fix(tensor): throw on undefined tensor access (#22)

Merge pull request #23 from Nozom1ff/perf/matmul

fix(tests): fix build failed.

feat(matmul): add 3D x 3D batch matrix multiplication support

Merge pull request #21 from Nozom1ff/master

Merge branch 'feat/rms_norm'

feat(op): rms_norm with CPU and CUDA support

perf(cuda): optimize cuda reduce sum using warp shuffle.

Merge pull request #20 from Nozom1ff/master

feat(nn): implement Embedding layer with CPU and CUDA support

feat(layer): embedding - header file

perf(mat): use float4-optimized v7 matmul kernel to reach 89% of PyTorch

chore(build): remove contributor's CMake changes.

Merge pull request #19 from Nozom1ff/feat/leaky-relu

feat(op): leaky_relu