rockchip-linux
ｐａｄｄｌｅＯＣＲ：ｏｎｎｘ模型转ｒｋｎｎ报错。
环境：ｕｂｕｎｔｕ１８.０４ ，ｒｋ３３９９ｐｒｏ
省略前面初始化部分 ：
I End importing onnx... done --> Building model I Generate input meta ... I Load input meta I Generate input meta ... D import clients finished I Load net... I Load data... I Load input meta I Start quantization... W:tensorflow:From /data/pengshan/miniconda3/envs/board/lib/python3.6/site-packages/rknn/api/rknn.py:278: The name tf.reset_default_graph is deprecated. Please use tf.compat.v1.reset_default_graph instead.
D import clients finished
D iterations: 2, batch_size: 1
I Quantization start...
D set up a quantize net
D *********** Setup input meta ***********
D import clients finished
D *********** Setup database (1) ***********
D Setup provider layer "text_input_layer":
D Lids: ['x_355']
D Shapes: [[1, 960, 960, 3]]
D Data types: ['float32']
D Sparse tensors: []
D Tensor names(H5FS only): []
W:tensorflow:From /data/pengshan/miniconda3/envs/board/lib/python3.6/site-packages/tensorflow/python/ops/control_flow_ops.py:1814: py_func (from tensorflow.python.ops.script_ops) is deprecated and will be removed in a future version.
Instructions for updating:
tf.py_func is deprecated in TF V2. Instead, there are two
options available in V2.
- tf.py_function takes a python function which manipulates tf eager
tensors instead of numpy arrays. It's easy to convert a tf eager tensor to
an ndarray (just call tensor.numpy()) but having access to eager tensors
means tf.py_functions can use accelerators such as GPUs as well as
being differentiable using a gradient tape.
- tf.numpy_function maintains the semantics of the deprecated tf.py_func
(it is not differentiable, and manipulates numpy arrays). It drops the
stateful argument making all functions stateful.
D Add preprocess "[('reverse_channel', False), ('mean', [0.485, 0.456, 0.406]), ('scale', [4.366812227074235, 4.464285714285714, 4.444444444444445]), ('preproc_node_params', ordereddict([('add_preproc_node', False), ('preproc_type', 'IMAGE_RGB'), ('preproc_perm', [0, 1, 2, 3])]))]" for "x_355" D *********** Setup input meta complete *********** D Process x_355 ... D RKNN output shape(input): (1 960 960 3) D Tensor @x_355:out0 type: asymmetric_affine D Real output shape: (1, 960, 960, 3) D Process Conv_Conv_0_259 ... D RKNN output shape(convolution): (1 480 480 8) D Tensor @Conv_Conv_0_259:out0 type: asymmetric_affine 2022-06-07 15:18:43.700453: W tensorflow/compiler/jit/mark_for_compilation_pass.cc:1412] (One-time warning): Not using XLA:CPU for cluster because envvar TF_XLA_FLAGS=--tf_xla_cpu_global_jit was not set. If you want XLA:CPU, either set that envvar, or use experimental_jit_scope to enable XLA:CPU. To confirm that XLA is active, pass --vmodule=xla_compilation_cache=1 (as a proper command-line flag, not via TF_XLA_FLAGS) or set the envvar XLA_FLAGS=--xla_hlo_profile. D Real output shape: (1, 480, 480, 8) D Process Initializer_Constant_1_270 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_0_260 ... D RKNN output shape(add): (1 480 480 8) D Tensor @Add_Add_0_260:out0 type: asymmetric_affine D Real output shape: (1, 480, 480, 8) D Process Clip_Clip_0_247 ... D RKNN output shape(relun): (1 480 480 8) D Tensor @Clip_Clip_0_247:out0 type: asymmetric_affine D Real output shape: (1, 480, 480, 8) D Process Mul_Mul_0_233 ... D RKNN output shape(multiply): (1 480 480 8) D Tensor @Mul_Mul_0_233:out0 type: asymmetric_affine D Real output shape: (1, 480, 480, 8) D Process Initializer_Constant_0_234 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_0_218 ... D RKNN output shape(Divide): (1 480 480 8) D Tensor @Div_Div_0_218:out0 type: asymmetric_affine D Real output shape: (1, 480, 480, 8) D Process Conv_Conv_1_293 ... D RKNN output shape(convolution): (1 480 480 8) D Tensor @Conv_Conv_1_293:out0 type: asymmetric_affine D Real output shape: (1, 480, 480, 8) D Process Relu_Relu_0_282 ... D RKNN output shape(relu): (1 480 480 8) D Tensor @Relu_Relu_0_282:out0 type: asymmetric_affine D Real output shape: (1, 480, 480, 8) D Process Conv_Conv_2_271 ... D RKNN output shape(convolution): (1 480 480 8) D Tensor @Conv_Conv_2_271:out0 type: asymmetric_affine D Real