Skip to content

Releases: ezmsg-org/ezmsg-sigproc

v3.8.0

Choose a tag to compare

@cboulay cboulay released this 05 Sep 22:04
2523824

What's Changed

  • Resolve the operating axis from chunk_dim, and deprecate the axis settings by @cboulay in #239

Full Changelog: v3.7.0...v3.8.0

v3.7.0

Choose a tag to compare

@cboulay cboulay released this 04 Sep 18:37

What's Changed

  • Read the MLX cache after eviction has run, not before by @cboulay in #236
  • Assert the bound MLX honours, not the limit it was given by @cboulay in #238
  • Key state resets on the message layout, via baseproc's axis-aware default by @cboulay in #233
  • Concat: cache the axes that describe the stream, watch the ones it caches by @cboulay in #234
  • Downsample: the dimension is not configurable by @cboulay in #235

Full Changelog: v3.6.0...v3.7.0

v3.6.0

Choose a tag to compare

@cboulay cboulay released this 25 Aug 18:00
3807028

What's Changed

  • Add a causal half-Gaussian option to the Gaussian smoothing filter by @cboulay in #229
  • Give the parameterless math Units a usable SETTINGS type by @cboulay in #230

Full Changelog: v3.5.0...v3.6.0

v3.5.0

Choose a tag to compare

@cboulay cboulay released this 25 Aug 16:49
5a9a10b

What's Changed

  • Add the Anscombe variance-stabilizing transform and its inverse by @cboulay in #228

Full Changelog: v3.4.0...v3.5.0

v3.4.0

Choose a tag to compare

@cboulay cboulay released this 24 Aug 22:39
58cc93b

What's Changed

  • Bound the MLX buffer cache instead of letting it track the machine by @cboulay in #227

Full Changelog: v3.3.0...v3.4.0

v3.3.0

Choose a tag to compare

@cboulay cboulay released this 22 Aug 22:55

What's Changed

  • Make the evaluation barrier a graph decision (Materialize modes) by @cboulay in #220
  • Let the caller decide whether band power stalls on the device by @cboulay in #221
  • Document where a Materialize node belongs, with the measurements by @cboulay in #225
  • Fix Downsample and Log crashing on MLX arrays by @cboulay in #222
  • Stop paying for the EWMA bias correction once it rounds to 1 by @cboulay in #223
  • Fold the AffineTransform bias into the matmul instead of concatenating a ones column by @cboulay in #224

Full Changelog: v3.2.0...v3.3.0

v3.2.0

Choose a tag to compare

@cboulay cboulay released this 14 Aug 14:50
8784ed8

What's Changed

  • Copy, don't retain, message data carried across calls by @cboulay in #218
  • Use None, not inf, for "no limit" resample settings by @cboulay in #219

Full Changelog: v3.1.0...v3.2.0

v3.1.0

Choose a tag to compare

@cboulay cboulay released this 10 Aug 22:23
4490d6e

What's Changed

  • EWMA: hoist the filter axis to last before scipy's IIR loop by @cboulay in #216

Full Changelog: v3.0.0...v3.1.0

v3.0.0

Choose a tag to compare

@cboulay cboulay released this 09 Aug 23:26
073b01e

What's Changed

  • Affine/CAR: one channel-grouping spec, and kernels chosen from the weights by @cboulay in #212
  • Window: stop re-copying the backlog, and add opt-in batcher mode by @cboulay in #213

Full Changelog: v2.38.0...v3.0.0

v2.38.0

Choose a tag to compare

@cboulay cboulay released this 09 Aug 05:08
cbf6db7

What's Changed

  • Aggregate: ARGMIN/ARGMAX return coordinates; fix RangedAggregate async passthrough by @cboulay in #199
  • EWMA/scaler: add opt-in reset_on_resume for passthrough gaps by @cboulay in #200
  • ButterworthZeroPhase: replace the ring buffer with a carry-over tail by @cboulay in #201
  • HybridBuffer: configurable sample axis; preserve the input layout by @cboulay in #202
  • MLX SOS: pick the serial kernel by channel count, not just pole radius by @cboulay in #203
  • Bump pre-commit ruff to v0.15.12 to match the dev dependency by @cboulay in #204
  • Filter: split large scipy IIR chunks across threads by @cboulay in #205
  • Filter: give numpy FIR a real multichannel path by @cboulay in #206
  • Filter: call scipy's SOS kernel directly; drop dead _sosfilt_xp by @cboulay in #207

Full Changelog: v2.37.0...v2.38.0