~/talk

DimWit - Tensor programming with wit

Talkat Scala Days 2026 · 2026-10-12..10-13· [machine-learning, typescript]
links:none on file
links:none on file
slides:none on file

DimWit implements named tensor notation in Scala by encoding dimension names in types, letting the compiler verify axis-aligned operations; it wraps JAX via Scala Py and avoids reimplementing autodiff, JIT, or GPU code. Tracing overhead rises modestly (roughly 1s to 2s) but runtime uses JIT-compiled native code, so execution cost is unchanged. A full trained GPT-2 124M model is trained on 10 billion tokens in about three days on a DGX Spark GPU and runs on par with JAX.

AI-written summary · about

  1. 0:00Introduction to Dimwit
  2. 1:13Limitations of JAX in Python
  3. 2:24Named tensor dimensions
  4. 3:36Architecture and performance
  5. 4:51Verification and conclusion
~/search
type to search 6,518 conferences, 143,918 speakers and 192,133 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
← →
on a talk: 5 seconds back / forward
0–9
on a talk: jump to 0%–90% of the recording
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.