
# How to Run Mixtral 8X7B

Mixtral is the latest open source model from Mistral AI. The model was release with no documentation, explanation or other comment from Mistral AI on Dec 8th via a single Tweet of the [magnet link](https://x.com/MistralAI/status/1733150512395038967?s=20). Mistral AI has several open source LLM models that are popular including Mistral 7B. 

Mixtral 8X7B is notable in that it is a mixture of experts (MoE) model with exceptional ability. This guide uses some hacky implementations to get it to run. Once the model is out for a few months, it will certainly gain more support with open source tools.

To run it on Vast.ai, use a 8X 4090/3090 or a 4X A6000/A40 instance. You will need 120GB of total GPU RAM on the instance for the model. We use the typical development Pytorch recommended template and a modified version of Illama to run inference. This is a base model with no instruction fine tuning, so proper prompt techniques are helpful. 

## Steps:

1. **Select Template**:
   - Choose the [Devel template](https://cloud.vast.ai/?ref_id=43484&template_id=e4c5e88bc289f4eecb0c955c4fe7430d) for Pytorch by clicking that link or selecting it in Templates.

2. **Rent a Server**:
   - Select a 8X 4090/3090 or 4X A40/A6000
   - Add at least 120GB of disk space
   - Click the Rent button
   - Purchase credits if needed.
   - Refer to the [quickstart guide](https://vast.ai/docs/overview/quickstart) for help.

3. **SSH into the Machine**:
   - Obtain SSH information from the instance card by clicking on the >_ button 
   - Connect to the instance via the SSH command. It will look like this but have the port and server IP info of your instance:
     ```
     ssh -p <yourport> root@<yourserverip> -L 8080:localhost:8080
     ```

4. **Download Mixtral 8X7B Model Weights**:
   - Use a torrent file. This typically takes about 15 minutes, depending on server download speed.
   - To download the torrent file, run:
     ```
     apt install transmission-cli
     transmission-cli magnet:?xt=urn:btih:5546272da9065eddeb6fcd7ffddeef5b75be79a7&dn=mixtral-8x7b-32kseqlen&tr=udp%3A%2F%http://2Fopentracker.i2p.rocks%3A6969%2Fannounce&tr=http%3A%2F%http://2Ftracker.openbittorrent.com%3A80%2Fannounce
     ```

5. **Install Illama for Inference**:
   - Clone the repository:
     ```
     git clone https://github.com/dzhulgakov/llama-mistral
     ```
   - Install dependencies:
     ```
     pip install fire sentencepiece
     ```

6. **Run Mixtral 8X7B!**:
   - Load the model weights and run the prompts in `example_text_completion.py`. If you use 8 GPUS, append "--num-gpus 8" at the end of the run command like in the example below. If you use 2 GPUS, remove that line.:
     ```
     cd llama-mistral
     python example_text_completion.py ../Downloads/mixtral-8x7b-32kseqlen/ ../Downloads/mixtral-8x7b-32kseqlen/tokenizer.model --num-gpus 8
     ```
	 And there you go! You will get the output of the prompts that are currently listed 

7. **Modify Prompts**:
   - Install Nano:
     ```
     apt install nano
     ```
   - Edit the file:
     ```
     nano example_text_completion.py
     ```
   - Find and modify the prompts at the bottom.
   - Re-run the python file to get new responses.

Thank you for using Vast.ai to run the latest open source LLM, Mixtral 8X7B. Drop us a line and let us know what you think of the model.


Running the latest open source model from Mistral AI called Mixtral 8x7b on Vast.ai using 8X 4090s or 8X 3090s

How to run Mixtral 8X7B


# Google Colab vs. Vast: Which Platform Is Best for Your Projects?

If you’ve ever dabbled in machine learning, data science, or coding in general, chances are you’ve heard of Google Colaboratory – better known as **Google Colab**. But what exactly is it, what are its limitations, and how can you make the best use of it?

Let's take a closer look at these questions and more…

## What Is Google Colab?

In a nutshell, **Google Colab** is a free, browser-based platform that lets you run and write code in a Jupyter notebook environment without having to set up anything locally on your machine. It's widely used for Python-based projects, especially in machine learning and data analysis.

Colab offers an easy starting point for experiments, since you can use the Jupyter notebook interface to write and execute code, visualize results, and keep everything neatly organized in one place. This type of interactive programming allows you to experiment with code in real time, making it ideal for data science and research projects that require immediate feedback and iteration.

### Some other benefits include:

- **Accessibility** – The browser-based platform allows you to access your projects from anywhere.
- **Integration with Google Drive** – Save your notebooks and datasets directly to your Google Drive.
- **Collaboration** – Easily share your notebooks with others for real-time collaboration across teams.
- **Wealth of Resources** – Plenty of tutorials, sample notebooks, and community support are available in the Colab ecosystem, making it easier for you to learn and troubleshoot along the way.

