-_0_0_0/inductor_output_code_ch44xxkifazlcpkp6mi44xhqeej2j5mbgwmesiwx6y3oajzmixxp_7.html

<pre style="background-color:#ffffff;">
<span style="font-style:italic;color:#969896;"># AOT ID: [&#39;0_inference&#39;]
</span><span style="font-weight:bold;color:#a71d5d;">from </span><span style="color:#323232;">ctypes </span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">c_void_p, c_long, c_int
</span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">torch
</span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">math
</span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">random
</span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">os
</span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">tempfile
</span><span style="font-weight:bold;color:#a71d5d;">from </span><span style="color:#323232;">math </span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">inf, nan
</span><span style="font-weight:bold;color:#a71d5d;">from </span><span style="color:#323232;">cmath </span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">nanj
</span><span style="font-weight:bold;color:#a71d5d;">from </span><span style="color:#323232;">torch._inductor.hooks </span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">run_intermediate_hooks
</span><span style="font-weight:bold;color:#a71d5d;">from </span><span style="color:#323232;">torch._inductor.utils </span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">maybe_profile
</span><span style="font-weight:bold;color:#a71d5d;">from </span><span style="color:#323232;">torch._inductor.codegen.memory_planning </span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">_align </span><span style="font-weight:bold;color:#a71d5d;">as </span><span style="color:#323232;">align
</span><span style="font-weight:bold;color:#a71d5d;">from </span><span style="color:#323232;">torch </span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">device, empty_strided
</span><span style="font-weight:bold;color:#a71d5d;">from </span><span style="color:#323232;">torch._inductor.async_compile </span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">AsyncCompile
</span><span style="font-weight:bold;color:#a71d5d;">from </span><span style="color:#323232;">torch._inductor.select_algorithm </span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">extern_kernels
</span><span style="font-weight:bold;color:#a71d5d;">from </span><span style="color:#323232;">torch._inductor.codegen.multi_kernel </span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">MultiKernelCall
</span><span style="color:#323232;">
</span><span style="color:#323232;">aten </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">torch.ops.aten
</span><span style="color:#323232;">inductor_ops </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">torch.ops.inductor
</span><span style="color:#323232;">_quantized </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">torch.ops._quantized
</span><span style="color:#323232;">assert_size_stride </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">torch.</span><span style="color:#0086b3;">_C</span><span style="color:#323232;">._dynamo.guards.assert_size_stride
</span><span style="color:#323232;">empty_strided_cpu </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">torch.</span><span style="color:#0086b3;">_C</span><span style="color:#323232;">._dynamo.guards._empty_strided_cpu
</span><span style="color:#323232;">empty_strided_cuda </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">torch.</span><span style="color:#0086b3;">_C</span><span style="color:#323232;">._dynamo.guards._empty_strided_cuda
</span><span style="color:#323232;">empty_strided_xpu </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">torch.</span><span style="color:#0086b3;">_C</span><span style="color:#323232;">._dynamo.guards._empty_strided_xpu
</span><span style="color:#323232;">reinterpret_tensor </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">torch.</span><span style="color:#0086b3;">_C</span><span style="color:#323232;">._dynamo.guards._reinterpret_tensor
</span><span style="color:#323232;">alloc_from_pool </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">torch.ops.inductor._alloc_from_pool
