Skip to content

Make EMA alpha configurable instead of hard-coded 0.5#56

Merged
lbedogni merged 1 commit into
mainfrom
fix/sciot-009-configurable-ema-alpha
Jul 13, 2026
Merged

Make EMA alpha configurable instead of hard-coded 0.5#56
lbedogni merged 1 commit into
mainfrom
fix/sciot-009-configurable-ema-alpha

Conversation

@ff225

@ff225 ff225 commented Jul 13, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • Adds an optional offloading_algo.ema_alpha setting (validated in src/sciot/config.py, default 0.5) so the EMA smoothing factor for device/edge per-layer inference times in RequestHandler.handle_device_inference_result is no longer hard-coded.
  • Documents the new key in src/server/settings.yaml and docs/config/server.full.yaml.

Closes #9

Test plan

  • uv run pytest tests/unit/test_config_validation.py tests/unit/test_request_handler_helpers.py -q
  • Fast test suite (/fast-tests) — 181 passed, 1 skipped

RequestHandler.handle_device_inference_result hard-coded alpha = 0.5 for
the EMA smoothing of device/edge per-layer inference times. Add an
optional offloading_algo.ema_alpha setting (validated in sciot.config,
defaulting to 0.5) and load it via load_offloading_ema_alpha_config().

Resolves #9
@lbedogni lbedogni merged commit b3996bd into main Jul 13, 2026
3 checks passed
@lbedogni lbedogni deleted the fix/sciot-009-configurable-ema-alpha branch July 13, 2026 14:26
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Make EMA alpha configurable instead of hard-coded 0.5

2 participants