codellama README

The project's README from github.com/meta-llama/codellama. As one markdown file: README.md.

Introducing Code Llama

Code Llama is a family of large language models for code based on Llama 2 providing state-of-the-art performance among open models, infilling capabilities, support for large input contexts, and zero-shot instruction following ability for programming tasks. We provide multiple flavors to cover a wide range of applications: foundation models (Code Llama), Python specializations (Code Llama - Python), and instruction-following models (Code Llama - Instruct) with 7B, 13B and 34B parameters each. All models are trained on sequences of 16k tokens and show improvements on inputs with up to 100k tokens. 7B and 13B Code Llama and Code Llama - Instruct variants support infilling based on surrounding content. Code Llama was developed by fine-tuning Llama 2 using a higher sampling of code. As with Llama 2, we applied considerable safety mitigations to the fine-tuned versions of the model. For detailed information on model training, architecture and parameters, evaluations, responsible AI and safety refer to our research paper. Output generated by code generation features of the Llama Materials, including Code Llama, may be subject to third party licenses, including, without limitation, open source licenses.

We are unlocking the power of large language models and our latest version of Code Llama is now accessible to individuals, creators, researchers and businesses of all sizes so that they can experiment, innovate and scale their ideas responsibly. This release includes model weights and starting code for pretrained and fine-tuned Llama language models — ranging from 7B to 34B parameters.

This repository is intended as a minimal example to load Code Llama models and run inference.

Download

In order to download the model weights and tokenizers, please visit the Meta website and accept our License.

Once your request is approved, you will receive a signed URL over email. Then run the download.sh script, passing the URL provided when prompted to start the download. Make sure that you copy the URL text itself, do not use the 'Copy link address' option when you right click the URL. If the copied URL text starts with: https://download.llamameta.net, you copied it correctly. If the copied URL text starts with: https://l.facebook.com, you copied it the wrong way.

Pre-requisites: make sure you have wget and md5sum installed. Then to run the script: bash download.sh.

Keep in mind that the links expire after 24 hours and a certain amount of downloads. If you start seeing errors such as 403: Forbidden, you can always re-request a link.

Model sizes

Model Size
7B ~12.55GB
13B 24GB
34B 63GB
70B 131GB

Setup

In a conda environment with PyTorch / CUDA available, clone the repo and run in the top-level directory:

pip install -e .

Inference

Different models require different model-parallel (MP) values:

Model MP
7B 1
13B 2
34B 4
70B 8

All models, except the 70B python and instruct versions, support sequence lengths up to 100,000 tokens, but we pre-allocate the cache according to max_seq_len and max_batch_size values. So set those according to your hardware and use-case.

Pretrained Code Models

The Code Llama and Code Llama - Python models are not fine-tuned to follow instructions. They should be prompted so that the expected answer is the natural continuation of the prompt.

See example_completion.py for some examples. To illustrate, see command below to run it with the CodeLlama-7b model (nproc_per_node needs to be set to the MP value):

torchrun --nproc_per_node 1 example_completion.py \
    --ckpt_dir CodeLlama-7b/ \
    --tokenizer_path CodeLlama-7b/tokenizer.model \
    --max_seq_len 128 --max_batch_size 4

Pretrained code models are: the Code Llama models CodeLlama-7b, CodeLlama-13b, CodeLlama-34b, CodeLlama-70b and the Code Llama - Python models CodeLlama-7b-Python, CodeLlama-13b-Python, CodeLlama-34b-Python, CodeLlama-70b-Python.

Code Infilling

Code Llama and Code Llama - Instruct 7B and 13B models are capable of filling in code given the surrounding context.

See example_infilling.py for some examples. The CodeLlama-7b model can be run for infilling with the command below (nproc_per_node needs to be set to the MP value):

torchrun --nproc_per_node 1 example_infilling.py \
    --ckpt_dir CodeLlama-7b/ \
    --tokenizer_path CodeLlama-7b/tokenizer.model \
    --max_seq_len 192 --max_batch_size 4

Pretrained infilling models are: the Code Llama models CodeLlama-7b and CodeLlama-13b and the Code Llama - Instruct models CodeLlama-7b-Instruct, CodeLlama-13b-Instruct.

Fine-tuned Instruction Models

Code Llama - Instruct models are fine-tuned to follow instructions. To get the expected features and performance for the 7B, 13B and 34B variants, a specific formatting defined in chat_completion() needs to be followed, including the INST and <<SYS>> tags, BOS and EOS tokens, and the whitespaces and linebreaks in between (we recommend calling strip() on inputs to avoid double-spaces). CodeLlama-70b-Instruct requires a separate turn-based prompt format defined in dialog_prompt_tokens(). You can use chat_completion() directly to generate answers with all instruct models; it will automatically perform the required formatting.

