Project proposal | |
Dear Students, I hope you are doing well. As we approach the next semester, I wanted to bring your attention to the upcoming full-semester project. The project can be completed individually or in groups of two. Successful completion will grant you 3 points (Project course). In some special cases, we may consider granting an additional 3 points. If you have any questions, please feel free to approach Moshe (umoshekimhiatcampus.technion.ac.il). Please add me to the CC . ************************************************************************************************************************************************************************* Post-Training Quantization for Lidar-Camera Fusion Framework Real-time sensor fusion is critical for autonomous vehicles, which rely on a complex interplay of LiDAR, camera images, and radar data. To make critical navigation decisions in real-time, onboard deep learning models must efficiently process and integrate this sensor data directly within the vehicle, all while operating under limited hardware resources. Quantization techniques offer a promising approach. By reducing the precision of the model’s calculations, quantization can significantly decrease inference time without sacrificing accuracy. This enables faster decision-making for autonomous vehicles, a crucial factor for safe and reliable navigation. In this project, you will use Post-Training Quantization techniques for both Image Encoder [1] and PointCloud [2] to improve the inference time of multi-modality models for autonomous driving [3]. [1] https://arxiv.org/pdf/2102.05426.pdf [2] https://arxiv.org/pdf/2401.15865.pdf [3] https://arxiv.org/pdf/2205.13790.pdf Supervisor(s): Moshe Kimhi and Dr Chaim Baskin Requirements: Must have knowledge of Python and the Deep Learning framework (either from a course or previous projects). Advantage: Knowledge of LIDAR, PointCloud, and other 3D data formats. ********************************************************************************************************************************************************************************* Best, Chaim |
פורסם ב-1/5/2024, 11:15:30 Created on 1/5/2024, 11:15:30 Создано1/5/2024, 11:15:30 تم النشر ب-1/5/2024, 11:15:30 |
HW3 part3 fix and hint | |
Dear students, A couple of important notes regarding HW3 part3: 1) In the cell with the test on the Encoder implementation, please comment out/ erase the row that creates the random x This was accidentally left from an older version and causes the test to fail. Regarding the EncoderLayer, there is no bug in the test. 2) HINT: If you are failing the test, you might want to check what happens to the rows in the attention matrix corresponding to padding after performing the softmax. Make sure to fix the output of the softmax if needed. 3) Regarding the padding: note that the indexes with the value 0 are the ones that should be masked out, and not the opposite. 4) Encoder output: The output should be of dimensions [BatchX1], and not [Batch] as written in the documentation. |
פורסם ב-4/4/2024, 08:37:39 Created on 4/4/2024, 08:37:39 Создано4/4/2024, 08:37:39 تم النشر ب-4/4/2024, 08:37:39 |
HW3 is out | |
Dear students please read in the course website https://vistalab-technion.github.io/cs236781/assignments/hw3 |
פורסם ב-21/3/2024, 08:58:13 Created on 21/3/2024, 08:58:13 Создано21/3/2024, 08:58:13 تم النشر ب-21/3/2024, 08:58:13 |
HW2 is out | |
Please read here https://vistalab-technion.github.io/cs236781/assignments/hw2 good luck |
פורסם ב-18/2/2024, 09:39:52 Created on 18/2/2024, 09:39:52 Создано18/2/2024, 09:39:52 تم النشر ب-18/2/2024, 09:39:52 |
Welcome to CS236781: Deep Learning on Computational Accelerators! | |
Dear Students, We welcome you to the course and hope you're doing fine in this difficult time. The official course website is at https://vistalab-technion.github.io/cs236781/ Please read the updates we posted there, starting from the welcome messege: https://vistalab-technion.github.io/cs236781/2023/12/27/welcome/ And register to the course piazza https://piazza.com/technion.ac.il/spring2024/236781 We wish you a productive and enjoyable semester, The course staff. |
פורסם ב-16/1/2024, 16:04:49 Created on 16/1/2024, 16:04:49 Создано16/1/2024, 16:04:49 تم النشر ب-16/1/2024, 16:04:49 |
First lecture- change date | |
Dear All, The first lecture will be at 17.01 (Wed) at 12:30-14:30 at Tuab 2 instead of 18.01. See you all, Chaim |
פורסם ב-8/1/2024, 11:27:20 Created on 8/1/2024, 11:27:20 Создано8/1/2024, 11:27:20 تم النشر ب-8/1/2024, 11:27:20 |