output shape: (1, 480, 480, 8) D Process Relu_Relu_1_248 ... D RKNN output shape(relu): (1 480 480 8) D Tensor @Relu_Relu_1_248:out0 type: asymmetric_affine D Real output shape: (1, 480, 480, 8) D Process Conv_Conv_3_235 ... D RKNN output shape(convolution): (1 480 480 8) D Tensor @Conv_Conv_3_235:out0 type: asymmetric_affine D Real output shape: (1, 480, 480, 8) D Process Add_Add_1_206 ... D RKNN output shape(add): (1 480 480 8) D Tensor @Add_Add_1_206:out0 type: asymmetric_affine D Real output shape: (1, 480, 480, 8) D Process Conv_Conv_4_205 ... D RKNN output shape(convolution): (1 480 480 32) D Tensor @Conv_Conv_4_205:out0 type: asymmetric_affine D Real output shape: (1, 480, 480, 32) D Process Relu_Relu_2_176 ... D RKNN output shape(relu): (1 480 480 32) D Tensor @Relu_Relu_2_176:out0 type: asymmetric_affine D Real output shape: (1, 480, 480, 32) D Process Conv_Conv_5_153 ... D RKNN output shape(convolution): (1 240 240 32) D Tensor @Conv_Conv_5_153:out0 type: asymmetric_affine D Real output shape: (1, 240, 240, 32) D Process Relu_Relu_3_111 ... D RKNN output shape(relu): (1 240 240 32) D Tensor @Relu_Relu_3_111:out0 type: asymmetric_affine D Real output shape: (1, 240, 240, 32) D Process Conv_Conv_6_91 ... D RKNN output shape(convolution): (1 240 240 16) D Tensor @Conv_Conv_6_91:out0 type: asymmetric_affine D Real output shape: (1, 240, 240, 16) D Process Conv_Conv_7_192 ... D RKNN output shape(convolution): (1 240 240 40) D Tensor @Conv_Conv_7_192:out0 type: asymmetric_affine D Real output shape: (1, 240, 240, 40) D Process Relu_Relu_4_177 ... D RKNN output shape(relu): (1 240 240 40) D Tensor @Relu_Relu_4_177:out0 type: asymmetric_affine D Real output shape: (1, 240, 240, 40) D Process Conv_Conv_8_154 ... D RKNN output shape(convolution): (1 240 240 40) D Tensor @Conv_Conv_8_154:out0 type: asymmetric_affine D Real output shape: (1, 240, 240, 40) D Process Relu_Relu_5_112 ... D RKNN output shape(relu): (1 240 240 40) D Tensor @Relu_Relu_5_112:out0 type: asymmetric_affine D Real output shape: (1, 240, 240, 40) D Process Conv_Conv_9_92 ... D RKNN output shape(convolution): (1 240 240 16) D Tensor @Conv_Conv_9_92:out0 type: asymmetric_affine D Real output shape: (1, 240, 240, 16) D Process Add_Add_2_55 ... D RKNN output shape(add): (1 240 240 16) D Tensor @Add_Add_2_55:out0 type: asymmetric_affine D Real output shape: (1, 240, 240, 16) D Process Conv_Conv_10_231 ... D RKNN output shape(convolution): (1 240 240 40) D Tensor @Conv_Conv_10_231:out0 type: asymmetric_affine D Real output shape: (1, 240, 240, 40) D Process Relu_Relu_6_202 ... D RKNN output shape(relu): (1 240 240 40) D Tensor @Relu_Relu_6_202:out0 type: asymmetric_affine D Real output shape: (1, 240, 240, 40) D Process Conv_Conv_11_187 ... D RKNN output shape(convolution): (1 120 120 40) D Tensor @Conv_Conv_11_187:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 40) D Process Relu_Relu_7_148 ... D RKNN output shape(relu): (1 120 120 40) D Tensor @Relu_Relu_7_148:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 40) D Process Conv_Conv_12_127 ... D RKNN output shape(convolution): (1 120 120 24) D Tensor @Conv_Conv_12_127:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 24) D Process Conv_Conv_13_217 ... D RKNN output shape(convolution): (1 120 120 64) D Tensor @Conv_Conv_13_217:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 64) D Process Relu_Relu_8_203 ... D RKNN output shape(relu): (1 120 120 64) D Tensor @Relu_Relu_8_203:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 64) D Process Conv_Conv_14_188 ... D RKNN output shape(convolution): (1 120 120 64) D Tensor @Conv_Conv_14_188:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 64) D Process Relu_Relu_9_149 ... D RKNN output shape(relu): (1 120 120 64) D Tensor @Relu_Relu_9_149:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 64) D Process Conv_Conv_15_128 ... D RKNN output shape(convolution): (1 120 120 24) D Tensor @Conv_Conv_15_128:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 24) D Process Add_Add_3_87 ... D RKNN output shape(add): (1 120 120 24) D Tensor @Add_Add_3_87:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 