Alveo U50 ml-suite Unknown: exceptions.RuntimeError: Could not init FPGA: xclbin

[jaydevshelat]

Hi,

I am using Alveo U50 board. I am trying to run the ml-suite tensorflow example. I have tried command line as well as Jupyter notebook. I get the same error (show below).

My board is powered-up & initialized. One observation is the utilities (for e.g. docker_run.sh script) don't recognize my board/platform automatically.

I am fairly sure this is because the docker container does not have an overlaybin for U50.

Does anyone know if this works on U50? If yes, where to get the correct docker container?

Error received:

`
save graph at ./inception_v1_baseline-fpga.pb
[XBLAS] # kernels: 1
|
ERROR: No devices found
ERROR: Failed to find an OpenCL platform
2020-04-15 22:35:22.928304: W tensorflow/core/framework/op_kernel.cc:1261] Unknown: exceptions.RuntimeError: Could not init FPGA: xclbin /opt/ml-suite/overlaybins/(unknown)/overlay_4.xclbin lib_path /opt/ml-suite/xfdnn/rt/libs/libxfdnn.so
Traceback (most recent call last):

File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/script_ops.py", line 206, in call
ret = func(*args)

File "/opt/ml-suite/xfdnn/rt/xdnn_rt_base.py", line 134, in forward_exec
layer_outputs = layer.forward_exec( layer_inputs )

File "./xfdnn/tools/emu/fpga_pydot_layer.py", line 103, in forward_exec

File "/opt/ml-suite/xfdnn/rt/xdnn_env.py", line 83, in createHandle
raise RuntimeError("Could not init FPGA: xclbin %s lib_path %s" % (self._xdnnParams['xclbin'], self._xdnnParams['lib_path']))

RuntimeError: Could not init FPGA: xclbin /opt/ml-suite/overlaybins/(unknown)/overlay_4.xclbin lib_path /opt/ml-suite/xfdnn/rt/libs/libxfdnn.so

Traceback (most recent call last):
File "run.py", line 134, in
top5_accuracy(graph, args.input_nodes, args.output_nodes, iter_cnt, batch_size, args.label_offset)
File "/opt/ml-suite/examples/tensorflow/utils.py", line 106, in top5_accuracy
predictions = sess.run(output_tensor, feed_dict={tensor: inputs[name] for name, tensor in input_tensors.items()})
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 929, in run
run_metadata_ptr)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 1152, in _run
feed_dict_tensor, options, run_metadata)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 1328, in _do_run
run_metadata)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 1348, in _do_call
raise type(e)(node_def, op, message)
tensorflow.python.framework.errors_impl.UnknownError: exceptions.RuntimeError: Could not init FPGA: xclbin /opt/ml-suite/overlaybins/(unknown)/overlay_4.xclbin lib_path /opt/ml-suite/xfdnn/rt/libs/libxfdnn.so`

[wilderfield]

Alveo U50 is only supported in the Xilinx/Vitis-AI github repository. Vitis-AI is evolved from MLSuite and has a broader scope. It supports more boards, plus edge devices. https://github.com/Xilinx/Vitis-AI

…

On Wed, Apr 15, 2020 at 3:44 PM Jaydev ***@***.***> wrote: Hi, I am using Alveo U50 board. I am trying to run the ml-suite tensorflow example. I have tried command line as well as Jupyter notebook. I get the same error (show below). My board is powered-up & initialized. One observation is the utilities (for e.g. docker_run.sh script) don't recognize my board/platform automatically. I am fairly sure this is because the docker container does not have an overlaybin for U50. Does anyone know if this works on U50? If yes, where to get the correct docker container? Error received: `... save graph at ./inception_v1_baseline-fpga.pb [XBLAS] # kernels: 1 ERROR: No devices found ERROR: Failed to find an OpenCL platform 2020-04-15 22:35:22.928304: W tensorflow/core/framework/op_kernel.cc:1261] Unknown: exceptions.RuntimeError: Could not init FPGA: xclbin /opt/ml-suite/overlaybins/(unknown)/overlay_4.xclbin lib_path /opt/ml-suite/xfdnn/rt/libs/libxfdnn.so Traceback (most recent call last): File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/script_ops.py", line 206, in *call* ret = func(*args) File "/opt/ml-suite/xfdnn/rt/xdnn_rt_base.py", line 134, in forward_exec layer_outputs = layer.forward_exec( layer_inputs ) File "./xfdnn/tools/emu/fpga_pydot_layer.py", line 103, in forward_exec File "/opt/ml-suite/xfdnn/rt/xdnn_env.py", line 83, in createHandle raise RuntimeError("Could not init FPGA: xclbin %s lib_path %s" % (self._xdnnParams['xclbin'], self._xdnnParams['lib_path'])) RuntimeError: Could not init FPGA: xclbin /opt/ml-suite/overlaybins/(unknown)/overlay_4.xclbin lib_path /opt/ml-suite/xfdnn/rt/libs/libxfdnn.so Traceback (most recent call last): File "run.py", line 134, in top5_accuracy(graph, args.input_nodes, args.output_nodes, iter_cnt, batch_size, args.label_offset) File "/opt/ml-suite/examples/tensorflow/utils.py", line 106, in top5_accuracy predictions = sess.run(output_tensor, feed_dict={tensor: inputs[name] for name, tensor in input_tensors.items()}) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 929, in run run_metadata_ptr) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 1152, in _run feed_dict_tensor, options, run_metadata) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 1328, in _do_run run_metadata) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 1348, in _do_call raise type(e)(node_def, op, message) tensorflow.python.framework.errors_impl.UnknownError: exceptions.RuntimeError: Could not init FPGA: xclbin /opt/ml-suite/overlaybins/(unknown)/overlay_4.xclbin lib_path /opt/ml-suite/xfdnn/rt/libs/libxfdnn.so` — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#127>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADX4NJ62QLS46BGPZNNIYETRMY2FZANCNFSM4MI6Q4YQ> .

[jaydevshelat]

Thank You @wilderfield !
One Suggestion: It would be nice to have target platform in the readme. Even in Vitis-AI it looks like U50 is not well supported (compared to U200/U250/ZCU boards). Again, having this information in different readme links would be of great help to a Xilinx beginner!