Usmann Khan

Can you provide the parameters used to reproduct the results in the paper? epochs, local epochs, learning rate, etc

startup: ``` INFO[09-26|16:43:34.300] Build info git_branch=alpha git_tag=v2022.09.03 git_commit=32bd69e5316050005e34448ec6b0165f97173d50 INFO[09-26|16:43:34.300] Starting Erigon on Ethereum mainnet... INFO[09-26|16:43:34.301] Maximum peer count ETH=100 total=100 INFO[09-26|16:43:34.301] starting HTTP APIs APIs=eth,erigon,engine INFO[09-26|16:43:34.301] torrent verbosity level=WRN INFO[09-26|16:43:36.404]...

Lighthouse finished syncing and now I'm seeing ``` INFO[09-27|06:33:15.259] [Downloader] Downloaded PoS Headers now=15605445 blk/sec=582.467 INFO[09-27|06:33:15.264] [txpool] stat block=14999999 pending=9798 baseFee=0 queued=3057 alloc=3.5GB sys=3.8GB INFO[09-27|06:33:45.259] [Downloader] Downloaded PoS Headers now=15596795...

``` print(pred_conf.size(), pred_cls.size()) print(pred_conf.view(num_samples,-1,1).size()) print(pred_cls.view(num_samples, -1, self.num_classes).size()) print((pred_boxes[..., :4].view(num_samples, -1, 4) * self.stride).size()) print(pred_boxes[..., 4:].view(num_samples, -1, 2).size()) ``` prints: ``` torch.Size([1, 3, 19, 19]) torch.Size([1, 3, 19, 19, 3]) torch.Size([1,...

~Thanks @usherbob. I assume you mean the filters=30 in the immediately prior layer. Do you know how it is calculated? Is it:~ ~[(3 classes) + 7] * (3 channels) =...

@usherbob Yeah I am but instead of converting lyft->kitti I converted this repo to work with the lyft dataset. What kind of performance are you seeing? It was training very...

Nope I never got that far. Did you try visualizing them first?

Did the visualisations look good? What did you do for z and height, just use the mean?

Hm, I don't know. The only thing I can think of is the height and z-position. Actually that is the reason I didn't continue. I had a suspicion that it...

How long did it take you to train your model? Like in min/epoch?