initial commit

LICENSE_ai8x.md

@@ -0,0 +1,33 @@
+# LICENSE
+
+Copyright (C) 2019-2023 Maxim Integrated Products, Inc., All rights Reserved.
+
+This software is protected by copyright laws of the United States and
+of foreign countries. This material may also be protected by patent laws
+and technology transfer regulations of the United States and of foreign
+countries. This software is furnished under a license agreement and/or a
+nondisclosure agreement and may only be used or reproduced in accordance
+with the terms of those agreements. Dissemination of this information to
+any party or parties not specified in the license agreement and/or
+nondisclosure agreement is expressly prohibited.
+
+The above copyright notice and this permission notice shall be included
+in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL MAXIM INTEGRATED BE LIABLE FOR ANY CLAIM, DAMAGES
+OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+OTHER DEALINGS IN THE SOFTWARE.
+
+Except as contained in this notice, the name of Maxim Integrated
+Products, Inc. shall not be used except as stated in the Maxim Integrated
+Products, Inc. Branding Policy.
+
+The mere transfer of this software does not imply any licenses
+of trade secrets, proprietary technology, copyrights, patents,
+trademarks, maskwork rights, or any other form of intellectual
+property whatsoever. Maxim Integrated Products, Inc. retains all
+ownership rights.

README.md

@@ -0,0 +1,77 @@
+## DEX: Data Channel Extension for Efficient CNN Inference on Tiny AI Accelerators (NeurIPS '24)
+
+This is the official PyTorch Implementation of "DEX: Data Channel Extension for Efficient CNN Inference on Tiny AI Accelerators" (NeurIPS '24) 
+by 
+[Taesik Gong](https://taesikgong.com/), 
+[Fahim Kawsar](https://www.fahim-kawsar.net/), and
+[Chulhong Min](https://chulhongmin.com/).
+
+The codebase is forked from the [AI8X Training repository](https://github.com/analogdevicesinc/ai8x-training), the official repository for MAX78000 and MAX78002 AI accelerators, and additional implementations are added for the DEX project.
+
+### Tested Environment
+
+- Ubuntu 22.04
+- NVIDIA A40 GPUs
+
+### Step-by-Step Guide
+
+
+1. Move to the root directory:
+   ```bash
+   cd data-channel-extension
+   ```
+
+2. Build the Docker image:
+   ```bash
+   ./docker/build_image.sh
+   ```
+
+3. Run the Docker container:
+   ```bash
+   ./docker/run_container.sh
+   ```
+
+4. Execute the container:
+   ```bash
+   ./docker/exec_container.sh
+   ```
+
+   This starts the SSH service in the Docker container.
+
+   **Password**: `root`
+
+5. Run the training script:
+   ```bash
+   ./scripts/dex/train_all.sh
+   ```
+
+6. Download and extract datasets:
+
+   - **Caltech101**:
+     - Download: [Caltech101](https://drive.google.com/file/d/137RyRjvTBkBiIfeYBNZBtViDHQ6_Ewsp)
+     - Extract under `data/Caltech101/caltech101/101_ObjectCategories`
+
+   - **Caltech256**:
+     - Download: [Caltech256](https://drive.google.com/file/d/1r6o0pSROcV1_VwT4oSjA2FBUSCWGuxLK)
+     - Extract under `data/Caltech256/caltech256/256_ObjectCategories/`
+
+   - **ImageNette**:
+     - Download: [ImageNette](https://s3.amazonaws.com/fast-ai-imageclas/imagenette2.tgz)
+     - Extract under `data/Imagenette/`
+
+   - **Food101**:
+     - The required data will be automatically downloaded.
+
+7. Modify the `train_all.sh` file to run specific experiments as needed. By default, it launches training for ImageNette with SimpleNet.
+
+
+### Citation
+
+```
+@inproceedings{gong2024dex,
+    title={{DEX}: Data Channel Extension for Efficient CNN Inference on Tiny AI Accelerators},
+    author={Gong, Taesik and Kawsar, Fahim and Min, Chulhong},
+    booktitle={Thirty-eighth Conference on Neural Information Processing Systems},
+    year={2024}
+}
+```