Product & Environment
- Product: MLX-A1
- Compiler Version: 1.2.0
- OS: (container) mobilint/qbcompiler:1.2-cpu-ubuntu22.04
Description
The quantize and compile subcommands of python -m qbcompiler fail and the CLI cannot produce the .mxq model. Also the failure is not logged to the terminal.
I have confirmed the Python API (qbcompiler.mxq_compile) works fine.
Logs & Attachments
$ python3 -m qbcompiler compile --model simple-conv.onnx --backend onnx \
> --target-device aries-rb --output y.mxq --use-random-calib
[2026-08-05 12:53:05] qbcompiler.model_dict.common._reshape_util.reshape_check : [INFO] Parser is configured to use reshape_check_util (Python implementation).
2026-08-05 12:53:06 | WARNING | mblt.mblt.operator.activation_functions | ACTIVATION_FUNCTIONS and _OPS_PER_INPUT_ELEMENT keys mismatch: missing=['InverseSigmoid'], extra=[]
2026-08-05 12:53:06 | INFO | mblt.qbcompiler.model_dict_new.parser.backend.onnx.load_onnx_model | LOAD simple-conv.onnx, onnx opset=17, ir_version=8
2026-08-05 12:53:06 | INFO | mblt.qbcompiler.model_dict_new.parser.backend.onnx.inference | run onnx inference session
2026-08-05 12:53:06 | INFO | mblt.qbcompiler.model_dict_new.parser.backend.onnx.inference | onnx inference session done.
2026-08-05 12:53:06 | INFO | mblt.qbcompiler.model_dict_new.parser.backend.onnx.onnx_graph_builder | onnx opset=17, ir_version=8 optypes={'Relu', 'Conv'}
2026-08-05 12:53:06 | INFO | mblt.qbcompiler.model_dict_new.parser.parser |
==================================================================
Summary by Operator Type[Model Loaded]
==================================================================
OpType | Count | Unsupported | Supported(%) | GOPS | GMAC
---------------+-------+-------------+--------------+------+------
Convolution | 1| 1| 0.0| 3.70| 1.85
Input | 1| 1| 0.0| 0.00| 0.00
InputConstant | 2| 2| 0.0| 0.00| 0.00
Output | 1| 1| 0.0| 0.00| 0.00
Relu | 1| 1| 0.0| 0.00| 0.00
Transpose | 2| 2| 0.0| 0.00| 0.00
---------------+-------+-------------+--------------+------+------
All | 8| 8| 0.0| 3.70| 1.85
==================================================================
2026-08-05 12:53:06 | INFO | mblt.qbcompiler.model_dict_new.parser.transform.strip_yolo_decode_transform | Applying YOLO decode removal rules.
2026-08-05 12:53:06 | INFO | mblt.qbcompiler.model_dict_new.parser.transform.strip_yolo_decode_transform | ================ [OPTIMIZE stage=yolo decode removal, num_ops=8] ================
2026-08-05 12:53:06 | INFO | mblt.mblt.transform.basic_transform | optimize_level=999.
2026-08-05 12:53:06 | INFO | mblt.mblt.transform.basic_transform | ================ [OPTIMIZE stage=1, num_ops=6] ================
2026-08-05 12:53:06 | INFO | mblt.mblt.transform.basic_transform | ================ [OPTIMIZE stage=2, num_ops=6] ================
2026-08-05 12:53:06 | INFO | mblt.mblt.transform.basic_transform | ================ [OPTIMIZE stage=3, num_ops=6] ================
2026-08-05 12:53:06 | INFO | mblt.mblt.transform.basic_transform | ================ [OPTIMIZE stage=4, num_ops=6] ================
2026-08-05 12:53:06 | INFO | mblt.mblt.transform.basic_transform | ================ [OPTIMIZE stage=5, num_ops=6] ================
2026-08-05 12:53:06 | INFO | mblt.mblt.transform.basic_transform | ================ [OPTIMIZE stage=6, num_ops=6] ================
2026-08-05 12:53:06 | INFO | mblt.mblt.transform.basic_transform | ================ [OPTIMIZE stage=7, num_ops=6] ================
2026-08-05 12:53:06 | INFO | mblt.mblt.transform.basic_transform | ================ [OPTIMIZE stage=8, num_ops=6] ================
2026-08-05 12:53:06 | INFO | mblt.mblt.transform.basic_transform | ================ [OPTIMIZE stage=9, num_ops=6] ================
2026-08-05 12:53:06 | INFO | mblt.qbcompiler.model_dict_new.parser.parser |
================================================================
Summary by Operator Type[HL compilation done. Device=aries-rb]
================================================================
OpType | Count | Unsupported | Supported(%) | GOPS | GMAC
-------------+-------+-------------+--------------+------+------
Convolution | 1| 0| 100.0| 3.70| 1.85
Input | 1| 1| 0.0| 0.00| 0.00
Output | 1| 1| 0.0| 0.00| 0.00
Relu | 1| 0| 100.0| 0.00| 0.00
Transpose | 2| 2| 0.0| 0.00| 0.00
-------------+-------+-------------+--------------+------+------
All | 6| 4| 33.33| 3.70| 1.85
================================================================
2026-08-05 12:53:06 | INFO | mblt.qbcompiler.model_dict_new.parser.parser |
[Parser]
Number of Subgraphs: 1
graph order: [0]
subgraph DAG: persisted (0 edges, 1 parallel levels)
DAG levels (parallel groups):
L0: [sg0]
stateful_lstm_subgraphs: {}
backend: onnx
model inputs:
- name: images_channel_last, src_shape: (1,224,224,3)
model outputs:
- name: c0_channel_last/relu, src_shape: (1,224,224,16)
2026-08-05 12:53:06 | INFO | mblt.qbcompiler.model_dict_new.parser.parser | Write mblt: /tmp/qbc-compile-jzwhf1qy/intermediate.mblt
$ echo $?
1
$ ls -l y.mxq
ls: cannot access 'y.mxq': No such file or directory
simple-conv.onnx (2.0 KB)