Business Context
A semiconductor manufacturer processes high-resolution wafer images to detect defects. The volume of data requires a Convolutional Neural Network (CNN). The environment is a distributed SMP server, and the initialization must explicitly reserve computing threads to ensure resource allocation aligns with subsequent heavy training tasks.
About the Set : deepLearn
Creation and training of deep neural networks.
Discover all actions of deepLearn Data Preparation
Creation of an image metadata table (placeholder for actual image loading).
Copied!
| 1 | DATA casuser.wafer_images_meta; |
| 2 | LENGTH path $100 label $20; |
| 3 | INPUT path $ label $; |
| 4 | DATALINES; |
| 5 | '/mnt/imgs/wafer1.jpg' 'defect' |
| 6 | '/mnt/imgs/wafer2.jpg' 'clean' |
| 7 | '/mnt/imgs/wafer3.jpg' 'clean' |
| 8 | '/mnt/imgs/wafer4.jpg' 'defect' |
| 9 | ; |
| 10 | RUN; |
Étapes de réalisation
1
Initialization of the CNN model with specific thread allocation (16 threads).
Copied!
| 1 | PROC CAS; |
| 2 | DEEPLEARN.buildModel / |
| 3 | modelTable={name='wafer_defect_cnn', replace=true} |
| 4 | type='CNN' |
| 5 | nThreads=16; |
| 6 | RUN; |
Expected Result
The 'wafer_defect_cnn' table is created. The log should reflect the initialization of a CNN type model and usage of the specified thread count (or system cap), preparing the environment for parallelized layer addition.