LLaMA Cog template 🦙

This is a monorepo for building multiple Llama models using Cog:

• llama-2-13b
• llama-2-13b-chat
• llama-2-13b-transformers
• llama-2-70b
• llama-2-70b-chat
• llama-2-7b
• llama-2-7b-chat
• llama-2-7b-transformers
• llama-2-7b-vllm

See replicate.com/meta.

---

NOTE: This is an experimental branch that depends on exllama

For now, you should:

git clone https://github.com/turboderp/exllama
cd exllama
git checkout e8a544f95b3fd64dfa5549eeeafb85b1ac71a793

We're working on a proper integration.

This Cog template works with LLaMA 1 & 2 versions.

LLaMA is a new open-source language model from Meta Research that performs as well as closed-source models.

This is a guide to running LLaMA using in the cloud using Replicate. You'll use the Cog command-line tool to package the model and push it to Replicate as a web interface and API.

This template can be used to run the 7B, 13B, and 70B versions of LLaMA and LLaMA2 and it also works with fine-tuned models.

Note: Please verify the system prompt for LLaMA or LLAMA2 and update it accordingly.

Note: LLaMA is for research purposes only. It is not intended for commercial use. Check the license of LLaMA & LLaMA2 on the official LLaMA website of Meta Platforms, Inc.

Prerequisites

• LLaMA weights. The weights for LLaMA have not yet been released publicly. To apply for access, fill out the Meta Research form to be able to download the weights.
• GPU machine. You'll need a Linux machine with an NVIDIA GPU attached and the NVIDIA Container Toolkit installed. If you don't already have access to a machine with a GPU, check out our guide to getting a GPU machine.
• Docker. You'll be using the Cog command-line tool to build and push a model. Cog uses Docker to create containers for models.

Step 0: Install Cog

First, install Cog:

sudo curl -o /usr/local/bin/cog -L "https://github.com/replicate/cog/releases/latest/download/cog_$(uname -s)_$(uname -m)"
sudo chmod +x /usr/local/bin/cog

Step 1: Set up weights

Replicate currently supports the 7B model size.

Put your downloaded weights in a folder called unconverted-weights. The folder hierarchy should look something like this:

unconverted-weights
├── 7B
│   ├── checklist.chk
│   ├── consolidated.00.pth
│   └── params.json
├── tokenizer.model
└── tokenizer_checklist.chk

Convert the weights from a PyTorch checkpoint to a transformers-compatible format using the this command:

cog run python -m transformers.models.llama.convert_llama_weights_to_hf --input_dir unconverted-weights --model_size 7B --output_dir weights

You final directory structure should look like this:

weights
├── config.json
├── generation_config.json
├── pytorch_model-00001-of-00002.bin
├── pytorch_model-00002-of-00002.bin
├── pytorch_model.bin.index.json
├── special_tokens_map.json
├── tokenizer.model
└── tokenizer_config.json

Once you've done this, you should uncomment unconverted-weights in your .dockerignore file. This ensures that unconverted-weights aren't built into the resulting cog image.

Step 2: Tenzorize the weights from the transformers-compatible/huggingface format (this will allow cold-starts to happen much faster):

Run convert_to_tensors.py to tenzorize the weights from the previous transformers-compatible/huggingface format:

cog run python convert_to_tensors.py

This will tensorize your weights and write the tensorized weights to ./llama_weights/llama-7b/llama_7b_fp16.tensors if you have a GPU available and .../llama_7b_fp32.tensors if you don't. (To load the tensorized model instead of the transformers-compatible/huggingface weights, verify that DEFAULT_MODEL_NAME in config.py is set to the path of your tensorized weights.)

• Make sure **.tensors is not in your .dockerignore: In your .dockerignore file, remove **.tensors. This line will ignore all files that end with .tensors, no matter where they are in the directory structure.

Step 3: Run the model

You can run the model locally to test it:

cog predict -i prompt="Simply put, the theory of relativity states that"

LLaMA is not fine-tuned to answer questions. You should construct your prompt so that the expected answer is the natural continuation of your prompt.

Here are a few examples from the LLaMA FAQ:

• Do not prompt with "What is the meaning of life? Be concise and do not repeat yourself." but with "I believe the meaning of life is"
• Do not prompt with "Explain the theory of relativity." but with "Simply put, the theory of relativity states that"
• Do not prompt with "Ten easy steps to build a website..." but with "Building a website can be done in 10 simple steps:\n"

Step 4: Create a model on Replicate

Go to replicate.com/create to create a Replicate model.

Make sure to specify "private" to keep the model private.

Step 5: Configure the model to run on A100 GPUs

Replicate supports running models on a variety of GPUs. The default GPU type is a T4, but for best performance you'll want to configure your model to run on an A100.

Click on the "Settings" tab on your model page, scroll down to "GPU hardware", and select "A100". Then click "Save".

Step 6: Push the model to Replicate

Log in to Replicate:

sudo cog login

Push the contents of your current directory to Replicate, using the model name you specified in step 3:

sudo cog push r8.im/username/modelname

Note: if you get an error while pushing your model indicating that your model does not exist on Replicate (even if it was successfully created on the Replicate dashboard), make sure to use the "sudo" command in the "cog login" in terminal.

Learn more about pushing models to Replicate.

Step 7: Run the model on Replicate

Now that you've pushed the model to Replicate, you can run it from the website or with an API.

To use your model in the browser, go to your model page.

To use your model with an API, click on the "API" tab on your model page. You'll see commands to run the model with cURL, Python, etc.

To learn more about how to use Replicate, check out our documentation.

Contributors ✨

This template was generated by Marco Mascorro (@mascobot), with some modifications to the original cog LLaMA template and with the help of the cog and Replicate documentation that wonderful people put together. See all contributors below.

This project follows the all-contributors specification. Contributions of any kind welcome!