24) D Process Conv_Conv_16_204 ... D RKNN output shape(convolution): (1 120 120 64) D Tensor @Conv_Conv_16_204:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 64) D Process Relu_Relu_10_189 ... D RKNN output shape(relu): (1 120 120 64) D Tensor @Relu_Relu_10_189:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 64) D Process Conv_Conv_17_172 ... D RKNN output shape(convolution): (1 120 120 64) D Tensor @Conv_Conv_17_172:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 64) D Process Relu_Relu_11_129 ... D RKNN output shape(relu): (1 120 120 64) D Tensor @Relu_Relu_11_129:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 64) D Process Conv_Conv_18_107 ... D RKNN output shape(convolution): (1 120 120 24) D Tensor @Conv_Conv_18_107:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 24) D Process Add_Add_4_70 ... D RKNN output shape(add): (1 120 120 24) D Tensor @Add_Add_4_70:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 24) D Process Conv_Conv_19_350 ... D RKNN output shape(convolution): (1 120 120 120) D Tensor @Conv_Conv_19_350:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 120) D Process Initializer_Constant_3_353 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_5_351 ... D RKNN output shape(add): (1 120 120 120) D Tensor @Add_Add_5_351:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 120) D Process Clip_Clip_1_347 ... D RKNN output shape(relun): (1 120 120 120) D Tensor @Clip_Clip_1_347:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 120) D Process Mul_Mul_1_341 ... D RKNN output shape(multiply): (1 120 120 120) D Tensor @Mul_Mul_1_341:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 120) D Process Initializer_Constant_2_334 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_1_333 ... D RKNN output shape(Divide): (1 120 120 120) D Tensor @Div_Div_1_333:out0 type: asymmetric_affine D Real output shape: (1, 120, 120, 120) D Process Conv_Conv_20_323 ... D RKNN output shape(convolution): (1 60 60 120) D Tensor @Conv_Conv_20_323:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 120) D Process Initializer_Constant_5_335 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_6_324 ... D RKNN output shape(add): (1 60 60 120) D Tensor @Add_Add_6_324:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 120) D Process Clip_Clip_2_316 ... D RKNN output shape(relun): (1 60 60 120) D Tensor @Clip_Clip_2_316:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 120) D Process Mul_Mul_2_307 ... D RKNN output shape(multiply): (1 60 60 120) D Tensor @Mul_Mul_2_307:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 120) D Process Initializer_Constant_4_308 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_2_301 ... D RKNN output shape(Divide): (1 60 60 120) D Tensor @Div_Div_2_301:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 120) D Process Conv_Conv_21_295 ... D RKNN output shape(convolution): (1 60 60 40) D Tensor @Conv_Conv_21_295:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 40) D Process Conv_Conv_22_352 ... D RKNN output shape(convolution): (1 60 60 104) D Tensor @Conv_Conv_22_352:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 104) D Process Initializer_Constant_7_354 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_7_349 ... D RKNN output shape(add): (1 60 60 104) D Tensor @Add_Add_7_349:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 104) D Process Clip_Clip_3_344 ... D RKNN output shape(relun): (1 60 60 104) D Tensor @Clip_Clip_3_344:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 104) D Process Mul_Mul_3_342 ... D RKNN output shape(multiply): (1 60 60 104) D Tensor @Mul_Mul_3_342:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 104) D Process Initializer_Constant_6_337 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_3_336 ... D RKNN output shape(Divide): (1 60 60 104) D Tensor @Div_Div_3_336:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 104) D Process Conv_Conv_23_325 ... D RKNN output shape(convolution): (1 60 60 104) D Tensor @Conv_Conv_23_325:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 104) D Process Initializer_Constant_9_338 