undefined symbol: gzeof

[Novelfor]

aimet 1.35 call QuantizationSimModel error

File "/data/miniconda3/envs/py310/lib/python3.10/site-packages/aimet_torch/v1/quantsim.py", line 307, in __init__
    self._realize_quant_wrappers_in_model(self.model)
  File "/data/miniconda3/envs/py310/lib/python3.10/site-packages/aimet_torch/v1/quantsim.py", line 322, in _realize_quant_wrappers_in_model
    self._realize_quant_wrappers_in_model(module_ref)
  File "/data/miniconda3/envs/py310/lib/python3.10/site-packages/aimet_torch/v1/quantsim.py", line 318, in _realize_quant_wrappers_in_model
    quantized_module = module_ref.realize_v1_wrapper()
  File "/data/miniconda3/envs/py310/lib/python3.10/site-packages/aimet_torch/quantsim_config/builder.py", line 191, in realize_v1_wrapper
    quantized_module = StaticGridQuantWrapper(self._module_to_wrap, self._weight_bw, self._activation_bw,
  File "/data/miniconda3/envs/py310/lib/python3.10/site-packages/aimet_torch/v1/qc_quantize_op.py", line 701, in __init__
    super().__init__(module_to_wrap, weight_bw, activation_bw, round_mode, quant_scheme,
  File "/data/miniconda3/envs/py310/lib/python3.10/site-packages/aimet_torch/v1/qc_quantize_op.py", line 227, in __init__
    self.output_quantizers = [tensor_quantizer_factory(activation_bw, round_mode,
  File "/data/miniconda3/envs/py310/lib/python3.10/site-packages/aimet_torch/v1/qc_quantize_op.py", line 227, in <listcomp>
    self.output_quantizers = [tensor_quantizer_factory(activation_bw, round_mode,
  File "/data/miniconda3/envs/py310/lib/python3.10/site-packages/aimet_torch/v1/qc_quantize_op.py", line 98, in tensor_quantizer_factory
    tensor_quantizer = StaticGridPerTensorQuantizer(bitwidth, round_mode, quant_scheme,
  File "/data/miniconda3/envs/py310/lib/python3.10/site-packages/aimet_torch/v1/tensor_quantizer.py", line 424, in __init__
    self._cppOp = [AimetTensorQuantizer(quant_scheme)]
  File "/data/miniconda3/envs/py310/lib/python3.10/site-packages/aimet_common/aimet_tensor_quantizer.py", line 61, in __init__
    raise RuntimeError(f"Unable to initialize class AimetTensorQuantizer: {ERROR_MESSAGE}")
RuntimeError: Unable to initialize class AimetTensorQuantizer: AimetTensorQuantizer import failed with the following error:

/data/miniconda3/envs/py310/lib/python3.10/site-packages/aimet_common/_libpymo.cpython-310-x86_64-linux-gnu.so: undefined symbol: gzeof

i use nm and ldd command to _libpymo.cpython-310-x86_64-linux-gnu.so

ldd /data/miniconda3/envs/py310/lib/python3.10/site-packages/aimet_common/_libpymo.cpython-310-x86_64-linux-gnu.so
        linux-vdso.so.1 (0x00007ffce6bdc000)                                                                                                                       libpython3.10.so.1.0 => /lib/x86_64-linux-gnu/libpython3.10.so.1.0 (0x00007fd7f31a0000)
        libcublas.so.12 => /data/cuda/cuda-12.1/cuda/lib64/libcublas.so.12 (0x00007fd7ec800000)
        liblapacke.so.3 => /lib/x86_64-linux-gnu/liblapacke.so.3 (0x00007fd7ec599000)
        libstdc++.so.6 => /lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007fd7ec36d000)
        libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fd7f30b9000)
        libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007fd7f3097000)
        libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fd7ec145000)
        /lib64/ld-linux-x86-64.so.2 (0x00007fd7f5069000)
        libexpat.so.1 => /lib/x86_64-linux-gnu/libexpat.so.1 (0x00007fd7ec114000)
        libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007fd7ec0f8000)
        libcublasLt.so.12 => /data/cuda/cuda-12.1/cuda/lib64/libcublasLt.so.12 (0x00007fd7cb200000)
        librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007fd7f3092000)
        libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007fd7f308b000)
        libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007fd7f3086000)
        libblas.so.3 => /lib/x86_64-linux-gnu/libblas.so.3 (0x00007fd7ec052000)
        liblapack.so.3 => /lib/x86_64-linux-gnu/liblapack.so.3 (0x00007fd7caac7000)
        libtmglib.so.3 => /lib/x86_64-linux-gnu/libtmglib.so.3 (0x00007fd7caa5c000)
        libgfortran.so.5 => /lib/x86_64-linux-gnu/libgfortran.so.5 (0x00007fd7ca780000)                                                                            libquadmath.so.0 => /lib/x86_64-linux-gnu/libquadmath.so.0 (0x00007fd7ca738000)

 nm -D /data/miniconda3/envs/py310/lib/python3.10/site-packages/aimet_common/_libpymo.cpython-310-x86_64-linux-gnu.so | grep gzeof
                 U gzeof

I install aimet_torch by this command

pip install aimet_torch-1.35.0.cu121-cp310-cp310-manylinux_2_34_x86_64.whl  --index-url https://download.pytorch.org/whl/cu121

[1826133674]

I also met this error. By the way,

[huisai]

+1

[erickksun]

lib/python3.10/site-packages/aimet_common$ patchelf --add-needed libz.so _libpymo.cpython-310-x86_64-linux-gnu.so
It works.
enjoy.

[quic-akhobare]

Thanks for the report - could you please paste details of your environment? E.g. which ubuntu version? Which PyTorch and CUDA version are you using?

[quic-kyunggeu]

This issue was fixed by #3497, but AIMET 1.35 was released before my fix.
I'll cherry-pick my fix to 1.35 and make a quick bugfix release

[quic-kyunggeu]

Update: This bug only exists in 1.35; it didn't exist in <=1.34 and is fixed in >=2.0.0.
@Novelfor @1826133674 Could you please consider upgrading to 2.0.0?