Hello everybody,
Wandering if somebody managed to properly compile the Depth-Anything V2 for the Hailo-8 and if there are some data & benchmarks (e.g. fps for running the small, baseline and large models) for running it on the hardware. I’m new to this platform and trying to compile it for a few days now, but no luck so far (I’ve managed to make the conversion from ONNX, optimize it, but with many errors, but I’m stuck at compiling it).
I’m asking, since I want to see if it’s worth investing more time to make it work, or just search for some other hardware.
Best regards
Hey @Michael_Dragos ,
Welcome to the Hailo Community!
The size of this model is not the issue ( small and maybe basee but not large ) i assume because we have larger models by Parameters and FLOPS but it has a lot of transformer (VIT-B block) , you might need to tweak it to compile , can you provide the errors you are getting to help you with the compilation!
Hello, those are the errors I’m getting during the compiler phase. I’m running the script on x86 Ubuntu machine.
[error] Mapping Failed (allocation time: 1m 59s)
No successful assignments: format_conversion1_defuse_reshape_hxf_to_w_transposed errors:
Agent infeasible
format_conversion1_defuse_width_feature_reshape errors:
Agent infeasible
format_conversion1_defuse_width_feature_reshape_concat1_transpose errors:
Agent infeasible
concat1 errors:
Agent infeasible
concat1_auto_spatial_reshape_from_concat1_to_ew_add1_defuse_width_feature_reshape_transpose errors:
Agent infeasible
auto_spatial_reshape_from_concat1_to_ew_add1_defuse_width_feature_reshape errors:
Agent infeasible
auto_spatial_reshape_from_concat1_to_ew_add1_defuse_reshape_f_to_hxw_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change51_to_slice4_defuse_reshape_hxf_to_w_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change51_to_slice4_defuse_width_feature_reshape errors:
Agent infeasible
slice4 errors:
Agent infeasible
slice4 failed on kernel validation: 16x4 is not supported in slice4
slice4 failed on kernel validation: More than 1 scs is not supported in current hw_arch for slice4
format_conversion5_defuse_width_feature_reshape errors:
Agent infeasible
format_conversion5_defuse_reshape_f_to_hxw_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change45_to_slice3_defuse_reshape_hxf_to_w_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change45_to_slice3_defuse_width_feature_reshape errors:
Agent infeasible
slice3 errors:
Agent infeasible
slice3 failed on kernel validation: 16x4 is not supported in slice3
slice3 failed on kernel validation: More than 1 scs is not supported in current hw_arch for slice3
format_conversion4_defuse_width_feature_reshape errors:
Agent infeasible
format_conversion4_defuse_reshape_f_to_hxw_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change38_to_slice2_defuse_reshape_hxf_to_w_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change38_to_slice2_defuse_width_feature_reshape errors:
Agent infeasible
slice2 errors:
Agent infeasible
slice2 failed on kernel validation: 16x4 is not supported in slice2
slice2 failed on kernel validation: More than 1 scs is not supported in current hw_arch for slice2
format_conversion3_defuse_width_feature_reshape errors:
Agent infeasible
format_conversion3_defuse_reshape_f_to_hxw_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change31_to_slice1_defuse_reshape_hxf_to_w_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change31_to_slice1_defuse_width_feature_reshape errors:
Agent infeasible