These features make Google Colab a convenient tool for a wide range of users. However, it does have some significant limitations that may prompt you to look for other solutions – such as running a Jupyter instance right here on **Vast** and uploading your notebook directly.

## Why Choose Vast Over Google Colab?

While Google Colab is helpful for smaller-scale projects or initial experimentation, **Vast's** cloud GPU platform offers a much more flexible, powerful, and reliable experience through our Jupyter integration. Here’s why you should consider using Vast instead.

### No Resource Limits

With Google Colab, free-tier users face dynamic usage limits with idle timeout periods and resource caps that sometimes fluctuate. The paid plans have their own set of restrictions as well. For example, notebooks can run for up to 12 hours on a free plan, and the highest-tier paid plan supports continuous code execution for at most 24 hours.

But with **Vast**, there are no imposed time limits, so your projects can run uninterrupted for as long as you need. Whether your project takes hours or days, you won't be cut off or forced to restart.

> **Note**: Google Colab does include the option to pay for a dedicated virtual machine (VM) to get around usage limits, but Vast's pricing is extremely competitive as we are the market leader in low-cost cloud GPU rental.

### Customizable GPU Power for Every Project

Google Colab offers different VMs based on the plan you choose. The free tier only allows access to VMs with a standard system memory profile, restricting the compute power you can use. Paid versions of Colab provide access to machines with a higher system memory profile, but you still don't get to select specific GPUs unless you opt to manually pay for a dedicated machine.

**Vast** gives you access right from the start to powerful and customizable hardware options that are also highly affordable. You can select the exact GPU setup you need, whether you're running small tasks or large-scale machine learning models. For high-end performance, Vast supports multi-GPU configurations, making it easy to scale up your projects without hitting resource bottlenecks.

### Simplicity and Reliability

**Vast** simplifies the entire process by running your Jupyter notebook directly in the cloud with no need for SSH or complex setups. Jupyter by itself is much more reliable than Google Colab and doesn't require setting up SSH keys – you can always open a terminal right inside the browser.

And while Google Colab does facilitate sharing and collaboration, **Vast** offers the same, with the added benefit of running on a more stable platform without timeouts or interruptions. Team members can collaborate in real time, sharing notebooks and results without worrying about session limits.

## Making the Switch to Vast

If you've been using Google Colab for simple notebooks and want to step up your workflow, switching to Vast is straightforward. Simply download your notebook from Colab as a `.ipynb` file, run a Vast Jupyter instance with the recommended [PyTorch template](https://cloud.vast.ai/?ref_id=43484&template_id=f5540ef1a1398b8499546edb53dae704), and upload your notebook into Jupyter directly.

For a step-by-step guide on running Jupyter notebooks on Vast, check out our detailed doc [here](https://vast.ai/docs/use-cases/google-colab).

While Google Colab is a useful starting point for basic tasks, it often falls short when you're dealing with more complex workflows. **Vast** provides a more robust and customizable solution through its Jupyter notebook integration, with greater control, higher performance, and fewer interruptions.

With **Vast**, you'll enjoy the freedom to push your projects even further. **Explore the possibilities today!**


Google Colab is a free, browser-based platform that lets you run and write code in a Jupyter notebook environment without having to set up anything locally on your machine.

Google Colab Explained: Simplifying Your Workflow with Cloud Tools


# Serving vLLM Embeddings on Vast.ai

## Background:

[vLLM](https://github.com/vllm-project/vllm) is a developer-friendly framework for serving large language models. It is very fast and focused on high throughput serving multiple simultaneous calls for production use cases. `vLLM` is now more flexible than ever as it also supports embedding models. This brings vLLM's dynamic batching and Paged Attention to embedding models for much faster throughput, all from the docker image that developers are used to.

This guide will show you how to setup vLLM to serve embedding models on Vast.

## Setup and Querying

First, we setup our environment and vast api key

```bash
pip install --upgrade vastai
```

Once you create your account, you can go [here](https://cloud.vast.ai/cli/) to find your API Key.

```bash
vastai set api-key <Your-API-Key-Here>
```

To serve an embedding model, we'll want a machine with a static IP address + ports to forward. We only will need a single GPU, which can have a small amount of RAM since embedding models are small. We'll prefer a modern GPU so we will add a filter for a compute cap of more than 800.

```bash
vastai search offers 'compute_cap >= 800 gpu_ram >= 24 num_gpus = 1 static_ip=true direct_port_count > 1 cuda_vers >= 12.4'
```

## Deploying the Image:

The easiest way to deploy this instance is to use the command line. Copy and Paste a specific instance id you choose from the list above into `instance-id` below.