</span><span style="color:#323232;">async_compile </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">AsyncCompile()
</span><span style="color:#323232;">empty_strided_p2p </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">torch.</span><span style="color:#0086b3;">_C</span><span style="color:#323232;">._distributed_c10d._SymmetricMemory.empty_strided_p2p
</span><span style="color:#323232;">
</span><span style="color:#323232;">
</span><span style="color:#323232;">async_compile.wait(</span><span style="color:#62a35c;">globals</span><span style="color:#323232;">())
</span><span style="font-weight:bold;color:#a71d5d;">del </span><span style="color:#323232;">async_compile
</span><span style="color:#323232;">
</span><span style="font-weight:bold;color:#a71d5d;">def </span><span style="font-weight:bold;color:#323232;">call</span><span style="color:#323232;">(args):
</span><span style="color:#323232;">    arg0_1, arg1_1, arg2_1 </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">args
</span><span style="color:#323232;">    args.clear()
</span><span style="color:#323232;">    assert_size_stride(arg0_1, (</span><span style="color:#0086b3;">4096</span><span style="color:#323232;">, </span><span style="color:#0086b3;">8192</span><span style="color:#323232;">), (</span><span style="color:#0086b3;">8192</span><span style="color:#323232;">, </span><span style="color:#0086b3;">1</span><span style="color:#323232;">))
</span><span style="color:#323232;">    assert_size_stride(arg1_1, (</span><span style="color:#0086b3;">4096</span><span style="color:#323232;">, ), (</span><span style="color:#0086b3;">1</span><span style="color:#323232;">, ))
</span><span style="color:#323232;">    assert_size_stride(arg2_1, (</span><span style="color:#0086b3;">4096</span><span style="color:#323232;">, ), (</span><span style="color:#0086b3;">1</span><span style="color:#323232;">, ))
</span><span style="color:#323232;">    </span><span style="font-weight:bold;color:#a71d5d;">with </span><span style="color:#323232;">torch.cuda._DeviceGuard(</span><span style="color:#0086b3;">0</span><span style="color:#323232;">):
</span><span style="color:#323232;">        torch.cuda.set_device(</span><span style="color:#0086b3;">0</span><span style="color:#323232;">)
</span><span style="color:#323232;">        </span><span style="font-style:italic;color:#969896;"># Topologically Sorted Source Nodes: [scaled_fake_quantize], Original ATen: [quark.scaled_fake_quantize]
</span><span style="color:#323232;">        buf0 </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">torch.ops.quark.scaled_fake_quantize.default(</span><span style="color:#183691;">&#39;int4&#39;</span><span style="color:#323232;">, arg0_1, arg1_1, arg2_1, </span><span style="color:#0086b3;">0</span><span style="color:#323232;">, </span><span style="color:#0086b3;">0</span><span style="color:#323232;">, </span><span style="color:#0086b3;">0.0</span><span style="color:#323232;">, </span><span style="color:#0086b3;">15.0</span><span style="color:#323232;">, </span><span style="color:#0086b3;">0</span><span style="color:#323232;">, </span><span style="color:#183691;">&#39;per_channel&#39;</span><span style="color:#323232;">, </span><span style="color:#183691;">&#39;haha&#39;</span><span style="color:#323232;">)
</span><span style="color:#323232;">        </span><span style="font-weight:bold;color:#a71d5d;">del </span><span style="color:#323232;">arg0_1
</span><span style="color:#323232;">        </span><span style="font-weight:bold;color:#a71d5d;">del </span><span style="color:#323232;">arg1_1
</span><span style="color:#323232;">        </span><span style="font-weight:bold;color:#a71d5d;">del </span><span style="color:#323232;">arg2_1
</span><span style="color:#323232;">        buf1 </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">buf0
</span><span style="color:#323232;">        assert_size_stride(buf1, (</span><span style="color:#0086b3;">4096</span><span style="color:#323232;">, </span><span style="color:#0086b3;">8192</span><span style="color:#323232;">), (</span><span style="color:#0086b3;">8192</span><span style="color:#323232;">, </span><span style="color:#0086b3;">1</span><span style="color:#323232;">))
</span><span style="color:#323232;">        </span><span style="font-weight:bold;color:#a71d5d;">del </span><span style="color:#323232;">buf0
