| | --- |
| | license: llama3 |
| | library_name: transformers |
| | tags: |
| | - code |
| | model-index: |
| | - name: Code Millenials |
| | results: |
| | - task: |
| | type: text-generation |
| | dataset: |
| | name: HumanEval |
| | type: openai_humaneval |
| | metrics: |
| | - type: pass@1 |
| | value: 0.671 |
| | name: pass@1 |
| | verified: false |
| | --- |
| | |
| |
|
| | # Bud Code Millenials 8B |
| |
|
| | Welcome to our Code Model repository! Our model is specifically fine-tuned for code generation tasks. Bud Millenial Code Gen open-source models are currently the State of the Art (SOTA) for code generation, beating all the existing models of all sizes. We have achieved a HumanEval value of 80.48 @ Pass 1, beating proprietary models like Gemini Ultra, Claude, GPT-3.5 etc. by a large margin, and on par with GPT-4 (HumanEval ~ 82. Ref. WizardCoder). Our proprietary model (Bud Code Jr) beats GPT-4 as well with a HumanEval value of 88.2 & a context size of 168K, we will be releasing an API for Researchers, Enterprises, and potential Partners by January 2024 end. If interested, please reach out to jithinvg@bud.studio |
| | ### News 🔥🔥🔥 |
| |
|
| | - [2024/04/21] We released **Code Millenials 8B** , which achieves the **67.1 pass@1** on the [HumanEval Benchmarks](https://github.com/openai/human-eval). |
| | - [2024/01/09] We released **Code Millenials 3B** , which achieves the **56.09 pass@1** on the [HumanEval Benchmarks](https://github.com/openai/human-eval). |
| | - [2024/01/09] We released **Code Millenials 1B** , which achieves the **51.82 pass@1** on the [HumanEval Benchmarks](https://github.com/openai/human-eval). |
| | - [2024/01/03] We released **Code Millenials 34B** , which achieves the **80.48 pass@1** on the [HumanEval Benchmarks](https://github.com/openai/human-eval). |
| | - [2024/01/02] We released **Code Millenials 13B** , which achieves the **76.21 pass@1** on the [HumanEval Benchmarks](https://github.com/openai/human-eval). |
| |
|
| |
|
| | ### HumanEval |
| |
|
| | <p align="center" width="100%"> |
| | <a ><img src="https://raw.githubusercontent.com/BudEcosystem/code-millenials/main/assets/result.png" alt="CodeMillenials" style="width: 100%; min-width: 300px; display: block; margin: auto;"></a> |
| | </p> |
| |
|
| | For the millenial models, the eval script in the github repo is used for the above result. |
| |
|
| | Note: The humaneval values of other models are taken from the official repos of [WizardCoder](https://github.com/nlpxucan/WizardLM), [DeepseekCoder](https://github.com/deepseek-ai/deepseek-coder), [Gemini](https://deepmind.google/technologies/gemini/#capabilities) etc. |
| |
|
| |
|
| | ### Models |
| |
|
| | | Model | Checkpoint | HumanEval (+) | MBPP (+) | |
| | |---------|-------------|---------------|----------| |
| | |Code Millenials 34B | <a href="https://huggingface.co/budecosystem/code-millenials-34b" target="_blank">HF Link</a> | 80.48 (75) | 74.68 (62.9) | |
| | |Code Millenials 13B | <a href="https://huggingface.co/budecosystem/code-millenials-13b" target="_blank">HF Link</a> | 76.21 (69.5) | 70.17 (57.6) | |
| | |Code Millenials 8B | <a href="https://huggingface.co/budecosystem/code-millenials-8b" target="_blank">HF Link</a> | 67.1 (61.6) | - | |
| | |Code Millenials 3B | <a href="https://huggingface.co/budecosystem/code-millenials-3b" target="_blank">HF Link</a> | 56.09 (52.43) | 55.13 (47.11) | |
| | |Code Millenials 1B | <a href="https://huggingface.co/budecosystem/code-millenials-1b" target="_blank">HF Link</a> | 51.82 (48.17) | 53.13 (44.61) | |
| |
|
| |
|
| |
|
| |
|
| | ### 🚀 Quick Start |
| |
|
| | Inference code using the pre-trained model from the Hugging Face model hub |
| |
|
| | ```python |
| | import torch |
| | from transformers import AutoTokenizer, AutoModelForCausalLM |
| | |
| | tokenizer = AutoTokenizer.from_pretrained("budecosystem/code-millenials-8b") |
| | model = AutoModelForCausalLM.from_pretrained("budecosystem/code-millenials-8b") |
| | |
| | template = """You are an exceptionally intelligent coding assistant that consistently delivers accurate and reliable responses to user instructions. |
| | |
| | ### Instruction: {instruction} |
| | |
| | ### Response:""" |
| | |
| | instruction = <Your code instruction here> |
| | |
| | prompt = template.format(instruction=instruction) |
| | |
| | inputs = tokenizer(prompt, return_tensors="pt") |
| | sample = model.generate(**inputs, max_length=128) |
| | print(tokenizer.decode(sample[0])) |
| | |
| | ``` |
| |
|
| |
|
| | ## Training details |
| |
|
| | The model is trained of 8 A100 80GB for approximately 50hrs. |
| |
|
| | | Hyperparameters | Value | |
| | | :----------------------------| :-----: | |
| | | per_device_train_batch_size | 8 | |
| | | gradient_accumulation_steps | 1 | |
| | | epoch | 3 | |
| | | steps | 8628 | |
| | | learning_rate | 2e-5 | |
| | | lr schedular type | cosine | |
| | | warmup ratio | 0.1 | |
| | | optimizer | adamw | |
| | | fp16 | True | |
| | | GPU | 8 A100 80GB | |
| | |
| | ### Important Note |
| | |
| | - **Bias, Risks, and Limitations:** Model may sometimes make errors, produce misleading contents, or struggle to manage tasks that are not related to coding. |
