Really cool! But right as it was nearing 4,000, it seems to have corrupted itself and no longer got any scores above 0. Not sure if that's a code bug or a neural net issue.
avg500 -4.6 last 500 episodes
peak 3959.3 best window
roll/s 20.68 20-step avg
progress 4388 562749 episodes
c1b 9 hours ago [-]
Yes it just collapses eventually — never stabilizes.
The training process is flawed, I suspect it has to do with the fact that some weights blow up over time, you can see in “weights” tab.
But at around 4K avg score you should see it solve the env almost every time.
Just a demo :) optimized for speed over stability.
Reward structure: Step: -1 Dot: +100 Win: +1000
so ~4k is max theoretical score on 6x6.
ticulatedspline 6 hours ago [-]
maybe because it doesn't understand "done"? perfect play is impossible, random variance will cause scores to drop even if the model plays well and "wins". feels like it would get stuck in a loop trying to improve what can't be improved.
fc417fc802 5 hours ago [-]
The optimizer doesn't need to understand anything it's just an iterated mathematical construct. The author simply didn't bother to implement the necessary details to ensure numerical stability.
Alternatively it might be a problem with the scoring model in the end game.
jesuo 5 hours ago [-]
feels like it would get stuck in a loop trying to improve what can't be improved.
That is the point, there is nothing on an intention that we cannot improve, the goal here is no more than 1 unique iteration of the same path
r3trohack3r 10 hours ago [-]
I think I noticed it reach “end game.” The snake reaches a point where, if it gets any longer, it is out of squares and hits its own tail. So it finds the route through the squares that it can infinitely loop, never eats the ball, and score starts dropping and goes negative.
9 hours ago [-]
9 hours ago [-]
starshadowx2 3 hours ago [-]
FYI this website sets off a bunch of Bitdefender alerts as being a suspicious web page. I assume probably false positives or something but still something you might want to look into.
"The page https://ppo.gradexp.xyz/ has been detected with suspicious activity. It is not recommended to continue browsing this website."
I noticed snake gets penalized for not getting to the apple early, is that what you really want? Snake is about how long it gets not about the balance between length and wall clock time
foo12bar 7 hours ago [-]
But if not the snake could go into an infinite loop, never growing, never eating.
jesuo 7 hours ago [-]
Poorly programmed, it doesn't learn from its mistakes, the games get stuck in a loop because the snake doesn't capture a piece but the piece remains and there's a gap, constantly moving the snake along the same path with negative scores in an infinite loop leaving an unaltered yin and yang ;) there's a repetitive pattern in these infinite games between the position of the gap and the piece
spectre9 7 hours ago [-]
Did you let it train? This doesn’t happen for me
jesuo 6 hours ago [-]
Yes, thousands of games, you can see how it happens in the displayed game matrix, there comes a point when they all enter those loops
https://ibb.co/bM4RPzPb
spectre9 6 hours ago [-]
Makes sense, author mentioned training collapses eventually
12 hours ago [-]
beardsciences 12 hours ago [-]
My average eventually made it to about 3900, and then stagnated between 3600-3900. I'm curious if this is universal behavior or not. I'm up to about 5k steps.
neduma 12 hours ago [-]
More details and implementation notes please?
cshimmin 8 hours ago [-]
It's on the page, if you click the little info icon in the upper-right. Here's the text but there's some nice graphics there too:
Snake Game, training entirely in the browser. Built on tinygrad: the rollout / targets / train graphs are TinyJits authored in Python, then compiled once to WGSL and replayed here under WebGPU.
Observation: flat 10×10 board (100) + 4-dim prev-action one-hot = 104 dims. fc_pi.weight is zero-init so the opening policy is uniform over the legal actions; fc_v uses tinygrad's default Kaiming init.
Per rollout: T=24 × N=384 parallel snakes (9,216 transitions), then K=3 epochs × 4 mini-batches of PPO updates. GAE γ=0.99, λ=0.95; AdamW wd=0.01; ratio clip ε=0.1; grad-norm 0.5; Huber value β=1, val_coef=1; entropy bonus 0.008333333333333333.
Action mask + value clip + KL early stop. The 4-dim prev_a obs tail lets fc_pi zero the U-turn logit (the env silently overrides same-axis reversals anyway). Value loss is max(huber(v_new−td), huber(v_clip−td)) at ε=0.2. Approx-KL is sampled after each epoch and breaks the loop at 1.5·kl_target.
mavdol04 8 hours ago [-]
That's cool, i did exactly the same few years ago
LowLevelKernel 10 hours ago [-]
Link to repo?
insane_dreamer 4 hours ago [-]
sound cool; would like to show my kid for education; doesn't work on Mac/Safari though (no webGPU)
bozhark 9 hours ago [-]
Crashed
th1nhng0 11 hours ago [-]
cool project
jmclnx 11 hours ago [-]
> WebGPU not available in this browser
Looks like this is for Linux and Windows, on NetBSD I get this issue :(
NoboruWataya 10 hours ago [-]
I got this in Firefox on Linux, just had to enable WebGPU in about:config (`dom.webgpu.enabled` = true).
jmclnx 10 hours ago [-]
Did not know that existed, I enabled it but no luck. Must be a NetBSD thing based upon this new message:
> WebGPU is not yet available in Release or late Beta builds.
redshiftza 10 hours ago [-]
If you are using brave (which i assume also applies to chrome) , there is a menu at brave://flags , you can enable unsafe web GPU from there
austinthetaco 11 hours ago [-]
my training on a 10x10 just randomly broke. i got to like 3600 then the graph went flat, the viewer on the left just showed it constantly restarting the game, and the scores in the negative. my average is now -10.
Rendered at 04:30:37 GMT+0000 (Coordinated Universal Time) with Vercel.
avg500 -4.6 last 500 episodes
peak 3959.3 best window
roll/s 20.68 20-step avg
progress 4388 562749 episodes
But at around 4K avg score you should see it solve the env almost every time.
Just a demo :) optimized for speed over stability.
Reward structure: Step: -1 Dot: +100 Win: +1000 so ~4k is max theoretical score on 6x6.
Alternatively it might be a problem with the scoring model in the end game.
That is the point, there is nothing on an intention that we cannot improve, the goal here is no more than 1 unique iteration of the same path
"The page https://ppo.gradexp.xyz/ has been detected with suspicious activity. It is not recommended to continue browsing this website."
Same for:
https://ppo.gradexp.xyz/version.js
https://ppo.gradexp.xyz/dist/sizes.js
https://ppo.gradexp.xyz/dist/size_6/manifest.j
https://ppo.gradexp.xyz/dist/size_6/weights.safetens
https://ppo.gradexp.xyz/dist/sokol/demo.wa
I noticed that if you go from training to watch and then back, the training temporarily drop significantly in score.
trained and made a viz for the model and then made it displace text.
should probably do a proper write-up:https://x.com/i/status/2038367016969724259
Looks like this is for Linux and Windows, on NetBSD I get this issue :(
> WebGPU is not yet available in Release or late Beta builds.