ns_autodeploy [ERROR] Model Configuration Send Status = 1

[pfeatherstone]

I'm evaluating my TFLite model an i get the following debug messages:

ns_autodeploy --tflite-file=classifier.tflite
2025-06-27 13:04:09.425420: I external/local_tsl/tsl/cuda/cudart_stub.cc:32] Could not find cuda drivers on your machine, GPU will not be used.
2025-06-27 13:04:09.428410: I external/local_tsl/tsl/cuda/cudart_stub.cc:32] Could not find cuda drivers on your machine, GPU will not be used.
2025-06-27 13:04:09.438774: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:479] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2025-06-27 13:04:09.460133: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:10575] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2025-06-27 13:04:09.460197: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1442] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2025-06-27 13:04:09.472212: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2025-06-27 13:04:10.151783: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT
projects/autodeploy
[NS] Model Name automatically set to: classifier
[NS] Running 2 Stage Autodeploy for Platform: apollo510_evb
[NS] Max Arena Size for apollo510_evb: 2458 KB
[NS] Best 142KB model location for apollo510_evb: TCM
[NS] Using AmbiqSuite Version: R5.2.0
[NS] Using TensorFlow Version: ns_tflm_2025_03_06
INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
[NS] Static Analysis Warning: TANH is not in optimized in mainstream TFLM, consider using Ambiq NS-TFLM
[NS] *** Stage [1/2]: Create and fine-tune EVB model characterization image
[NS] Compiling and deploying Baseline image: arena size = 2458k, arena location = SRAM model_location = TCM, Resource Variables count = 0
[ERROR] Model Configuration Send Status = 1
[ERROR] This may be caused by allocating too little memory for the tensor arena.
[ERROR] This script uses TFLM's arena_used_bytes() function to determine the arena size,
[ERROR] which has a bug where it does not account for scratch buffer padding.
[ERROR] To manually add a padding for scratch buffers, use the --arena-size-scratch-buffer-padding option.
Model Configuration Failed

Is there a way to get a more helpful message than:

[ERROR] Model Configuration Send Status = 1

The TFLite model is correct and runs fine in a normal enviornment using LiteRT

[pfeatherstone]

If i use a full float32 model it works, if i use a full int8 model, it works, if i use a partially quantized model, that's when it breaks with the error message above. The way i was exporting from Keras to TFLite (and quantizing at the same time) was:

    converter = tf.lite.TFLiteConverter.from_keras_model(model_keras)
    converter.optimizations = [tf.lite.Optimize.DEFAULT]
    model_tflite = converter.convert()
    with open(tflite_path, 'wb') as f:
        f.write(model_tflite)

Using tf.lite.Optimize.DEFAULT without representative_dataset uses partial quantization. Most of the layers are float32 except the conv2d layers which use int8 weights.
So, does neuralSPOT not support partial quantization?