feat: 2.6.0 inference#174
Merged
calvinleng-science merged 20 commits intomainfrom Apr 24, 2026
Merged
Conversation
Replace host-side SNPE converter invocation with a Docker-based approach. The container (Python 3.10 + pinned deps) eliminates Python version and numpy compatibility issues on the host. - Add model-converter/ with Dockerfile and self-contained convert.py - Rewrite onnx_to_dlc.py to orchestrate Docker (auto-builds image) - Bind-mount SNPE SDK at runtime (Qualcomm license compliant) - Add --snpe-root CLI arg to deploy-model - Remove unused onnx_transforms.py (logic moved into container) - Fix -u shorthand conflict between --username and global --uri
…w up on overwrite confirmation
…runtime, updated build to package onnxruntime into the deb
…in scifi-headstage-shared-libraries into the resulting app .deb as that blocks installations of scifi-headstage-shared-libraries. additionally, tap names will now wrap instead of truncate in the rich UI
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
This introduces the command "deploy-model" that allows users to either deploy a .pt or .onnx model onto the device.
If the user specifies --quantize and --input-list, the model can be used by the DSP runtime. To allow DSP runtime, the user must specify the root directory for v2.34 of Qualcomm's QAIRT to enable conversion of their model to .dlc. This is required since Qualcomm does not allow redistribution of their software.
If the user does not specify --quantize and --input-list, then the model will simply be uploaded as an .onnx model (we will convert .pt models to .onnx) and simply just use onnxruntime's CPU runtime.
Changes
Testing
Students at Neurohack 2026 were able to use our inference pipeline with Synapse Apps and the deploy-model utility. As for DSP runtime, I have tested it with my own apps and have confirmed that we are able to perform inference with the DSP runtime.
Example DSP runtime user-flow:
Example CPU runtime user-flow:
