Nothing helps at all.
Ok so this all started off approximately 4 days ago. My fortnite application began acting up, so I uninstalled and reinstalled. After doing so, my cursor would be able to move to my other monitor when the selected video setting was fullscreen. I tested other games, and the issue did not occur in Aimgod, but occurred in both CS:GO and apex legends. For the past three days, I have spent 12+ hours a day scouring the internet looking for a fix to this issue. I have uninstalled and reinstalled graphics driver tens of times (safe mode, ddu, without geforce experience, after factory resetting pc etc.) I have followed the fullscreen optimization registry fix, no luck. I have uninstalled and reinstalled monitor drivers tens of times. I have factory reset pc 3+ times, formatted hard drives, all of the fixes that I could find online. After going through 14 hours of troubleshooting steps today, it now appears that EVERY game in windows is now having this issue. It is extremely frustrating as I am a semi professional player, and am missing out greatly on practice and tournaments. I really don’t know what to do. My current ideas are: Hard drives still contain incorrect files, registry will not reset and is continuing to run incorrect files, graphics card is completely fucked even though partially functioning correctly, monitor is screwed (have talked to ACER support about this issue, $500 monitor and there is no downloadable windows 10 driver, only running generic pnp when it used to display as Acer Predator XB271h. Also, my monitor has been glitching out when using a stretched resolution, changing resolution in general (Windows sets it to 1600x1050 ((not sure if that’s specific number but it sets it to 16;10 ( not sure x 1050)). When this issue occurs and I change my resolution to 1920x1080, it sets my monitors refresh rate to 60hz, and when I change the refresh rate back to 144hz, the resolution goes to 640x480p. Please if anyone could help me in any way that would be greatly appreciated. Don’t need common answers, as I’ve already scoured the internet on thousands of forums trying to find fixes. Please, can a technological genius help me out.
submitted by
reverttoseason3 to
Windows10
[D] Determined: new PyTorch API, Model Registry, and optimized HP tuning
Hi everyone,
It's been a few months since we
announced Determined's open-source
deep learning training platform on Reddit. We've been thrilled at the reception we've gotten so far, and we really appreciate all the feedback. We've been working hard to improve Determined, in part based on the feedback we got from Reddit last time, so I wanted to share a few updates:
(1)
New PyTorch API Our initial PyTorch API supported relatively simple models with a single model graph, optimizer, and learning rate scheduler. That made it difficult to define more complex models like GANs. We've since revised this API to make it more flexible to support any number of models, optimizers and LR schedulers, and to give developers control over how optimizers/LR schedulers are used in the training loop. Check out
this example of a GAN in Determined. We would love feedback on the new API -- the
PyTorch tutorial walks through how to take an existing PyTorch model and adapt it to work with Determined.
(2)
Built-in Model Registry Determined now includes a lightweight model registry to make it simpler to identify promising models, to version those models, and to promote models from research to production. There are simple APIs to manage model versions and load a model from the registry. For more details, check out this
blog post or the
model registry docs.
(3)
Optimized HP tuning Determined has supported built-in hyperparameter tuning for a long time, but we recently tuned our Hyperband implementation to significantly improve its efficiency and scalability. The new implementation is 20-30% faster, depending on the model and hyperparameter space. This blog post on
how to build an HP tuning system in practice has some benchmarks of the new algorithm.
(4)
Kubernetes Support Determined doesn't require Kubernetes, but if you're using Kubernetes to manage your GPUs, Determined can now integrate into a Kubernetes-based cluster in a natural way. Check out this
blog post on Determined + k8s for more details.
As always, we would love your feedback!
submitted by
neilc to
MachineLearning