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_8_326 ... D RKNN output shape(add): (1 60 60 104) D Tensor @Add_Add_8_326:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 104) D Process Clip_Clip_4_318 ... D RKNN output shape(relun): (1 60 60 104) D Tensor @Clip_Clip_4_318:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 104) D Process Mul_Mul_4_309 ... D RKNN output shape(multiply): (1 60 60 104) D Tensor @Mul_Mul_4_309:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 104) D Process Initializer_Constant_8_310 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_4_302 ... D RKNN output shape(Divide): (1 60 60 104) D Tensor @Div_Div_4_302:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 104) D Process Conv_Conv_24_296 ... D RKNN output shape(convolution): (1 60 60 40) D Tensor @Conv_Conv_24_296:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 40) D Process Add_Add_9_279 ... D RKNN output shape(add): (1 60 60 40) D Tensor @Add_Add_9_279:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 40) D Process Conv_Conv_25_343 ... D RKNN output shape(convolution): (1 60 60 96) D Tensor @Conv_Conv_25_343:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Initializer_Constant_11_345 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_10_340 ... D RKNN output shape(add): (1 60 60 96) D Tensor @Add_Add_10_340:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Clip_Clip_5_331 ... D RKNN output shape(relun): (1 60 60 96) D Tensor @Clip_Clip_5_331:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Mul_Mul_5_330 ... D RKNN output shape(multiply): (1 60 60 96) D Tensor @Mul_Mul_5_330:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Initializer_Constant_10_328 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_5_327 ... D RKNN output shape(Divide): (1 60 60 96) D Tensor @Div_Div_5_327:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Conv_Conv_26_319 ... D RKNN output shape(convolution): (1 60 60 96) D Tensor @Conv_Conv_26_319:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Initializer_Constant_13_329 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_11_320 ... D RKNN output shape(add): (1 60 60 96) D Tensor @Add_Add_11_320:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Clip_Clip_6_312 ... D RKNN output shape(relun): (1 60 60 96) D Tensor @Clip_Clip_6_312:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Mul_Mul_6_303 ... D RKNN output shape(multiply): (1 60 60 96) D Tensor @Mul_Mul_6_303:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Initializer_Constant_12_304 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_6_297 ... D RKNN output shape(Divide): (1 60 60 96) D Tensor @Div_Div_6_297:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Conv_Conv_27_289 ... D RKNN output shape(convolution): (1 60 60 40) D Tensor @Conv_Conv_27_289:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 40) D Process Add_Add_12_266 ... D RKNN output shape(add): (1 60 60 40) D Tensor @Add_Add_12_266:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 40) D Process Conv_Conv_28_321 ... D RKNN output shape(convolution): (1 60 60 96) D Tensor @Conv_Conv_28_321:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Initializer_Constant_15_332 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_13_322 ... D RKNN output shape(add): (1 60 60 96) D Tensor @Add_Add_13_322:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Clip_Clip_7_314 ... D RKNN output shape(relun): (1 60 60 96) D Tensor @Clip_Clip_7_314:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Mul_Mul_7_305 ... D RKNN output shape(multiply): (1 60 60 96) D Tensor @Mul_Mul_7_305:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Initializer_Constant_14_306 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_7_299 ... D RKNN output shape(Divide): (1 60 60 96) D Tensor @Div_Div_7_299:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Conv_Conv_29_298 ... D RKNN output shape(convolution): (1 60 60 96) D Tensor @Conv_Conv_29_298:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Initializer_Constant_17_300 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_14_294 ... D RKNN output