ssd-mobilenet.tflite.mlir.config.yaml

---

name: ssd-mobilenet-tflite
gops: [2.715259726]
shapes:
  - [1, 3, 320, 320]


mlir_transform:
  model_transform.py
    --model_name $(name)
    --model_def
    $(home)/lite-model_ssd_mobilenet_v1_100_320_uint8_default_1.tflite
    --pixel_format=rgb
    --test_input=$(root)/dataset/samples/cat.jpg
    --input_shapes [$(shape_param)]
    --test_result $(name)_top_outputs.npz
    --mlir $(workdir)/transformed.mlir

BM1684X:
  deploy:
    - model_deploy.py  --mlir $(workdir)/transformed.mlir
        --quantize INT8
        --asymmetric
        --chip bm1684x
        --test_input $(workdir)/$(name)_in_f32.npz
        --test_reference $(name)_top_outputs.npz
        --model $(workdir)/$(name)_bm1684x_int8_asym.bmodel