How to apply best history after Auto Scheduler for relay.vm.compile

namduc · August 29, 2021, 8:08am

I’m using Auto Scheduler to find the best performance for the model. However, when I apply best history like this:

    with auto_scheduler.ApplyHistoryBest(log_file):
        with tvm.transform.PassContext(opt_level=3, disabled_pass=["FoldScaleAxis"], config={"relay.backend.use_auto_scheduler": True}):
            vm_exec = relay.vm.compile(mod, target=TARGET, params=params)

A lot of log throw to terminal: Cannot find config for target... and model after compile not faster:

Cannot find config for target=llvm -keys=cpu -libs=mkl -link-params=0 -mcpu=core-avx2, workload=('conv2d_NCHWc.x86' .....

Because my model not able to run with relay.build
How I could be applied log file with relay.vm.compile !!?

elvin-n · August 29, 2021, 8:31pm

virtual machine execution cannot be tuned so far

namduc · August 30, 2021, 3:29am

Are there any other solutions !? I spent a lot of time tuning the model. But now I can’t use it

jwfromm · August 30, 2021, 3:04pm

Hi @namduc, deploying with the VM after autoscheduling shoudl be fine and it’s not clear why autoscheduler thinks your logs dont apply to your model. Would it be possible for you to post your tuning script as well?

namduc · August 30, 2021, 3:31pm

@jwfromm Thanks for your support! I use the model architecture customized from the Maskrcnn model
Here is my tuning script:

import tvm
from tvm import relay, auto_scheduler
from tvm.runtime.vm import VirtualMachine

TARGET = tvm.target.Target("llvm -mcpu=broadwell")
log_file = "card_extraction-autoschedule.json"

dummy_input = torch.randn(1, 3, 800, 800,device='cpu', requires_grad=True)
model = torch.jit.trace(model, dummy_input)
mod, params = relay.frontend.from_pytorch(model, input_infos=[('input0', dummy_input.shape)])

print("Extract tasks...")
tasks, task_weights = auto_scheduler.extract_tasks(mod["main"], params, TARGET)

for idx, task in enumerate(tasks):
     print("========== Task %d  (workload key: %s) ==========" % (idx, task.workload_key))
     print(task.compute_dag)

def run_tuning():
     print("Begin tuning...")
     tuner = auto_scheduler.TaskScheduler(tasks, task_weights)
     tune_option = auto_scheduler.TuningOptions(
     num_measure_trials=20000,  
         runner=auto_scheduler.LocalRunner(repeat=10, enable_cpu_cache_flush=True),
         measure_callbacks=[auto_scheduler.RecordToFile(log_file)],
      )
      tuner.tune(tune_option)

run_tuning()

# I apply log file here to compiling model
 with auto_scheduler.ApplyHistoryBest(log_file):
      with tvm.transform.PassContext(opt_level=3, disabled_pass=["FoldScaleAxis"], config={"relay.backend.use_auto_scheduler": True}):
         vm_exec = relay.vm.compile(mod, target=TARGET, params=params)
    
dev = tvm.cpu()
vm = VirtualMachine(vm_exec, dev)
start_t = time.time()
vm.set_input("main", **{"input0": sample.cpu().numpy()})
tvm_res = vm.run()
print(tvm_res[0].numpy().tolist())
print("Inference time of model after tuning: {:0.4f}".format(time.time() - start_t))

jwfromm · August 30, 2021, 5:23pm

Thanks for the sample, I don’t see anything that should be causing issues. Any chance you can post your log file for me to take a look? Also, I recommend benchmarking more than a single run. The first run of a model is much slower than following runs. Consider using benchmark as well. Here’s a sample of how to do it with the VM.

lib = vm.compile(mod, target=target, params=params)
exe = runtime.vm.VirtualMachine(lib, dev)
data = tvm.nd.array(np.random.rand(1, 1, 28, 28).astype("float32"), device=dev)
result = exe.benchmark(dev, data, func_name="main", repeat=2, number=1, end_to_end=True)

AndrewZhaoLuo · August 30, 2021, 8:24pm

Ah I’ve seen this error before.

github.com

AndrewZhaoLuo/TVM-Sandbox/blob/main/autoscheduler/tune_network_cuda.py

"""
Minimum example showing using VM causes autoscheduling logs not to be used properly -- missing configs
"""

import numpy as np

import tvm
from tvm import relay, auto_scheduler
import tvm.relay.testing
from tvm.contrib import graph_executor


def get_network(name, batch_size, layout="NHWC", dtype="float32"):
    """Get the symbol definition and random weight of a network"""

    # auto-scheduler prefers NHWC layout
    if layout == "NHWC":
        image_shape = (224, 224, 3)
    elif layout == "NCHW":
        image_shape = (3, 224, 224)

This file has been truncated. show original

^---- here is a script that in the past could produce a similar error as the one you see OP.

That being said I recently upgraded TVM and runnign this script no longer gives me this error. I suggest grabbing the latests TVM?