slice1 errors:
Agent infeasible
slice1 failed on kernel validation: 16x4 is not supported in slice1
slice1 failed on kernel validation: More than 1 scs is not supported in current hw_arch for slice1
format_conversion2_defuse_width_feature_reshape errors:
Agent infeasible
format_conversion2_defuse_reshape_f_to_hxw_transposed errors:
Agent infeasible
[error] Failed to produce compiled graph
[error] BackendAllocatorException: Compilation failed: No successful assignments: format_conversion1_defuse_reshape_hxf_to_w_transposed errors:
Agent infeasible
format_conversion1_defuse_width_feature_reshape errors:
Agent infeasible
format_conversion1_defuse_width_feature_reshape_concat1_transpose errors:
Agent infeasible
concat1 errors:
Agent infeasible
concat1_auto_spatial_reshape_from_concat1_to_ew_add1_defuse_width_feature_reshape_transpose errors:
Agent infeasible
auto_spatial_reshape_from_concat1_to_ew_add1_defuse_width_feature_reshape errors:
Agent infeasible
auto_spatial_reshape_from_concat1_to_ew_add1_defuse_reshape_f_to_hxw_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change51_to_slice4_defuse_reshape_hxf_to_w_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change51_to_slice4_defuse_width_feature_reshape errors:
Agent infeasible
slice4 errors:
Agent infeasible
slice4 failed on kernel validation: 16x4 is not supported in slice4
slice4 failed on kernel validation: More than 1 scs is not supported in current hw_arch for slice4
format_conversion5_defuse_width_feature_reshape errors:
Agent infeasible
format_conversion5_defuse_reshape_f_to_hxw_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change45_to_slice3_defuse_reshape_hxf_to_w_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change45_to_slice3_defuse_width_feature_reshape errors:
Agent infeasible
slice3 errors:
Agent infeasible
slice3 failed on kernel validation: 16x4 is not supported in slice3
slice3 failed on kernel validation: More than 1 scs is not supported in current hw_arch for slice3
format_conversion4_defuse_width_feature_reshape errors:
Agent infeasible
format_conversion4_defuse_reshape_f_to_hxw_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change38_to_slice2_defuse_reshape_hxf_to_w_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change38_to_slice2_defuse_width_feature_reshape errors:
Agent infeasible
slice2 errors:
Agent infeasible
slice2 failed on kernel validation: 16x4 is not supported in slice2
slice2 failed on kernel validation: More than 1 scs is not supported in current hw_arch for slice2
format_conversion3_defuse_width_feature_reshape errors:
Agent infeasible
format_conversion3_defuse_reshape_f_to_hxw_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change31_to_slice1_defuse_reshape_hxf_to_w_transposed errors:
Agent infeasible
auto_spatial_reshape_from_precision_change31_to_slice1_defuse_width_feature_reshape errors:
Agent infeasible
slice1 errors:
Agent infeasible
slice1 failed on kernel validation: 16x4 is not supported in slice1
slice1 failed on kernel validation: More than 1 scs is not supported in current hw_arch for slice1
format_conversion2_defuse_width_feature_reshape errors:
Agent infeasible
format_conversion2_defuse_reshape_f_to_hxw_transposed errors:
Agent infeasible
Also, during the optimization phase, I’m getting this error:
18/20 [04:59<00:19, 9.66s/entries]2025-04-16 10:46:14.441675: I tensorflow/core/common_runtime/executor.cc:1197] [/device:CPU:0] (DEBUG INFO) Executor start aborting (this does not indicate an error and you can ignore this message): INVALID_ARGUMENT: You must feed a value for placeholder tensor 'Placeholder/_0' with dtype float and shape [1,518,518,3]
