16
Model Information
The Meta Llama 3.1 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction tuned generative models in 8B, 70B and 405B sizes (text in/text out). The Llama 3.1 instruction tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source and closed chat models on common industry benchmarks.

This repository contains meta-llama/Meta-Llama-3.1-405B quantized using bitsandbytes from BF16 down to NF4 with a block size of 64, and storage type torch.bfloat16.

Model Usage
In order to run the inference with Llama 3.1 405B BNB in NF4, around 220 GiB of VRAM are needed only for loading the model checkpoint, without including the KV cache or the CUDA graphs, meaning that there should be a bit over that VRAM available.

In order to use the current quantized model, support is offered for different solutions as transformers, or text-generation-inference.

🤗 transformers
In order to run the inference with Llama 3.1 405B BNB in NF4, both torch and bitsandbytes need to be installed as:

pip install "torch>=2.0.0" bitsandbytes --upgrade

Then, the latest version of transformers need to be installed, being 4.43.0 or higher, as:

pip install "transformers[accelerate]>=4.43.0" --upgrade

To run the inference on top of Llama 3.1 405B BNB in NF4 precision, the model can be instantiated as any other causal language modeling model via AutoModelForCausalLM and run the inference normally.https://huggingface.co/hugging-quants/Meta-Llama-3.1-405B-BNB-NF4-BF16

huggingface.co

hugging-quants/Meta-Llama-3.1-405B-BNB-NF4-BF16 · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

13:22 · Aug 8, 2024 · Thu

2024 The Hugging-quants/Meta-Llama-3.1-405B-BNB-NF4
Model Information
The Meta Llama 3.1 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction tuned generative models in 8B, 70B and 405B sizes (text in/text out). The Llama 3.1 instruction tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source and closed chat models on common industry benchmarks.

This repository contains meta-llama/Meta-Llama-3.1-405B quantized using bitsandbytes from BF16 down to NF4 with a block size of 64.

🤗 transformers
In order to run the inference with Llama 3.1 405B BNB in NF4, both torch and bitsandbytes need to be installed as:

pip install "torch>=2.0.0" bitsandbytes --upgrade

Then, the latest version of transformers need to be installed, being 4.43.0 or higher, as:

pip install "transformers[accelerate]>=4.43.0" --upgrade

To run the inference on top of Llama 3.1 405B BNB in NF4 precision, the model can be instantiated as any other causal language modeling model via AutoModelForCausalLM and run the inference normally.https://huggingface.co/hugging-quants/Meta-Llama-3.1-405B-BNB-NF4

huggingface.co

hugging-quants/Meta-Llama-3.1-405B-BNB-NF4 · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

13:20 · Aug 8, 2024 · Thu

Vintern-1B ❄️ (Viet-InternVL2-1B) [🤗 HF Demo] - The LLaVA 🌋 Challenger
We are excited to introduce Vintern-1B the Vietnamese 🇻🇳 multimodal model that combines the advanced Vietnamese language model Qwen2-0.5B-Instruct with the latest visual model, InternViT-300M-448px, CVPR 2024. This model excels in tasks such as OCR-VQA, Doc-VQA, and Chart-VQA,... With only 1 billion parameters, it is 4096 context length finetuned from the InternVL2-1B model on over 1.5 million specialized image-question-answer pairs for optical character recognition 🔍, text recognition 🔤, document extraction 📑, and more. The model can be integrated into various on-device applications 📱, demonstrating its versatility and robust capabilities.

Vintern-1B is a multimodal large language model series, featuring models of various sizes. For each size, we release instruction-tuned models optimized for multimodal tasks. Vintern-1B consists of InternViT-300M-448px, an MLP projector, and Qwen2-0.5B-Instruct.

Training details 📚
The fine-tuning dataset was meticulously sampled in part from the following datasets:

Viet-OCR-VQA
Viet-Doc-VQA
Viet-Doc-VQA-II
Benchmarks 📈
Since there are still many different metrics that need to be tested, we chose a quick and simple metric first to guide the development of our model. Our metric is inspired by Lavy's paper. For the time being, we are using GPT-4 to evaluate the quality of answers on two datasets: OpenViVQA and ViTextVQA. Detailed results can be found at the provided Here. The inputs are images, questions, labels, and predicted answers. The model will return a score from 0 to 10 for the corresponding answer quality. The results table is shown below.https://huggingface.co/5CD-AI/Viet-InternVL2-1B

huggingface.co

5CD-AI/Viet-InternVL2-1B · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

11:47 · Aug 8, 2024 · Thu

merve/sam2-hiera-small
SAM2-Hiera-small
This repository contains small variant of SAM2 model. SAM2 is the state-of-the-art mask generation model released by Meta.

Usage
You can use it like below. First install packaged version of SAM2.
pip install samv2 huggingface_hub

Each model requires different classes to infer.
For prompting:
Rred by: https://huggingface.co/merve/sam2-hiera-small

huggingface.co

merve/sam2-hiera-small · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

11:45 · Aug 8, 2024 · Thu

75.) Massed Compute (Cloud)
Best Deepfake Open Source App ROPE - So Easy To Use Full HD Feceswap DeepFace, No GPU Required Cloud