Build hangs at first step

[fpjentzsch]

When I try building the cybersecurity finn-example on the dev/tafk/de-docker branch (#61), the build hangs at the first step, no matter what the first step is. The build_dataflow.log is not created.

I don't think it is related to my environment setup, because the ./examples/bert/quicktest.sh works fine.

Steps to reproduce:

1. Clone Brainsmith (tested at this commit)
2. ./setup-venv.sh & source .venv/bin/activate
3. Get cybersecurity ONNX model from https://github.com/Xilinx/finn-examples/blob/main/build/cybersecurity-mlp/models/download-model.sh
4. Run smith unsw_nb15-mlp-w2a2.onnx cybsec.yaml with the following cybsec.yaml:

name: "cybsec_test"
clock_ns: 10.0
output: "estimates"
board: "RFSoC2x2"
save_intermediate_models: true
parallel_builds: 1
finn_config:
  target_fps: 1000000
  mvau_wwidth_max: 80
  output_dir: /home/felixj/WD/Brainsmith/examples/cybersecurity/build
design_space:
  kernels:
  kernels:
    - Thresholding
    - MVAU
  steps:
    - finn:qonnx_to_finn # happens even when I remove this line
    - "tidy_up"
    - "streamline"
    - "convert_to_hw"

I see this output:

Brainsmith DSE - Design Space Exploration
Model: unsw_nb15-mlp-w2a2.onnx
Blueprint: cybsec.yaml
[10/10/25 10:19:16] INFO Starting DSE with model=unsw_nb15-mlp-w2a2.onnx, blueprint=cybsec.yaml, output_dir=None dse.py:43
INFO Exploring design space for FPGA accelerator: dse_api.py:62
INFO Model: unsw_nb15-mlp-w2a2.onnx dse_api.py:63
INFO Blueprint: cybsec.yaml dse_api.py:64
INFO Output: /home/felixj/tmp_brainsmith/dse_20251010_101916 dse_api.py:65
[10/10/25 10:19:17] INFO Registered 60 qonnx transforms framework_adapters.py:376
INFO Registered 80 finn transforms framework_adapters.py:376
INFO Registered 27 finn backends framework_adapters.py:431
[10/10/25 10:19:27] INFO Registered 19 finn steps framework_adapters.py:514
INFO Framework initialization complete: framework_adapters.py:642
INFO - QONNX transforms: 60 framework_adapters.py:643
INFO - FINN transforms: 80 framework_adapters.py:644
INFO - FINN kernels: 35 framework_adapters.py:645
INFO - FINN backends: 27 framework_adapters.py:646
INFO - FINN kernel inference transforms: 18 framework_adapters.py:647
INFO - FINN build steps: 19 framework_adapters.py:648
INFO Design space: 4 steps, 2 kernels dse_api.py:74
INFO DSE tree: dse_api.py:79
INFO - Total paths: 1 dse_api.py:80
INFO - Total segments: 1 dse_api.py:81
INFO - Segment efficiency: 0.0% dse_api.py:82
INFO Starting design space exploration... dse_api.py:85
Executing tree with fail_fast=False
Output: /home/felixj/tmp_brainsmith/dse_20251010_101916
→ root

→ Executing: root
INFO FINN build: input=/home/felixj/tmp_brainsmith/dse_20251010_101916/root/input.onnx, output=/home/felixj/tmp_brainsmith/dse_20251010_101916/root finn_runner.py:95
INFO Executing FINN build with 4 steps finn_runner.py:121
Building dataflow accelerator from /home/felixj/tmp_brainsmith/dse_20251010_101916/root/input.onnx
Intermediate outputs will be generated in /home/felixj/tmp_brainsmith
Final outputs will be generated in .
Build log is at ./build_dataflow.log
Running step: step_qonnx_to_finn [1/4]

[tafk7]

The issue was a deadlock between FINN's stdout redirection and Rich Console logging, I've resolved the issue for now by always setting "verbose" to true in the DataflowBuildConfig. c8aa82f

FYI there's a lot of additional debug statements on that commit to troubleshoot an intermittent finn_xsi error I'm seeing only in CI/CD builds, but those will be removed before the merge to develop.