### Hosting a Single Embedding Model:

For now, we'll host just one embedding model.
The easiest way to deploy a single model on this instance is to use the command line. Copy and Paste a specific instance id you choose from the list above into `instance-id` below.

We particularly need `v2` so that we use the correct version of the api, `--port 8000` so it serves on the correct model, and `--model-id intfloat/e5-mistral-7b-instruct` to serve the correct model.

```bash
vastai create instance <instance-id> --image vllm/vllm-openai:latest --env '-p 8000:8000' --disk 40 --args --model intfloat/e5-mistral-7b-instruct
```

## Connecting and Testing:

To connect to your instance, we'll first need to get the IP address and port number. Once your instance is done setting up, you should see something like this:
![Instance_view](/uploads/serve_vllm_embeddings_instance_view.png)

Click on the highlighted button to see the IP address and correct port for our requests.

![IP_address_view](/uploads/serve_vllm_embeddings_ip_address_view.png)

## Connecting with the python requests library

```
import requests

headers = {
    'Content-Type': 'application/json',
}

json_data = {
    'model': 'intfloat/e5-mistral-7b-instruct',
    "input": ["San francisco is a"],
}

response = requests.post('http://<Instance-IP-Address>:<Port>/v1/embeddings', headers=headers, json=json_data)
print(response.content)
```

## Connecting with the OpenAI SDK

The OpenAI SDK also allows for generating embeddings. We can use it as well with the following code.

```bash
pip install openai
```

We will copy over the IP address and the port into the cell below.

```python

from openai import OpenAI

# Modify OpenAI's API key and API base to use Infinity's API server.
openai_api_key = "EMPTY"
openai_api_base = "http://<Instance-IP-Address>:<Port>/v1"
client = OpenAI(
    api_key=openai_api_key,
    base_url=openai_api_base,
)
model = "intfloat/e5-mistral-7b-instruct"
embeddings = client.embeddings.create(model=model, input="San francisco is a").data[0].embedding
print("Embeddings:")
print(embeddings)
```

Now we can integrate our embedding server into apps that use the OpenAI SDK.

## Next Steps:

Embeddings are an important part of GenAI applications along with running inference of Generative Models. Now with vLLM, you can run them with the same docker image on Vast for ultimate flexibility.


Serving vLLM Embeddings on Vast.ai


When it comes to high-performance computing (HPC), the NVIDIA H100 is among the best of the best GPUs on the market. Its predecessor, the NVIDIA A100, is also a very impressive GPU in its own right. From enterprise to exascale, these GPUs are enabling groundbreaking advancements in everything from AI research to complex scientific simulations.

Today we're taking a look at how the H100 and A100 compare with each other. Depending on your use case, one or the other might be better suited for your needs and preferences. So let's dive in!

## Introduction to the NVIDIA H100 and A100

As mentioned, the H100 GPU is the successor to the A100. It was specifically optimized for machine learning and designed to [deliver](https://developer.nvidia.com/blog/nvidia-hopper-architecture-in-depth/) an "order-of-magnitude performance leap for large-scale AI and HPC" over the A100, with "substantial improvements in architectural efficiency."

Based on the innovative Hopper architecture, the H100 can speed up large language models (LLMs) by 30X over the A100, according to NVIDIA. It features fourth-gen Tensor Cores and – unlike the A100 – a dedicated Transformer Engine to solve trillion-parameter language models.

Whether it's used in smaller enterprise systems or massive, unified GPU clusters, the H100 delivers efficient scalability thanks to its combination of fourth-gen NVLink (with 900 GB/s of GPU-to-GPU interconnect), NDR Quantum-2 InfiniBand networking (speeding up communication by every GPU across nodes), and PCIe Gen5. With superior FP64 and FP8 performance, the H100 is ideal for next-gen AI workloads.

The A100 GPU is known for its versatility, however. Built on the Ampere architecture, it was groundbreaking when released in 2020, providing up to 20X higher performance over the prior NVIDIA Volta generation. It can efficiently scale up or be partitioned using Multi-Instance GPU (MIG) into seven isolated GPU instances, just like the H100, which is ideal for shifting workload demands.

In short, the A100 excels in both training and inference tasks, and while it may not offer the sheer power of the H100, its robust FP64 performance makes it a great fit for scientific simulations and data analytics.

## Key Differences: H100 vs. A100

The H100 has an updated chip design and various other factors that distinguish it from the A100. The following are a few that are important to consider:

### Memory Bandwidth

The A100 has a memory bandwidth of about 2 TB/s, which is plenty for most use cases; however, the H100 offers an impressive 3.35 TB/s, making it ideal for workloads where data transfer bottlenecks may occur.