shape(add): (1 60 60 96) D Tensor @Add_Add_14_294:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Clip_Clip_8_291 ... D RKNN output shape(relun): (1 60 60 96) D Tensor @Clip_Clip_8_291:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Mul_Mul_8_281 ... D RKNN output shape(multiply): (1 60 60 96) D Tensor @Mul_Mul_8_281:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Initializer_Constant_16_269 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_8_268 ... D RKNN output shape(Divide): (1 60 60 96) D Tensor @Div_Div_8_268:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 96) D Process Conv_Conv_30_267 ... D RKNN output shape(convolution): (1 60 60 40) D Tensor @Conv_Conv_30_267:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 40) D Process Add_Add_15_257 ... D RKNN output shape(add): (1 60 60 40) D Tensor @Add_Add_15_257:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 40) D Process Conv_Conv_31_241 ... D RKNN output shape(convolution): (1 60 60 240) D Tensor @Conv_Conv_31_241:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 240) D Process Initializer_Constant_19_244 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_16_242 ... D RKNN output shape(add): (1 60 60 240) D Tensor @Add_Add_16_242:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 240) D Process Clip_Clip_9_228 ... D RKNN output shape(relun): (1 60 60 240) D Tensor @Clip_Clip_9_228:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 240) D Process Mul_Mul_9_213 ... D RKNN output shape(multiply): (1 60 60 240) D Tensor @Mul_Mul_9_213:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 240) D Process Initializer_Constant_18_198 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_9_197 ... D RKNN output shape(Divide): (1 60 60 240) D Tensor @Div_Div_9_197:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 240) D Process Conv_Conv_32_182 ... D RKNN output shape(convolution): (1 60 60 240) D Tensor @Conv_Conv_32_182:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 240) D Process Initializer_Constant_21_199 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_17_183 ... D RKNN output shape(add): (1 60 60 240) D Tensor @Add_Add_17_183:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 240) D Process Clip_Clip_10_164 ... D RKNN output shape(relun): (1 60 60 240) D Tensor @Clip_Clip_10_164:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 240) D Process Mul_Mul_10_142 ... D RKNN output shape(multiply): (1 60 60 240) D Tensor @Mul_Mul_10_142:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 240) D Process Initializer_Constant_20_143 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_10_122 ... D RKNN output shape(Divide): (1 60 60 240) D Tensor @Div_Div_10_122:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 240) D Process Conv_Conv_33_100 ... D RKNN output shape(convolution): (1 60 60 56) D Tensor @Conv_Conv_33_100:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 56) D Process Conv_Conv_34_243 ... D RKNN output shape(convolution): (1 60 60 336) D Tensor @Conv_Conv_34_243:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Initializer_Constant_23_245 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_18_232 ... D RKNN output shape(add): (1 60 60 336) D Tensor @Add_Add_18_232:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Clip_Clip_11_230 ... D RKNN output shape(relun): (1 60 60 336) D Tensor @Clip_Clip_11_230:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Mul_Mul_11_214 ... D RKNN output shape(multiply): (1 60 60 336) D Tensor @Mul_Mul_11_214:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Initializer_Constant_22_201 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_11_200 ... D RKNN output shape(Divide): (1 60 60 336) D Tensor @Div_Div_11_200:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Conv_Conv_35_184 ... D RKNN output shape(convolution): (1 60 60 336) D Tensor @Conv_Conv_35_184:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Initializer_Constant_25_186 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_19_185 ... D RKNN output shape(add): (1 60 60 336) D Tensor @Add_Add_19_185:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Clip_Clip_12_166 ... D RKNN output shape(relun): (1 60 60 336) D Tensor @Clip_Clip_12_166:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Mul_Mul_12_144 ... D RKNN output shape(multiply): (1 60 60 336) D Tensor @Mul_Mul_12_144:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Initializer_Constant_24_145 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_12_123 ... D RKNN output shape(Divide): (1 60 60 336) D Tensor @Div_Div_12_123:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Conv_Conv_36_101 ... D RKNN output shape(convolution): (1 60 60 56) D Tensor @Conv_Conv_36_101:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 56) D Process Add_Add_20_65 ... D RKNN output shape(add): (1 60 60 56) D Tensor @Add_Add_20_65:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 56) D Process Conv_Conv_37_283 ... D RKNN output shape(convolution): (1 60 60 336) D Tensor @Conv_Conv_37_283:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Initializer_Constant_27_286 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_21_284 ... D RKNN output shape(add): (1 60 60 336) D Tensor @Add_Add_21_284:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Clip_Clip_13_273 ... D RKNN output shape(relun): (1 60 60 336) D Tensor @Clip_Clip_13_273:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Mul_Mul_13_262 ... D RKNN output shape(multiply): (1 60 60 336) D Tensor @Mul_Mul_13_262:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Initializer_Constant_26_250 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_13_249 ... D RKNN output shape(Divide): (1 60 60 336) D Tensor @Div_Div_13_249:out0 type: asymmetric_affine D Real output shape: (1, 60, 60, 336) D Process Conv_Conv_38_236 ... D RKNN output shape(convolution): (1 30 30 336) D Tensor @Conv_Conv_38_236:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 336) D Process Initializer_Constant_29_251 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_22_237 ... D RKNN output shape(add): (1 30 30 336) D Tensor @Add_Add_22_237:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 336) D Process Clip_Clip_14_221 ... D RKNN output shape(relun): (1 30 30 336) D Tensor @Clip_Clip_14_221:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 336) D Process Mul_Mul_14_207 ... D RKNN output shape(multiply): (1 30 30 336) D Tensor @Mul_Mul_14_207:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 336) D Process Initializer_Constant_28_208 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_14_193 ... D RKNN output shape(Divide): (1 30 30 336) D Tensor @Div_Div_14_193:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 336) D Process Conv_Conv_39_179 ... D RKNN output shape(convolution): (1 30 30 80) D Tensor @Conv_Conv_39_179:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 80) D Process Conv_Conv_40_285 ... D RKNN output shape(convolution): (1 30 30 480) D Tensor @Conv_Conv_40_285:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Initializer_Constant_31_277 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_23_276 ... D RKNN output shape(add): (1 30 30 480) D Tensor @Add_Add_23_276:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Clip_Clip_15_275 ... D RKNN output shape(relun): (1 30 30 480) D Tensor @Clip_Clip_15_275:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Mul_Mul_15_263 ... D RKNN output shape(multiply): (1 30 30 480) D Tensor @Mul_Mul_15_263:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Initializer_Constant_30_253 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_15_252 ... D RKNN output shape(Divide): (1 30 30 480) D Tensor @Div_Div_15_252:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Conv_Conv_41_238 ... D RKNN output shape(convolution): (1 30 30 480) D Tensor @Conv_Conv_41_238:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Initializer_Constant_33_254 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_24_239 ... D RKNN output shape(add): (1 30 30 480) D Tensor @Add_Add_24_239:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Clip_Clip_16_223 ... D RKNN output shape(relun): (1 30 30 480) D Tensor @Clip_Clip_16_223:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Mul_Mul_16_209 ... D RKNN output shape(multiply): (1 30 30 480) D Tensor @Mul_Mul_16_209:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Initializer_Constant_32_210 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_16_194 ... D RKNN output shape(Divide): (1 30 30 480) D Tensor @Div_Div_16_194:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Conv_Conv_42_180 ... D RKNN output shape(convolution): (1 30 30 80) D Tensor @Conv_Conv_42_180:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 80) D Process Add_Add_25_137 ... D RKNN output shape(add): (1 30 30 80) D Tensor @Add_Add_25_137:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 80) D Process Conv_Conv_43_264 ... D RKNN output shape(convolution): (1 30 30 480) D Tensor @Conv_Conv_43_264:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Initializer_Constant_35_278 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_26_265 ... D RKNN output shape(add): (1 30 30 480) D Tensor @Add_Add_26_265:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Clip_Clip_17_256 ... D RKNN output shape(relun): (1 30 30 480) D Tensor @Clip_Clip_17_256:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Mul_Mul_17_240 ... D RKNN output shape(multiply): (1 30 30 480) D Tensor @Mul_Mul_17_240:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Initializer_Constant_34_225 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_17_224 ... D RKNN output shape(Divide): (1 30 30 480) D Tensor @Div_Div_17_224:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Conv_Conv_44_211 ... D RKNN output shape(convolution): (1 30 30 480) D Tensor @Conv_Conv_44_211:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Initializer_Constant_37_226 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_27_212 ... D RKNN output shape(add): (1 30 30 480) D Tensor @Add_Add_27_212:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Clip_Clip_18_196 ... D RKNN output shape(relun): (1 30 30 480) D Tensor @Clip_Clip_18_196:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Mul_Mul_18_181 ... D RKNN output shape(multiply): (1 30 30 480) D Tensor @Mul_Mul_18_181:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Initializer_Constant_36_162 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_18_160 ... D RKNN output shape(Divide): (1 30 30 480) D Tensor @Div_Div_18_160:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Conv_Conv_45_138 ... D RKNN output shape(convolution): (1 30 30 80) D Tensor @Conv_Conv_45_138:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 80) D Process Add_Add_28_116 ... D RKNN output shape(add): (1 30 30 80) D Tensor @Add_Add_28_116:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 80) D Process Conv_Conv_46_96 ... D RKNN output shape(convolution): (1 30 30 480) D Tensor @Conv_Conv_46_96:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Initializer_Constant_39_118 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Add_Add_29_97 ... D RKNN output shape(add): (1 30 30 480) D Tensor @Add_Add_29_97:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Clip_Clip_19_79 ... D RKNN output shape(relun): (1 30 30 480) D Tensor @Clip_Clip_19_79:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Mul_Mul_19_59 ... D RKNN output shape(multiply): (1 30 30 480) D Tensor @Mul_Mul_19_59:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Initializer_Constant_38_60 ... D RKNN output shape(variable): (1 1 1 1) D Real output shape: (1, 1, 1, 1) D Process Div_Div_19_45 ... D RKNN output shape(Divide): (1 30 30 480) D Tensor @Div_Div_19_45:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 480) D Process Conv_Conv_47_44 ... D RKNN output shape(convolution): (1 30 30 96) D Tensor @Conv_Conv_47_44:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 96) D Process GlobalAveragePool_GlobalAveragePool_0_161 ... D RKNN output shape(pooling): (1 1 1 96) D Tensor @GlobalAveragePool_GlobalAveragePool_0_161:out0 type: asymmetric_affine W:tensorflow:From /data/pengshan/miniconda3/envs/board/lib/python3.6/site-packages/rknn/api/rknn.py:278: The name tf.nn.avg_pool is deprecated. Please use tf.nn.avg_pool2d instead.