Actually, I’ve manage to go over the first compilation errors, but now, I’m getting this:
[warning] Failed to add spatial reshapes
[error] Mapping Failed (allocation time: 0s)
Reshape is needed for layers: ew_add1, precision_change0, reduce_mean1_layer_normalization1, ew_sub1_layer_normalization1, square1_layer_normalization1, ew_mult1_layer_normalization1, precision_change24, conv_feature_splitter1_1, conv_feature_splitter1_2, conv_feature_splitter1_3, matmul1, reduce_max_softmax1, ew_sub_softmax1, ne_activation_ew_sub_softmax1, reduce_sum_softmax1, ew_mult_softmax1, matmul2, conv3, precision_change1, reduce_mean1_layer_normalization2, ew_sub1_layer_normalization2, square1_layer_normalization2, ew_mult1_layer_normalization2, precision_change25, conv4, conv5, precision_change2, reduce_mean1_layer_normalization3, ew_sub1_layer_normalization3, square1_layer_normalization3, ew_mult1_layer_normalization3, precision_change26, conv_feature_splitter2_1, conv_feature_splitter2_2, conv_feature_splitter2_3, matmul3, reduce_max_softmax2, ew_sub_softmax2, reduce_sum_softmax2, ew_mult_softmax2, matmul4, conv7, precision_change3, reduce_mean1_layer_normalization4, ew_sub1_layer_normalization4, square1_layer_normalization4, ew_mult1_layer_normalization4, precision_change27, conv8, conv9, precision_change4, reduce_mean1_layer_normalization5, ew_sub1_layer_normalization5, square1_layer_normalization5, ew_mult1_layer_normalization5, precision_change28, conv_feature_splitter3_1, conv_feature_splitter3_2, conv_feature_splitter3_3, matmul5, reduce_max_softmax3, ew_sub_softmax3, reduce_sum_softmax3, ew_mult_softmax3, matmul6, conv11, precision_change5, reduce_mean1_layer_normalization6, ew_sub1_layer_normalization6, square1_layer_normalization6, ew_mult1_layer_normalization6, precision_change29, conv12, conv13, precision_change6, reduce_mean1_layer_normalization7, ew_sub1_layer_normalization7, square1_layer_normalization7, ew_mult1_layer_normalization7, precision_change30, conv_feature_splitter4_1, conv_feature_splitter4_2, conv_feature_splitter4_3, matmul7, reduce_max_softmax4, ew_sub_softmax4, ne_activation_ew_sub_softmax4, reduce_sum_softmax4, ew_mult_softmax4, matmul8, conv19, precision_change7, reduce_mean1_layer_normalization8, ew_sub1_layer_normalization8, square1_layer_normalization8, ew_mult1_layer_normalization8, precision_change31, conv20, conv21, precision_change8, reduce_mean1_layer_normalization9, ew_sub1_layer_normalization9, square1_layer_normalization9, ew_mult1_layer_normalization9, precision_change32, conv_feature_splitter5_1, conv_feature_splitter5_2, conv_feature_splitter5_3, matmul9, reduce_max_softmax5, ew_sub_softmax5, reduce_sum_softmax5, ew_mult_softmax5, matmul10, conv23, precision_change9, reduce_mean1_layer_normalization10, ew_sub1_layer_normalization10, square1_layer_normalization10, ew_mult1_layer_normalization10, precision_change33, conv24, conv25, precision_change10, reduce_mean1_layer_normalization11, ew_sub1_layer_normalization11, square1_layer_normalization11, ew_mult1_layer_normalization11, precision_change34, conv_feature_splitter6_1, conv_feature_splitter6_2, conv_feature_splitter6_3, matmul11, reduce_max_softmax6, ew_sub_softmax6, reduce_sum_softmax6, ew_mult_softmax6, matmul12, conv27, precision_change11, reduce_mean1_layer_normalization12, ew_sub1_layer_normalization12, square1_layer_normalization12, ew_mult1_layer_normalization12, precision_change35, conv28, conv29, precision_change12, reduce_mean1_layer_normalization13, ew_sub1_layer_normalization13, square1_layer_normalization13, ew_mult1_layer_normalization13, precision_change36, conv_feature_splitter7_1, conv_feature_splitter7_2, conv_feature_splitter7_3, matmul13, reduce_max_softmax7, ew_sub_softmax7, reduce_sum_softmax7, ew_mult_softmax7, matmul14, conv35, precision_change13, reduce_mean1_layer_normalization14, ew_sub1_layer_normalization14, square1_layer_normalization14, ew_mult1_layer_normalization14, precision_change37, conv36, conv37, precision_change14, reduce_mean1_layer_normalization15, ew_sub1_layer_normalization15, square1_layer_normalization15, ew_mult1_layer_normalization15, precision_change38, conv_feature_splitter8_1, conv_feature_splitter8_2, conv_feature_splitter8_3, matmul15, reduce_max_softmax8, ew_sub_softmax8, reduce_sum_softmax8, ew_mult_softmax8, matmul16, conv39, precision_change15, reduce_mean1_layer_normalization16, ew_sub1_layer_normalization16, square1_layer_normalization16, ew_mult1_layer_normalization16, precision_change39, conv40, conv41, precision_change16, reduce_mean1_layer_normalization17, ew_sub1_layer_normalization17, square1_layer_normalization17, ew_mult1_layer_normalization17, precision_change40, conv_feature_splitter9_1, conv_feature_splitter9_2, conv_feature_splitter9_3, matmul17, reduce_max_softmax9, ew_sub_softmax9, reduce_sum_softmax9, ew_mult_softmax9, matmul18, conv43, precision_change17, reduce_mean1_layer_normalization18, ew_sub1_layer_normalization18, square1_layer_normalization18, ew_mult1_layer_normalization18, precision_change41, conv44, conv45, precision_change18, reduce_mean1_layer_normalization19, ew_sub1_layer_normalization19, square1_layer_normalization19, ew_mult1_layer_normalization19, precision_change42, conv_feature_splitter10_1, conv_feature_splitter10_2, conv_feature_splitter10_3, matmul19, reduce_max_softmax10, ew_sub_softmax10, reduce_sum_softmax10, ew_mult_softmax10, matmul20, conv51, precision_change19, reduce_mean1_layer_normalization20, ew_sub1_layer_normalization20, square1_layer_normalization20, ew_mult1_layer_normalization20, precision_change43, conv52, conv53, precision_change20, reduce_mean1_layer_normalization21, ew_sub1_layer_normalization21, square1_layer_normalization21, ew_mult1_layer_normalization21, precision_change44, conv_feature_splitter11_1, conv_feature_splitter11_2, conv_feature_splitter11_3, matmul21, reduce_max_softmax11, ew_sub_softmax11, reduce_sum_softmax11, ew_mult_softmax11, matmul22, conv55, precision_change21, reduce_mean1_layer_normalization22, ew_sub1_layer_normalization22, square1_layer_normalization22, ew_mult1_layer_normalization22, precision_change45, conv56, conv57, precision_change22, reduce_mean1_layer_normalization23, ew_sub1_layer_normalization23, square1_layer_normalization23, ew_mult1_layer_normalization23, precision_change46, conv_feature_splitter12_1, conv_feature_splitter12_2, conv_feature_splitter12_3, matmul23, reduce_max_softmax12, ew_sub_softmax12, reduce_sum_softmax12, ew_mult_softmax12, matmul24, conv59, precision_change23, reduce_mean1_layer_normalization24, ew_sub1_layer_normalization24, square1_layer_normalization24, ew_mult1_layer_normalization24, precision_change47, conv60, conv61, reduce_mean1_layer_normalization25, ew_sub1_layer_normalization25, square1_layer_normalization25, ew_mult1_layer_normalization25, precision_change48, but adding a reshape has failed.
[error] Failed to produce compiled graph
[error] BackendAllocatorException: Compilation failed: Reshape is needed for layers: ew_add1, precision_change0, reduce_mean1_layer_normalization1, ew_sub1_layer_normalization1, square1_layer_normalization1, ew_mult1_layer_normalization1, precision_change24, conv_feature_splitter1_1, conv_feature_splitter1_2, conv_feature_splitter1_3, matmul1, reduce_max_softmax1, ew_sub_softmax1, ne_activation_ew_sub_softmax1, reduce_sum_softmax1, ew_mult_softmax1, matmul2, conv3, precision_change1, reduce_mean1_layer_normalization2, ew_sub1_layer_normalization2, square1_layer_normalization2, ew_mult1_layer_normalization2, precision_change25, conv4, conv5, precision_change2, reduce_mean1_layer_normalization3, ew_sub1_layer_normalization3, square1_layer_normalization3, ew_mult1_layer_normalization3, precision_change26, conv_feature_splitter2_1, conv_feature_splitter2_2, conv_feature_splitter2_3, matmul3, reduce_max_softmax2, ew_sub_softmax2, reduce_sum_softmax2, ew_mult_softmax2, matmul4, conv7, precision_change3, reduce_mean1_layer_normalization4, ew_sub1_layer_normalization4, square1_layer_normalization4, ew_mult1_layer_normalization4, precision_change27, conv8, conv9, precision_change4, reduce_mean1_layer_normalization5, ew_sub1_layer_normalization5, square1_layer_normalization5, ew_mult1_layer_normalization5, precision_change28, conv_feature_splitter3_1, conv_feature_splitter3_2, conv_feature_splitter3_3, matmul5, reduce_max_softmax3, ew_sub_softmax3, reduce_sum_softmax3, ew_mult_softmax3, matmul6, conv11, precision_change5, reduce_mean1_layer_normalization6, ew_sub1_layer_normalization6, square1_layer_normalization6, ew_mult1_layer_normalization6, precision_change29, conv12, conv13, precision_change6, reduce_mean1_layer_normalization7, ew_sub1_layer_normalization7, square1_layer_normalization7, ew_mult1_layer_normalization7, precision_change30, conv_feature_splitter4_1, conv_feature_splitter4_2, conv_feature_splitter4_3, matmul7, reduce_max_softmax4, ew_sub_softmax4, ne_activation_ew_sub_softmax4, reduce_sum_softmax4, ew_mult_softmax4, matmul8, conv19, precision_change7, reduce_mean1_layer_normalization8, ew_sub1_layer_normalization8, square1_layer_normalization8, ew_mult1_layer_normalization8, precision_change31, conv20, conv21, precision_change8, reduce_mean1_layer_normalization9, ew_sub1_layer_normalization9, square1_layer_normalization9, ew_mult1_layer_normalization9, precision_change32, conv_feature_splitter5_1, conv_feature_splitter5_2, conv_feature_splitter5_3, matmul9, reduce_max_softmax5, ew_sub_softmax5, reduce_sum_softmax5, ew_mult_softmax5, matmul10, conv23, precision_change9, reduce_mean1_layer_normalization10, ew_sub1_layer_normalization10, square1_layer_normalization10, ew_mult1_layer_normalization10, precision_change33, conv24, conv25, precision_change10, reduce_mean1_layer_normalization11, ew_sub1_layer_normalization11, square1_layer_normalization11, ew_mult1_layer_normalization11, precision_change34, conv_feature_splitter6_1, conv_feature_splitter6_2, conv_feature_splitter6_3, matmul11, reduce_max_softmax6, ew_sub_softmax6, reduce_sum_softmax6, ew_mult_softmax6, matmul12, conv27, precision_change11, reduce_mean1_layer_normalization12, ew_sub1_layer_normalization12, square1_layer_normalization12, ew_mult1_layer_normalization12, precision_change35, conv28, conv29, precision_change12, reduce_mean1_layer_normalization13, ew_sub1_layer_normalization13, square1_layer_normalization13, ew_mult1_layer_normalization13, precision_change36, conv_feature_splitter7_1, conv_feature_splitter7_2, conv_feature_splitter7_3, matmul13, reduce_max_softmax7, ew_sub_softmax7, reduce_sum_softmax7, ew_mult_softmax7, matmul14, conv35, precision_change13, reduce_mean1_layer_normalization14, ew_sub1_layer_normalization14, square1_layer_normalization14, ew_mult1_layer_normalization14, precision_change37, conv36, conv37, precision_change14, reduce_mean1_layer_normalization15, ew_sub1_layer_normalization15, square1_layer_normalization15, ew_mult1_layer_normalization15, precision_change38, conv_feature_splitter8_1, conv_feature_splitter8_2, conv_feature_splitter8_3, matmul15, reduce_max_softmax8, ew_sub_softmax8, reduce_sum_softmax8, ew_mult_softmax8, matmul16, conv39, precision_change15, reduce_mean1_layer_normalization16, ew_sub1_layer_normalization16, square1_layer_normalization16, ew_mult1_layer_normalization16, precision_change39, conv40, conv41, precision_change16, reduce_mean1_layer_normalization17, ew_sub1_layer_normalization17, square1_layer_normalization17, ew_mult1_layer_normalization17, precision_change40, conv_feature_splitter9_1, conv_feature_splitter9_2, conv_feature_splitter9_3, matmul17, reduce_max_softmax9, ew_sub_softmax9, reduce_sum_softmax9, ew_mult_softmax9, matmul18, conv43, precision_change17, reduce_mean1_layer_normalization18, ew_sub1_layer_normalization18, square1_layer_normalization18, ew_mult1_layer_normalization18, precision_change41, conv44, conv45, precision_change18, reduce_mean1_layer_normalization19, ew_sub1_layer_normalization19, square1_layer_normalization19, ew_mult1_layer_normalization19, precision_change42, conv_feature_splitter10_1, conv_feature_splitter10_2, conv_feature_splitter10_3, matmul19, reduce_max_softmax10, ew_sub_softmax10, reduce_sum_softmax10, ew_mult_softmax10, matmul20, conv51, precision_change19, reduce_mean1_layer_normalization20, ew_sub1_layer_normalization20, square1_layer_normalization20, ew_mult1_layer_normalization20, precision_change43, conv52, conv53, precision_change20, reduce_mean1_layer_normalization21, ew_sub1_layer_normalization21, square1_layer_normalization21, ew_mult1_layer_normalization21, precision_change44, conv_feature_splitter11_1, conv_feature_splitter11_2, conv_feature_splitter11_3, matmul21, reduce_max_softmax11, ew_sub_softmax11, reduce_sum_softmax11, ew_mult_softmax11, matmul22, conv55, precision_change21, reduce_mean1_layer_normalization22, ew_sub1_layer_normalization22, square1_layer_normalization22, ew_mult1_layer_normalization22, precision_change45, conv56, conv57, precision_change22, reduce_mean1_layer_normalization23, ew_sub1_layer_normalization23, square1_layer_normalization23, ew_mult1_layer_normalization23, precision_change46, conv_feature_splitter12_1, conv_feature_splitter12_2, conv_feature_splitter12_3, matmul23, reduce_max_softmax12, ew_sub_softmax12, reduce_sum_softmax12, ew_mult_softmax12, matmul24, conv59, precision_change23, reduce_mean1_layer_normalization24, ew_sub1_layer_normalization24, square1_layer_normalization24, ew_mult1_layer_normalization24, precision_change47, conv60, conv61, reduce_mean1_layer_normalization25, ew_sub1_layer_normalization25, square1_layer_normalization25, ew_mult1_layer_normalization25, precision_change48, but adding a reshape has failed.
Hey @Michael_Dragos,
This error typically happens when the DFC compiler struggles to automatically add reshape operations for certain layers. Let me explain what’s probably causing it and how you can fix it:
Common Causes
The compiler is having shape inference issues, usually because:
- It can’t figure out input/output shapes for operations like matmul, layer_norm, or conv
- Some operations require very specific shapes on Hailo hardware
- Your model contains dynamic dimensions (those problematic
Noneor[?, ?, ?, ?]from ONNX/TensorFlow)
Solutions to Try
-
Manually add reshape layers before/after the problematic layers to match what the compiler expects
-
Create a custom layer config file to explicitly define shapes:
{ "layer_config": { "matmul1": { "input_shape": [1, 64], "output_shape": [1, 64] }, "precision_change0": { "input_shape": [1, 64, 1], "output_shape": [1, 64] } } }
I’d recommend doing each step separately: parse → optimize → compile. This approach creates har files and gives you better control over each step. Try these reshape layer changes first, and if you run into another issue, just update us here.