You can also deploy additional classifiers for filtering out inputs and outputs that are deemed unsafe. See the llama-recipes repo for an example of how to add a safety checker to the inputs and outputs of your inference code.

Examples using CodeLlama-7b-Instruct:

torchrun --nproc_per_node 1 example_instructions.py \
    --ckpt_dir CodeLlama-7b-Instruct/ \
    --tokenizer_path CodeLlama-7b-Instruct/tokenizer.model \
    --max_seq_len 512 --max_batch_size 4

Fine-tuned instruction-following models are: the Code Llama - Instruct models CodeLlama-7b-Instruct, CodeLlama-13b-Instruct, CodeLlama-34b-Instruct, CodeLlama-70b-Instruct.

Code Llama is a new technology that carries potential risks with use. Testing conducted to date has not — and could not — cover all scenarios. In order to help developers address these risks, we have created the Responsible Use Guide. More details can be found in our research papers as well.

Issues

Please report any software “bug”, or other problems with the models through one of the following means:

Model Card

See MODEL_CARD.md for the model card of Code Llama.

License

Our model and weights are licensed for both researchers and commercial entities, upholding the principles of openness. Our mission is to empower individuals, and industry through this opportunity, while fostering an environment of discovery and ethical AI advancements.

See the LICENSE file, as well as our accompanying Acceptable Use Policy

References

  1. Code Llama Research Paper
  2. Code Llama Blog Post

codellama (meta-llama/codellama)

meta
Searching Database…

16.2K ·GitHub
Codellama
archived
What is this?This is Meta's official inference code for Code Llama, a family of large language models for code built on Llama 2. The repo itself is a minimal example for loading the models and running generation.
Are the model weights in this repo, or is it just code to run them?Just the code — you download the weights separately. You accept the license on Meta's website, get a signed URL by email, and fetch the checkpoints with the bundled download.sh script.
How do I get it running?Request access on the Meta website, run bash download.sh with the signed URL emailed to you, set up a conda environment with PyTorch and CUDA, then run the example scripts like example_completion.py or example_infilling.py.
What do I need installed before I start?A conda environment with PyTorch, plus wget and md5sum for the download script — the repo declares no other dependencies.
Anything annoying about getting the weights?Yes: download links expire after 24 hours and after a limited number of downloads, so re-request a URL if you hit 403 errors. Checkpoints are also huge — from ~12.55GB (7B) up to 63GB (34B) and 131GB (70B).
Who is Meta making this available to?Individuals, creators, researchers and businesses of all sizes — Meta positions Code Llama for anyone who wants to experiment with code LLMs, from hobby projects to commercial use.
Can I use Code Llama in commercial projects?Yes — the model and weights are licensed to both researchers and commercial entities. It is Meta's custom license (GitHub lists it as non-standard) with an Acceptable Use Policy, not a classic open-source license.
What kind of hardware does it actually run on?An NVIDIA GPU setup: inference needs PyTorch with CUDA. Larger models want several GPUs — the code uses model parallelism with MP 1 for 7B, 2 for 13B, 4 for 34B and 8 for 70B.
Which model variants are there to pick from?Three flavors: foundation Code Llama, Python-specialized Code Llama - Python, and instruction-following Code Llama - Instruct, each in 7B, 13B, 34B and 70B sizes. The 7B and 13B base and Instruct models can additionally fill in code mid-file.
Can it complete code in the middle of a file, not just continue from the end?Yes — infilling is supported by the 7B and 13B Code Llama and Code Llama - Instruct variants: they complete code based on the surrounding context, with examples in example_infilling.py.
How long can the input context be?Models are trained on 16k-token sequences and improve on inputs up to 100k tokens, except the 70B Python and Instruct versions. The KV cache is pre-allocated, so set max_seq_len and max_batch_size to match your hardware.
How do I prompt the plain pretrained models?They are not tuned to follow instructions — phrase prompts so the expected answer is the natural continuation of your text, as shown in example_completion.py. For instruction following use the Instruct variants via chat_completion(), which handles the required formatting.
Is this repo still maintained?No — meta-llama/codellama is archived on GitHub and frozen as-is. The code still serves as the reference for downloading and running the original Code Llama weights, but do not expect updates here.
What are similar projects?llama3, petals, airllm, ai-torture-chamber, StartLux-Decision, Qwen, Model-Optimizer, EvoVLM
Which topics is it in?meta, open-weights, llama, code-generation, llm-inference, large-language-models, inference, cuda, pytorch, deep-learning, gpu, machine-learning
Creator
meta-llama
Size
<0.1 MB
Stars
16.2K
Updated
CommitMay 21
Age
3y
Commits · 90 Days