nvidia/CUDA-Autocomplete

1---2license: other3license_name: nvidia-open-model-license4license_link: >-5  https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/6library_name: transformers7pipeline_tag: text-generation8tags:9  - code10  - cuda11  - fill-in-the-middle12  - nvidia13  - pytorch14datasets:15  - bigcode/the-stack-v216base_model: Qwen/Qwen2.5-Coder-7B17---18## Model Overview19NVIDIA CUDA Autocomplete is a fine-tuned version of Qwen/Qwen2.5-Coder-7B enhanced for CUDA code completion. The model takes as input two strings of code context: the prefix (code before the cursor) and the suffix (code after the cursor), and outputs several lines of code that logically continues the prefix. By analyzing the surrounding code structure, variable names, and CUDA-specific patterns, the model predicts the most likely next line of code, enabling intelligent autocomplete functionality for general programming and CUDA development in the Nsight Copilot extension for VSCode and Cursor.20 21_This model is ready for commercial/non-commercial use._22 23 24### License/Terms of Use25Use of this model is governed by the [NVIDIA Open Model License Agreement](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/).26 27Additional Information.For Qwen2.5-Coder-7B, [Apache License, Version 2.0](https://huggingface.co/Qwen/Qwen2.5-Coder-7B/blob/main/LICENSE).28 29### Deployment Geography30Global31 32### Use Case33This model is intended to be used for code completion in the Nsight Copilot extension for VSCode / Cursor.34 35 36### Release Date37Huggingface : 06/09/2026 via [https://huggingface.co/nvidia/CUDA-Autocomplete](https://huggingface.co/nvidia/CUDA-Autocomplete)38 39## Reference(s)40[Qwen2.5-Coder paper](https://arxiv.org/abs/2409.12186)41[Qwen2.5-Coder blog](https://qwenlm.github.io/blog/qwen2.5-coder-family/)42[Qwen2.5-Coder GitHub repository](https://github.com/QwenLM/Qwen2.5-Coder)43 44## Model Architecture45**Architecture Type:** Transformer46**Network Architecture:** Qwen2ForCausalLM47**This model was developed based on Qwen/Qwen2.5-Coder-7B.**48**Number of model parameters:** 7B (7*10^9)49 50 51## Input52**Input Type(s):** Code53**Input Format(s):** String of code (meant for prefix code and suffix code)54**Input Parameters:** One-Dimensional (1D)55**Other Properties Related to Input:**56- **Context Window:** The model processes sequential code text with prefix and suffix context57- **Encoding:** UTF-8 text encoding58- **Input Structure:** Fill-in-the-middle (FIM) format with prefix and suffix tokens59 60 61## Output62**Output Type(s):** Code63**Output Format:** String64**Output Parameters:** One-Dimensional (1D)65**Other Properties Related to Output:**66- **Output Length:** Single line of code completion67- **Generation Method:** Autoregressive token-by-token generation68- **Encoding:** UTF-8 text encoding69- **Output Structure:** Sequential code text that continues from the input prefix70 71 72Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA's hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions.73 74## Software Integration75**Runtime Engine(s):** vLLM76**Supported Hardware Microarchitecture Compatibility:**77* H10078* DGX Spark79**[Supported] Operating System(s):** Linux80 81The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.82 83 84## Model Version(s)85v0.3.086 87## Training, Testing, and Evaluation Datasets88 89 90### Training Dataset91 92* **Source:** Subset of [bigcode/the-stack-v2](https://huggingface.co/datasets/bigcode/the-stack-v2) & synthetically generated CUDA data using OSS models like GPT-OSS 120B93* **Data Modality:** Text94* **Text Training Data Size:** ~700000 samples95* **Data Collection Method by dataset:** Hybrid: Automated, Synthetic96* **Labeling Method by dataset:** Not Applicable97* **Properties (Quantity, Dataset Descriptions, Sensor(s)):** ~700,000 samples. Text modality (source code). Content includes open-source CUDA and general programming code collected from permissive-licensed repositories, as well as machine-generated synthetic CUDA code produced by OSS models. Primarily English-language code with CUDA-specific constructs and APIs. No sensor data involved.98 99### Testing Dataset100* **Source:** NVIDIA Internal Data101* **Data Collection Method by dataset:** Automated102* **Labeling Method by dataset:** Not Applicable103* **Properties (Quantity, Dataset Descriptions, Sensor(s)):** 2,156 samples. Text modality (source code). Content consists of internal proprietary CUDA and HPC library code (e.g., cuDNN, cuda-hpc) parsed from internal GitLab repositories. Code is CUDA-specific with domain-specific APIs and patterns. No sensor data involved.104 105 106### Evaluation Dataset107* **Source:** Subset of [bigcode/the-stack-v2](https://huggingface.co/datasets/bigcode/the-stack-v2)108* **Data Collection Method by dataset:** Automated109* **Labeling Method by dataset:** Not Applicable110* **Properties (Quantity, Dataset Descriptions, Sensor(s)):** ~33,000 samples. Each sample corresponds to a single source code file. Text modality (source code). Content includes open-source code collected from permissive-licensed repositories. CUDA and general programming code in English. No sensor data involved.111 112 113## Inference114**Acceleration Engine:** vLLM115 116**Test Hardware:**117* H100118* DGX Spark119 120## Ethical Considerations121NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications.  When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.122For more detailed information on ethical considerations for this model, please see the Model Card++ Explainability, Bias, Safety & Security, and Privacy Subcards.123Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns [here](https://app.intigriti.com/programs/nvidia/nvidiavdp/detail).124