</span><span style="color:#323232;">    </span><span style="font-weight:bold;color:#a71d5d;">return </span><span style="color:#323232;">(buf1, )
</span><span style="color:#323232;">
</span><span style="color:#323232;">
</span><span style="font-weight:bold;color:#a71d5d;">def </span><span style="font-weight:bold;color:#323232;">benchmark_compiled_module</span><span style="color:#323232;">(times</span><span style="font-weight:bold;color:#a71d5d;">=</span><span style="color:#0086b3;">10</span><span style="color:#323232;">, repeat</span><span style="font-weight:bold;color:#a71d5d;">=</span><span style="color:#0086b3;">10</span><span style="color:#323232;">):
</span><span style="color:#323232;">    </span><span style="font-weight:bold;color:#a71d5d;">from </span><span style="color:#323232;">torch._dynamo.testing </span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">rand_strided
</span><span style="color:#323232;">    </span><span style="font-weight:bold;color:#a71d5d;">from </span><span style="color:#323232;">torch._inductor.utils </span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">print_performance
</span><span style="color:#323232;">    arg0_1 </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">rand_strided((</span><span style="color:#0086b3;">4096</span><span style="color:#323232;">, </span><span style="color:#0086b3;">8192</span><span style="color:#323232;">), (</span><span style="color:#0086b3;">8192</span><span style="color:#323232;">, </span><span style="color:#0086b3;">1</span><span style="color:#323232;">), device</span><span style="font-weight:bold;color:#a71d5d;">=</span><span style="color:#183691;">&#39;cuda:0&#39;</span><span style="color:#323232;">, dtype</span><span style="font-weight:bold;color:#a71d5d;">=</span><span style="color:#323232;">torch.float32)
</span><span style="color:#323232;">    arg1_1 </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">rand_strided((</span><span style="color:#0086b3;">4096</span><span style="color:#323232;">, ), (</span><span style="color:#0086b3;">1</span><span style="color:#323232;">, ), device</span><span style="font-weight:bold;color:#a71d5d;">=</span><span style="color:#183691;">&#39;cuda:0&#39;</span><span style="color:#323232;">, dtype</span><span style="font-weight:bold;color:#a71d5d;">=</span><span style="color:#323232;">torch.float32)
</span><span style="color:#323232;">    arg2_1 </span><span style="font-weight:bold;color:#a71d5d;">= </span><span style="color:#323232;">rand_strided((</span><span style="color:#0086b3;">4096</span><span style="color:#323232;">, ), (</span><span style="color:#0086b3;">1</span><span style="color:#323232;">, ), device</span><span style="font-weight:bold;color:#a71d5d;">=</span><span style="color:#183691;">&#39;cuda:0&#39;</span><span style="color:#323232;">, dtype</span><span style="font-weight:bold;color:#a71d5d;">=</span><span style="color:#323232;">torch.int32)
</span><span style="color:#323232;">    fn </span><span style="font-weight:bold;color:#a71d5d;">= lambda</span><span style="color:#323232;">: call([arg0_1, arg1_1, arg2_1])
</span><span style="color:#323232;">    </span><span style="font-weight:bold;color:#a71d5d;">return </span><span style="color:#323232;">print_performance(fn, times</span><span style="font-weight:bold;color:#a71d5d;">=</span><span style="color:#323232;">times, repeat</span><span style="font-weight:bold;color:#a71d5d;">=</span><span style="color:#323232;">repeat)
</span><span style="color:#323232;">
</span><span style="color:#323232;">
</span><span style="font-weight:bold;color:#a71d5d;">if </span><span style="color:#323232;">__name__ </span><span style="font-weight:bold;color:#a71d5d;">== </span><span style="color:#183691;">&quot;__main__&quot;</span><span style="color:#323232;">:
</span><span style="color:#323232;">    </span><span style="font-weight:bold;color:#a71d5d;">from </span><span style="color:#323232;">torch._inductor.wrapper_benchmark </span><span style="font-weight:bold;color:#a71d5d;">import </span><span style="color:#323232;">compiled_module_main
</span><span style="color:#323232;">    compiled_module_main(</span><span style="color:#183691;">&#39;None&#39;</span><span style="color:#323232;">, benchmark_compiled_module)
</span></pre>