AI/ML●●●Banger
Watch a neural net discover molecules by arguing with itself
Two RL clones duel in-browser with epiplexity selection — no human tuning the explore-exploit knob.
WizardryRabbit HoleBig Brain
cmaicurious
102mo ago
![Weight Scope, watch and modify training of Deep Nets in real-time [video]](/screenshots/48993127.webp)
Real-time weight perturbation during training is a genuinely clever way to visualize network resilience.
ML researchers and students studying network dynamics
TensorBoard · Weights & Biases · Netron
Two RL clones duel in-browser with epiplexity selection — no human tuning the explore-exploit knob.
4-month VAE research artifact; reconstruction quality matters less than you'd think.
U-Net with 3-layer pixel-stack encoding emulates Foveon X3 sensors better than Lightroom presets.
16M trained params beats Gemini Embedding 2 on audio-text retrieval.
From-scratch 1B model training for $315 inverts the massive-budget assumption.
Fits a 325M model in 6GB VRAM where STE and float baselines crash.