D Real output shape: (1, 1, 1, 96) D Process Conv_Conv_48_139 ... D RKNN output shape(convolution): (1 1 1 24) D Tensor @Conv_Conv_48_139:out0 type: asymmetric_affine D Real output shape: (1, 1, 1, 24) D Process Reshape_Reshape_0_140 ... D RKNN output shape(variable): (1 1 1 24) D Real output shape: (1, 1, 1, 24) D Process Add_Add_30_119 ... D RKNN output shape(add): (1 1 1 24) D Tensor @Add_Add_30_119:out0 type: asymmetric_affine D Real output shape: (1, 1, 1, 24) D Process Relu_Relu_12_98 ... D RKNN output shape(relu): (1 1 1 24) D Tensor @Relu_Relu_12_98:out0 type: asymmetric_affine D Real output shape: (1, 1, 1, 24) D Process Conv_Conv_49_80 ... D RKNN output shape(convolution): (1 1 1 96) D Tensor @Conv_Conv_49_80:out0 type: asymmetric_affine D Real output shape: (1, 1, 1, 96) D Process Reshape_Reshape_1_81 ... D RKNN output shape(variable): (1 1 1 96) D Real output shape: (1, 1, 1, 96) D Process Add_Add_31_61 ... D RKNN output shape(add): (1 1 1 96) D Tensor @Add_Add_31_61:out0 type: asymmetric_affine D Real output shape: (1, 1, 1, 96) D Process Sigmoid_sigmoid_change_1_46 ... D RKNN output shape(sigmoid): (1 1 1 96) D Tensor @Sigmoid_sigmoid_change_1_46:out0 type: asymmetric_affine D Real output shape: (1, 1, 1, 96) D Process Mul_Mul_20_33 ... D RKNN output shape(multiply): (1 30 30 96) D Tensor @Mul_Mul_20_33:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 96) D Process Add_Add_32_32 ... D RKNN output shape(add): (1 30 30 96) D Tensor @Add_Add_32_32:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 96) D Process Conv_Conv_59_23 ... D RKNN output shape(convolution): (1 30 30 24) D Tensor @Conv_Conv_59_23:out0 type: asymmetric_affine D Real output shape: (1, 30, 30, 24) D Process GlobalAveragePool_GlobalAveragePool_4_141 ... D RKNN output shape(pooling): (1 -65 1 24) D Tensor @GlobalAveragePool_GlobalAveragePool_4_141:out0 type: asymmetric_affine E Catch exception when building RKNN model! E Traceback (most recent call last): E File "/data/pengshan/miniconda3/envs/board/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 1864, in _create_c_op E c_op = c_api.TF_FinishOperation(op_desc) E tensorflow.python.framework.errors_impl.InvalidArgumentError: Negative dimension size caused by subtracting 96 from 30 for 'GlobalAveragePool_GlobalAveragePool_4_141/AvgPool' (op: 'AvgPool') with input shapes: [1,30,30,24]. E During handling of the above exception, another exception occurred: E Traceback (most recent call last): E File "/data/pengshan/miniconda3/envs/board/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py", line 788, in _apply_op_helper E op_def=op_def) E File "/data/pengshan/miniconda3/envs/board/lib/python3.6/site-packages/tensorflow/python/util/deprecation.py", line 507, in new_func E return func(*args, **kwargs) E File "/data/pengshan/miniconda3/envs/board/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 3616, in create_op E op_def=op_def) E File "/data/pengshan/miniconda3/envs/board/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 2027, in init E control_input_ops) E File "/data/pengshan/miniconda3/envs/board/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 1867, in _create_c_op E raise ValueError(str(e)) E ValueError: Negative dimension size caused by subtracting 96 from 30 for 'GlobalAveragePool_GlobalAveragePool_4_141/AvgPool' (op: 'AvgPool') with input shapes: [1,30,30,24]. E Please feedback the detailed log file <log_feedback_to_the_rknn_toolkit_dev_team.log> to the RKNN Toolkit development team. E You can also check github issues: https://github.com/rockchip-linux/rknn-toolkit/issues
转换过程中出现这种问题，修改了不同的维度均报同样的错误。
