ViniciusSilvestre

I'm trying to use sb3 for a RL project. The "issue" im facing is the following: i have two gpu's in my system right now a RTX 3070 and a RTX 2070 super. the 3070 is running two instance of the env at the same time while the 2070 is running one instance of the env (limitations due to the way i'm doing a grid search for the optimal rewards). during the episodes everything runs fine, but when the 2070 reaches the rollout stage, it moves all the processing to the 3070 and stays at 0% usage. this is despite the fact i specified the device the model should run at. The model is trying to learn how to play a game i developed. The way thing are now its still faster than running with just the 3070 because of VRAM limitations, it can only fit two instances of the model in the 8GB frame buffer, but i cant help but feel im leaving performance behind. any ideas what could be happening? and if so, how to fix it? i tried specifying the device even more with: < model = PPO('MultiInputPolicy', env, verbose=1, tensorboard_log=log_path, device='cuda:1') with torch.cuda.device('cuda:1'): model.learn(total_timesteps=80000, callback=callback) > but it didn't help A1100a.txt Env.py PlagueGame.py Run Model 1.py Run Model 2.py Run Model 3.py

yeah, probaly those cards are "special" build for this scenarios, but ptobably too expensive and hard to find

Some dreams just die hard i guess, not possible acording to them, the performace should be faster than an gtx 960 even with crossfire off like you sad but assing the gpus to diferent vms is not possible Edit: I had the idea, because the card have a plx chip in the pcb, so (at least in my head), a acs overide might work, but im not about to drop my student money on something that may or may not work based on a guess from me. side note: having a hd 7990 would be dope though

They are full Tahiti, down clocked a bit https://www.techpowerup.com/gpu-specs/radeon-hd-7990.c2126 Edit: (I read 580 instead of 570 ops, sorry) A new rx 570 4GB is going about R$1000,00 more or less $269,90 + shipping which is a bit over budget right now, was my first choice tough still a bit over budget, but better Yep, L1 Might a better place for that question

I was born in Valença RJ, mountain climate, in summer temperaures barely reach 30 Celsius and my college is in Petropolis RJ, which is even "colder", winter gets to -2 celsius and summer berely scrapes the 27 mark

i live in a city thats has a cold climate, that isnt much of a issue yeah, that 7970 ghz was a monster card back in its day the s10000 is a good idea, might be hard to find though

In a gaming sense i agree with you, if i can find a good deal on a 970 i would go for it, but im kind of waiting on the second part, the virtualization is key here, as the second pcie slot on that mobo shares a iommu grup with everything else(network, storage etc) so, if the two gpus on the 7990 dont share a group, it could still be worth for me

Im in a bit of a dilemma here, i have a gtx 960 4gb card and a ryzen 7 2700x in a asus prime b350 plus mobo heres the situation, the hd 7990 is going to be a better gaming card in crossfire and when crossfire doesnt work, the perfomace is going to be pretty close to my 960 (im aware of crossfire issues) but, aside from that, im working on a project in my college that involves virtualization, and the 7990 is two gpus (no shit sherlock haha), could i possibily assing each gpu to a single vm (they might share the same iommu group)? if possible, this could actualy be a good deal for me

i have a gtx 960, and i would want that rx 480 for 3d content creation