### CUDA Cores and Tensor Cores

A higher number of CUDA and Tensor Cores means a GPU can perform more parallel computations simultaneously. With 14,592 CUDA Cores and 456 Tensor Cores, the H100 outperforms the A100’s 6,912 CUDA Cores and 432 Tensor Cores, delivering superior performance in applications that can leverage increased parallelism.

### Confidential Computing

The H100's Hopper architecture introduced NVIDIA's new confidential computing environment. It protects data-in-use by performing computation in a trusted execution environment (TEE) that is both hardware-based and attested. This means you essentially have absolute security for use cases where security, privacy, and regulatory compliance are paramount.

### Tensor Memory Accelerator (TMA)

Another feature of the H100's Hopper architecture is the TMA unit, which can transfer large blocks of data efficiently between global memory and shared memory. It supports asynchronous copies between thread blocks in a cluster as well, optimizing data movement and memory management for tensor operations. Ultimately, this enhances the efficiency and speed of large-scale AI and deep learning tasks.

### MIG Capability

The Multi-Instance GPU (MIG) feature is present in both the H100 and A100, providing flexible workload management so you can run multiple LLMs in parallel, for instance. The H100's MIG capability is more robust, however, with better resource allocation and more memory and compute power to split.

## Performance Benchmarks

According to NVIDIA's own benchmarks, the H100 offers up to 30X better inference performance than the A100. In another test where the A100 and H100 were used in clusters to train two large language models (LLMs), the H100 demonstrated remarkable speed improvements. The H100's new NVLink Switch System provided an even bigger boost, enabling the H100 cluster to train the models up to nine times faster than the A100 cluster:

![H100 vs A100 Speed Comparison](/uploads/Speed-up-over-A100.png)

## Choosing the Right GPU

So what does all this mean for you? The best GPU for your needs will depend on numerous factors. Here's a quick outline of the various features and specs of the NVIDIA A100 and H100:

| Feature                         | A100                   | H100                    |
|---------------------------------|------------------------|-------------------------|
| **GPU Architecture**            | Ampere                 | Hopper                  |
| **GPU Memory**                  | 40 or 80 GB HBM2e      | 80 MB HBM3              |
| **GPU Memory Bandwidth**        | 1.6 to 2 TB/s          | 3.35 TB/s               |
| **CUDA Cores**                  | 6912                   | 14,592                  |
| **FP64 TFLOPS**                 | 9.7                    | 33.5                    |
| **FP32 TFLOPS**                 | 19.6                   | 67                      |
| **TF32 Tensor Core Flops\***    | 156 \| 312             | 378 \| 756              |
| **FP16 Tensor Core Flops\***    | 312 \| 624             | 756 \| 1513             |
| **FP8 Tensor Core TFLOPS**      | N/A                    | 3958                    |
| **Peak INT8 TOPS\***            | 624 \| 1248            | 3958                    |
| **Media Engine**                | 0 NVENC<br>5 NVDEC<br>5 NVJPEG | 0 NVENC<br>7 NVDEC<br>7 NVJPEG |
| **L2 Cache**                    | 40 MB                  | 50 MB                   |
| **Power**                       | Up to 400 W            | Up to 700 W             |
| **Form Factor**                 | SXM4 - 8-way HGX       | SXM5 - 8-way HGX        |
| **Interconnect**                | PCIe 4.0 x16           | PCIe 5.0 x16            |

\*Without and with structured sparsity.

---

## Use Cases

When it comes to use cases, you may want to consider some general guidelines.

### The A100 is well suited for:
- Versatile AI workloads including training and inference.
- Applications needing a balance of performance and cost-efficiency.
- Scenarios where ultra-high precision isn't the priority.

### The H100 is ideal for:
- Cutting-edge AI training, large-scale simulations, and other HPC workloads.
- Environments where computational speed and memory bandwidth are critical.
- Future-proofing investments in AI infrastructure.

---

When deciding between the A100 and H100, consider your specific workload requirements. If you need top-tier double-precision performance and superior memory bandwidth, or you're dealing with next-gen HPC at datacenter scale and trillion-parameter AI, the H100 is the clear winner. For a more versatile and cost-effective solution that still delivers powerful AI performance, the A100 is a solid choice.

At Vast.ai, we understand the importance of finding the right GPU for your needs as well as your budget. Our cloud GPU rental marketplace offers a variety of machines at the lowest prices possible, saving you 5-6X on GPU compute. Whether you prefer on-demand pricing or interruptible instances through spot auction-based pricing, we've got you covered.

We invite you to discover what Vast can do for you and accelerate your projects with the right GPU today!


Explore the differences between NVIDIA's H100 and A100 GPUs, from architectural advancements to performance benchmarks.

Blog

How to run Mixtral 8X7B

How to Run Mixtral 8X7B

Steps:

Continue Reading: