Idon't know what to do

Ketul_Patel · July 17, 2025, 11:38pm

[info] Model Optimization Algorithm Quantization-Aware Fine-Tuning is done (completion time is 00:00:25.47)
[info] Layer Noise Analysis skipped
[info] Model Optimization is done
[info] Saved HAR to: /local/workspace/direction_classifier.har
[info] Loading model script commands to direction_classifier from /local/workspace/hailo_model_zoo/hailo_model_zoo/cfg/alls/generic/direction_classifier.alls
[info] ParsedPerformanceParam command, setting optimization_level(max=2)
[info] Loading network parameters
[info] Starting Hailo allocation and compilation flow
[error] Mapping Failed (allocation time: 0s)
Can’t find mutual format for fc1_d1 → ew_add0_ew_add_n_fc1_d0

[error] Failed to produce compiled graph
[error] BackendAllocatorException: Compilation failed: Can’t find mutual format for fc1_d1 → ew_add0_ew_add_n_fc1_d0
(hailo_virtualenv) hailo@116af785a8fc:/local/workspace$

omria · July 21, 2025, 1:10pm

Hey @Ketul_Patel,

Welcome to the Hailo Community!

That error you’re seeing:

Can't find mutual format for fc1_d1 → ew_add0_ew_add_n_fc1_d0

This indicates our Dataflow Compiler’s allocator couldn’t resolve the memory layout requirements between your fc1_d1 layer and the downstream element-wise add node. Our hardware optimally runs most operators in native NHCW format, and when connected operators have conflicting layout expectations, the compiler needs to insert explicit format conversions - when it can’t, you get this “mutual format” failure.

Our compiler’s default behavior is to only insert format conversions at input layers (through input_conversion), not between internal graph nodes. To resolve this, you’ll want to use a model-script to explicitly force a conversion/transpose immediately after fc1_d1. This ensures its output layout aligns with the other input to your element-wise add operation.

I can help you set up the model-script syntax if you need it. Are you working with a custom model architecture or a standard framework export?

Ketul_Patel · July 22, 2025, 8:49am

thank you for your genuine response, i am doing on my custom model architecture. also i will really thankful that you help me on my projerct

Ketul_Patel · July 22, 2025, 9:01am

file:///C:/Users/KetulP/Desktop/usb/direction_classifier_hailo_model.html#/model-details

this the .her file

i hop you, this will help you maybe

omria · July 23, 2025, 8:14am

Hey @Ketul_Patel,

That’s just your local file path - I can’t see what’s actually in your model from that. Could you share the .alls file contents and maybe a screenshot from Netron or something similar so I can see what we’re working with?

Ketul_Patel · July 23, 2025, 9:04am

i wouldn’t be able to send you images

new_normalization1 = normalization([127.5, 127.5, 127.5], [127.5, 127.5, 127.5])

quantization_param(fc1_d0, precision_mode=a16_w16)
quantization_param(fc1_d1, precision_mode=a16_w16)
quantization_param(fc2, precision_mode=a16_w16)

post_quantization_optimization(
finetune,
policy=enabled,
dataset_size=4665,
epochs=1,
learning_rate=0.0001,
loss_layer_names=[fc2],
loss_factors=[1.0],
loss_types=[l2]
)

model_optimization_flavor(optimization_level=1, compression_level=0)

performance_param(compiler_optimization_level=max)

Ketul_Patel · July 23, 2025, 9:04am

here is some error occurred

Ketul_Patel · July 23, 2025, 10:28am

Ketul_Patel · July 23, 2025, 10:29am

Ketul_Patel · July 23, 2025, 10:29am

Ketul_Patel · July 23, 2025, 10:29am

Ketul_Patel · July 23, 2025, 10:29am

Ketul_Patel · July 23, 2025, 10:29am

Ketul_Patel · July 23, 2025, 10:29am

Ketul_Patel · July 23, 2025, 10:29am

Ketul_Patel · July 30, 2025, 9:41am

Hello, I share you some screenshots please check it.

Ketul_Patel · August 29, 2025, 10:28am

hello sir please reply to this

omria · August 31, 2025, 8:57am

Hey @Ketul_Patel!

Looking at your screenshots, here’s the issue:

fc1_d0: Shape (1, 128) - flat/linear tensor format
fc1_d1: Shape (1, 16, 8) - 3D tensor format (likely NHCW or NHWC)

You’re trying to do an element-wise addition between these two, but they have completely different shapes and memory layouts. The Hailo compiler can’t automatically figure out how to convert between them for internal nodes (it only does this automatically at model inputs).

How To Fix

You need to reshape and transpose fc1_d1 to match fc1_d0’s flat format before the addition.

Updated Model Script

Here’s your fixed .alls script with the necessary changes:

new_normalization1 = normalization([127.5, 127.5, 127.5], [127.5, 127.5, 127.5])

# Quantization
quantization_param(fc1_d0, precision_mode=a16_w16)
quantization_param(fc1_d1, precision_mode=a16_w16)
quantization_param(fc2, precision_mode=a16_w16)

# Optimization
post_quantization_optimization(
    finetune,
    policy=enabled,
    dataset_size=4665,
    epochs=1,
    learning_rate=0.0001,
    loss_layer_names=[fc2],
    loss_factors=[1.0],
    loss_types=[l2]
)

model_optimization_flavor(optimization_level=1, compression_level=0)
performance_param(compiler_optimization_level=max)

Graph Connection Update

Important: In your model graph, make sure the element-wise Add operation connects to:

fc1_d0 (unchanged)
fc1_d1_reshaped (NOT the original fc1_d1)

So it should look like: Add(fc1_d0, fc1_d1_reshaped)

Why This Works

Transpose first: This handles any row/column memory order issues between the layouts
Reshape second: This flattens the 3D tensor (1, 16, 8) into the required (1, 128) shape
Compiler restriction: Hailo only inserts automatic format conversions at model entry points, not between internal nodes - that’s why we need to do this explicitly

Hope this helps!

Ketul_Patel · September 1, 2025, 9:03am

thank youfor your support i will check this

Ketul_Patel · September 1, 2025, 9:23am

omria:

new_normalization1 = normalization([127.5, 127.5, 127.5], [127.5, 127.5, 127.5])

# Fix the layout mismatch - force fc1_d1 to match fc1_d0's format
transpose(fc1_d1, perm=[0, 2, 1], name="fc1_d1_transposed")  # (1, 16, 8) → (1, 8, 16)
reshape(fc1_d1_transposed, shape=[1, 128], name="fc1_d1_reshaped")  # Now matches fc1_d0

# Your existing quantization config (this part was correct)
quantization_param(fc1_d0, precision_mode=a16_w16)
quantization_param(fc1_d1, precision_mode=a16_w16)
quantization_param(fc2, precision_mode=a16_w16)

# Post-quantization optimization
post_quantization_optimization(
    finetune,
    policy=enabled,
    dataset_size=4665,
    epochs=1,
    learning_rate=0.0001,
    loss_layer_names=[fc2],
    loss_factors=[1.0],
    loss_types=[l2]
)

# Optimization and performance parameters
model_optimization_flavor(optimization_level=1, compression_level=0)
performance_param(compiler_optimization_level=max)

ocal/shared_with_docker/direction_classifier.yaml --hw-arch hailo8l
Start run for network direction_classifier …
Initializing the hailo8l runner…
[info] Translation started on Tensorflow model direction_classifier
[info] Start nodes mapped from original model: ‘serving_default_rescaling_input:0’: ‘direction_classifier/input_layer1’.
[info] End nodes mapped from original model: ‘StatefulPartitionedCall:0’.
[info] Translation completed on Tensorflow model direction_classifier (completion time: 00:00:00.13)
[info] Saved HAR to: /local/workspace/direction_classifier.har
Preparing calibration data…
[info] Loading model script commands to direction_classifier from /local/shared_with_docker/direction_classifier.alls
Traceback (most recent call last):
File “/local/workspace/hailo_virtualenv/lib/python3.10/site-packages/hailo_sdk_client/sdk_backend/script_parser/model_script_parser.py”, line 381, in parse_script
script_grammar.parseString(input_script, parseAll=True)
File “/local/workspace/hailo_virtualenv/lib/python3.10/site-packages/pyparsing.py”, line 1955, in parseString
raise exc
File “/local/workspace/hailo_virtualenv/lib/python3.10/site-packages/pyparsing.py”, line 3814, in parseImpl
raise ParseException(instring, loc, self.errmsg, self)
pyparsing.ParseException: Expected end of text, found ‘#’ (at char 185), (line:4, col:56)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File “/local/workspace/hailo_virtualenv/bin/hailomz”, line 33, in
sys.exit(load_entry_point(‘hailo-model-zoo’, ‘console_scripts’, ‘hailomz’)())
File “/local/workspace/hailo_model_zoo/hailo_model_zoo/main.py”, line 122, in main
run(args)
File “/local/workspace/hailo_model_zoo/hailo_model_zoo/main.py”, line 111, in run
return handlersargs.command
File “/local/workspace/hailo_model_zoo/hailo_model_zoo/main_driver.py”, line 248, in compile
_ensure_optimized(runner, logger, args, network_info)
File “/local/workspace/hailo_model_zoo/hailo_model_zoo/main_driver.py”, line 91, in _ensure_optimized
optimize_model(
File “/local/workspace/hailo_model_zoo/hailo_model_zoo/core/main_utils.py”, line 351, in optimize_model
optimize_full_precision_model(runner, calib_feed_callback, logger, model_script, resize, input_conversion, classes)
File “/local/workspace/hailo_model_zoo/hailo_model_zoo/core/main_utils.py”, line 315, in optimize_full_precision_model
runner.load_model_script(model_script)
File “/local/workspace/hailo_virtualenv/lib/python3.10/site-packages/hailo_sdk_common/states/states.py”, line 16, in wrapped_func
return func(self, *args, **kwargs)
File “/local/workspace/hailo_virtualenv/lib/python3.10/site-packages/hailo_sdk_client/runner/client_runner.py”, line 497, in load_model_script
self._sdk_backend.load_model_script_from_file(model_script, append)
File “/local/workspace/hailo_virtualenv/lib/python3.10/site-packages/hailo_sdk_client/sdk_backend/sdk_backend.py”, line 494, in load_model_script_from_file
self._script_parser.parse_script_from_file(model_script_path, nms_config, append)
File “/local/workspace/hailo_virtualenv/lib/python3.10/site-packages/hailo_sdk_client/sdk_backend/script_parser/model_script_parser.py”, line 312, in parse_script_from_file
return self.parse_script(f.read(), append, nms_config_file)
File “/local/workspace/hailo_virtualenv/lib/python3.10/site-packages/hailo_sdk_client/sdk_backend/script_parser/model_script_parser.py”, line 389, in parse_script
raise BackendScriptParserException(f"Parsing failed at:\n{e.markInputline()}")
hailo_sdk_client.sdk_backend.sdk_backend_exceptions.BackendScriptParserException: Parsing failed at:
transpose(fc1_d1,perm=[0,2,1],name=“fc1_d1_transposed”)>!<#(1,16,8)→(1,8,16)

Topic		Replies	Views
1D form is not supported in layer ... of type EWAddLayer General dfc , error	14	234	January 17, 2025
loss layer name problem General	3	26	July 22, 2025
Running hailomz optimize reports layer yolov8n/conv41 doesn't have one output layer Community Projects raspberry-pi , hailo8 , error	26	1338	October 31, 2024
Problems optimizing (quantization) dinov2 - ONNX to HEF General dfc , raspberry-pi , hailo8 , error	6	733	November 21, 2024
Optimization custom model General optimization , hailo8	1	380	January 6, 2025

Idon't know what to do

How To Fix

Updated Model Script

Graph Connection Update

Why This Works

Related topics