Now that we have the new OpenStack platform we have used this to generate images from the Daniella network. Results are coming in faster than ever before. The M10 cards work very well and we can naturally break them down into multiple parallel jobs since there are 4 GPU’s per card.
Daniella generated some very good results at low resolution, and significantly faster (4 times faster). This means we can train more models, faster. You can see the test images here. We have modified the parameters of each, some produce better than others but we are generally very happy with the new platform and the performance so far.
0 Comments