Portuguese
mamba
mamba-3
ssm
muon
muonclip
qk-clip
adamw
optimizer
nope
hybrid-architecture
activation-checkpointing
3d-vision-language
depth-perception
point-cloud
diffusion
stable-diffusion
mixture-of-experts
kohonen
som-attention
random-forest
deep-neural-network
knowledge-distillation
bidirectional-distillation
engram-memory
multimodal
multi-token-prediction
dpo
rlhf
hh-rlhf
pcgrad
long-context
lra-qvit
engineer-agent
integral-state
cython
jev-classifiers
gumbel-softmax
quantization
riemannian-optimization
stiefel-manifold
cython-kernels
gradient-accumulation
vector-quantization
grayscale-detection
joint-perception
self-regulation
signal-table
abmo
ciar
growing-som
pt-br
Commit History
19 days ago
-
KHTST v11 — MAMBA-3 HÍBRIDO + NoPE (5:1) + MAMBA-3VL 3D-VL + CHECKPOINTING DE ATIVAÇÃO + ENSEMBLE RF×DNN×DESTILAÇÃO BIDIRECIONAL×ENGRAEM + 9 DATASETS DO REQUISITO + RLHF REAL: 12 tarefas com FONTE ÚNICA (Teo 22.1); N_MAX_TOKENS = entrada 256K + saída 128K = 393216 (Teo 22.1.1); critérios Jev DINÂMICOS de dados reais (DESCRICOES_TAREFAS removida, Teo 22.2); kernel trapezoidal estável com gradiente exato (Teos 22.2–22.5); checkpointing de ativação exato (Teo 22.8, Δ=0 medido); Hedge com regreto O(√(T ln M)) e Gibbs local (Teos 23.1–23.4); KD simétrica mode-covering+seeking (Teo 23.8); engram k-WTA (Teo 23.10); estados no Hub antes de 900MB com retomada real (Teo 23.13); orçamento 8h+1h; teste com DADOS REAIS APROVADO antes de treinos longos
Verified signature3c2b90f