{"id":30060990,"url":"https://github.com/bhimrazy/deepnvme-experiments","last_synced_at":"2026-02-11T04:01:16.204Z","repository":{"id":300012303,"uuid":"1004908200","full_name":"bhimrazy/deepnvme-experiments","owner":"bhimrazy","description":"A collection of experiments demonstrating simple file reads and writes involving CPU/GPU tensors using DeepNVMe.","archived":false,"fork":false,"pushed_at":"2025-07-05T18:11:29.000Z","size":16,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2025-08-08T01:54:07.952Z","etag":null,"topics":["deepnvme","deepspeed"],"latest_commit_sha":null,"homepage":"https://github.com/deepspeedai/DeepSpeedExamples/tree/master/deepnvme/file_access","language":null,"has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/bhimrazy.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2025-06-19T11:14:46.000Z","updated_at":"2025-07-05T18:11:32.000Z","dependencies_parsed_at":"2025-06-19T12:48:31.547Z","dependency_job_id":null,"html_url":"https://github.com/bhimrazy/deepnvme-experiments","commit_stats":null,"previous_names":["bhimrazy/deepnvme-experiments"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/bhimrazy/deepnvme-experiments","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/bhimrazy%2Fdeepnvme-experiments","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/bhimrazy%2Fdeepnvme-experiments/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/bhimrazy%2Fdeepnvme-experiments/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/bhimrazy%2Fdeepnvme-experiments/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/bhimrazy","download_url":"https://codeload.github.com/bhimrazy/deepnvme-experiments/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/bhimrazy%2Fdeepnvme-experiments/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":29326801,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-02-11T03:52:29.695Z","status":"ssl_error","status_checked_at":"2026-02-11T03:52:23.094Z","response_time":97,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["deepnvme","deepspeed"],"created_at":"2025-08-08T01:54:01.108Z","updated_at":"2026-02-11T04:01:16.182Z","avatar_url":"https://github.com/bhimrazy.png","language":null,"funding_links":[],"categories":[],"sub_categories":[],"readme":"# DeepNVMe Experiments\n\nA collection of experiments demonstrating simple file reads and writes involving CPU/GPU tensors using DeepNVMe.\n\n## Overview\n\nThis project guides you through running and experimenting with DeepNVMe for high-performance storage access using both CPU and GPU tensors. The scripts referenced here are cloned from the official DeepSpeedExamples repository.\n\n\u003cimg width=\"1330\" alt=\"image\" src=\"https://github.com/user-attachments/assets/91f35288-cce3-487c-9a7b-aa37b062c4ad\" /\u003e\n\n\u003e **Note:** The above [results](https://docs.google.com/spreadsheets/d/1Ffdm-IOCCDXg_fLR1EK9hhPAUeI3LFo1pZ87YuDK_K8/edit?usp=sharing) were obtained from experiments conducted on [Lightning AI Studios](https://lightning.ai/).\n## Prerequisites\n\n- Python 3.10+\n- Linux (tested on Ubuntu)\n- [DeepSpeed](https://www.deepspeed.ai/)\n- (Optional, for GPU tests) NVIDIA GPU with GPUDirect Storage support\n- `libaio` development libraries\n\n## Setup\n\n1. **Clone the Repository**\n    ```sh\n    git clone https://github.com/deepspeedai/DeepSpeedExamples.git\n    cd DeepSpeedExamples/deepnvme/file_access\n    ```\n\n2. **Install Dependencies**\n    ```sh\n    pip install deepspeed\n    sudo apt update\n    sudo apt install libaio-dev\n    ```\n    \u003e **Note:** If the `async_io` operator is unavailable, ensure `libaio-dev` is installed.\n\n3. **(Optional) GPUDirect Storage**\n    - For GPU-related tests, follow the [NVIDIA GDS troubleshooting guide](https://docs.nvidia.com/gpudirect-storage/troubleshooting-guide/index.html).\n\n4. **Check DeepSpeed Installation**\n    ```sh\n    ds_report\n    ```\n    ![image](https://github.com/user-attachments/assets/38a25e27-79ea-433b-8fff-edda261a8d4f)\n\n\n## Running Experiments\n\n1. **Prepare Output Directories**\n    ```sh\n    mkdir -p py_out aio_out\n    ```\n\n2. **CPU Tensor File Operations**\n    - **Store Tensor:**\n        ```sh\n        python py_store_cpu_tensor.py --nvme_folder py_out\n        ```\n    - **Load Tensor:**\n        ```sh\n        python py_load_cpu_tensor.py --input_file py_out/test_ouput_1024MB.pt\n        ```\n    \u003e **Tip:** Comment out the last line in the script to prevent the file from being unlinked (see script, e.g., line 23).\n\n3. **AIO CPU Tensor File Operations**\n    - **Store Tensor:**\n        ```sh\n        python aio_store_cpu_tensor.py --nvme_folder aio_out\n        ```\n    - **Load Tensor:**\n        ```sh\n        python aio_load_cpu_tensor.py --input_file aio_out/test_ouput_1024MB.pt\n        ```\n    \u003e **Note:** The first run may trigger compilation.  \n    \u003e **Tip:** Comment out the last line in the script to prevent the file from being unlinked (see script, e.g., line 44).\n\n## Performance Tuning\n\nUse `ds_nvme_tune` to automatically find optimal NVMe settings:\n\n```sh\nmkdir -p local_nvme\nds_nvme_tune --nvme_dir local_nvme         # For CPU\nds_nvme_tune --nvme_dir local_nvme --gpu   # For GPU\n```\n\nSee the [DeepSpeed DeepNVMe tutorial](https://www.deepspeed.ai/tutorials/deepnvme/?utm_source=chatgpt.com#performance-tuning) for more details.\n\n## References\n\n- [DeepSpeedExamples: deepnvme/file_access](https://github.com/deepspeedai/DeepSpeedExamples/tree/master/deepnvme/file_access)\n- [DeepSpeed DeepNVMe Tutorial](https://www.deepspeed.ai/tutorials/deepnvme)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fbhimrazy%2Fdeepnvme-experiments","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fbhimrazy%2Fdeepnvme-experiments","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fbhimrazy%2Fdeepnvme-experiments/lists"}