For whatever reason sometimes using the VM on autoscheduling logs hits the autotvm lowering pipeline instead of the autoscheduling lowering pipeline which is what causes this error. I haven’t really had time to investigate the reason why.

masahi · August 30, 2021, 8:32pm

MaskRCNN has dynamic-batch conv2d, conv2d transpose, and dense that cannot be tuned. So after tuning you would still see that warning because of those dynamic workloads.

namduc · August 31, 2021, 1:53am

@jwfromm Thanks
Here is a part of my log file:

{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [1, 1, 4], 1], ["SP", 3, 8, 400, [5, 4, 4], 1], ["SP", 3, 12, 400, [20, 20, 1], 1], ["SP", 3, 16, 8, [1, 1, 8], 1], ["SP", 3, 20, 3, [3], 1], ["SP", 3, 22, 7, [7], 1], ["SP", 3, 24, 7, [7], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 1], ["FSP", 6, 2, 2, 1], ["FSP", 6, 4, 3, 1], ["FSP", 6, 6, 4, 1], ["FSP", 6, 8, 5, 1], ["RE", 6, [0, 2, 4, 6, 8, 1, 3, 5, 7, 9]], ["CA", 3, 6, 4], ["CA", 1, 3, 9], ["FU", 6, [0, 1, 2, 3, 4]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$16"], ["AN", 1, 4, 2], ["AN", 3, 25, 2], ["AN", 6, 5, 2]]]], "r": [[0.0153098, 0.015353, 0.0153929, 0.0152849, 0.0153374, 0.0153348, 0.0152953, 0.015309, 0.0152716, 0.0153824], 0, 1.84983, 1630239320], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [4, 1, 1], 1], ["SP", 3, 8, 400, [25, 4, 4], 1], ["SP", 3, 12, 400, [1, 4, 5], 1], ["SP", 3, 16, 8, [8, 1, 1], 1], ["SP", 3, 20, 3, [3], 1], ["SP", 3, 22, 7, [1], 1], ["SP", 3, 24, 7, [1], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 2], ["FSP", 6, 3, 2, 2], ["FSP", 6, 6, 3, 2], ["FSP", 6, 9, 4, 2], ["FSP", 6, 12, 5, 2], ["RE", 6, [0, 3, 6, 9, 12, 1, 4, 7, 10, 13, 2, 5, 8, 11, 14]], ["CA", 3, 6, 9], ["CA", 1, 6, 2], ["FU", 6, [0, 1, 2]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$512"], ["AN", 1, 4, 2], ["AN", 3, 25, 2], ["AN", 6, 12, 2]]]], "r": [[0.215612, 0.214613, 0.214669, 0.214758, 0.214848, 0.218075, 0.214809, 0.21486, 0.215318, 0.214682], 0, 5.00403, 1630239322], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [1, 4, 2], 1], ["SP", 3, 8, 400, [1, 5, 40], 1], ["SP", 3, 12, 400, [1, 20, 10], 1], ["SP", 3, 16, 8, [1, 4, 1], 1], ["SP", 3, 20, 3, [3], 1], ["SP", 3, 22, 7, [1], 1], ["SP", 3, 24, 7, [7], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 1], ["FSP", 6, 2, 2, 1], ["FSP", 6, 4, 3, 1], ["FSP", 6, 6, 4, 1], ["FSP", 6, 8, 5, 1], ["RE", 6, [0, 2, 4, 6, 8, 1, 3, 5, 7, 9]], ["CA", 3, 6, 4], ["CR", 1], ["FU", 1, [0, 1, 2]], ["AN", 1, 0, 3], ["FU", 6, [0, 1, 2, 3, 4]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$0"], ["AN", 1, 2, 2], ["AN", 3, 25, 2], ["AN", 6, 5, 2]]]], "r": [[0.227671, 0.218972, 0.223055, 0.221485, 0.220519, 0.219144, 0.222717, 0.219439, 0.220186, 0.220731], 0, 5.57451, 1630239325], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [1, 1, 8], 1], ["SP", 3, 8, 400, [5, 16, 5], 1], ["SP", 3, 12, 400, [5, 8, 10], 1], ["SP", 3, 16, 8, [8, 1, 1], 1], ["SP", 3, 20, 3, [3], 1], ["SP", 3, 22, 7, [1], 1], ["SP", 3, 24, 7, [1], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["CA", 1, 3, 7], ["FU", 3, [0, 1, 2, 3, 4, 5, 6, 7]], ["AN", 3, 0, 3], ["FU", 6, [0, 1, 2]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$0"], ["AN", 1, 4, 2], ["AN", 3, 18, 2], ["AN", 6, 2, 2]]]], "r": [[0.363728, 0.370664, 0.368474, 0.368962, 0.367229, 0.368955, 0.373197, 0.366844, 0.368602, 0.375494], 0, 8.51309, 1630239329], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [8, 1, 1], 1], ["SP", 3, 8, 400, [40, 2, 1], 1], ["SP", 3, 12, 400, [10, 2, 4], 1], ["SP", 3, 16, 8, [2, 4, 1], 1], ["SP", 3, 20, 3, [1], 1], ["SP", 3, 22, 7, [7], 1], ["SP", 3, 24, 7, [7], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 2], ["FSP", 6, 3, 2, 2], ["FSP", 6, 6, 3, 2], ["FSP", 6, 9, 4, 2], ["FSP", 6, 12, 5, 2], ["RE", 6, [0, 3, 6, 9, 12, 1, 4, 7, 10, 13, 2, 5, 8, 11, 14]], ["CA", 3, 6, 9], ["CA", 1, 3, 9], ["FU", 6, [0, 1, 2, 3, 4, 5, 6]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$0"], ["AN", 1, 4, 2], ["AN", 3, 25, 2], ["AN", 6, 8, 2]]]], "r": [[0.0414936, 0.0406137, 0.0407088, 0.040657, 0.0406766, 0.0406778, 0.0406458, 0.0407365, 0.0406257, 0.0405932], 0, 1.27985, 1630239330], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [2, 1, 4], 1], ["SP", 3, 8, 400, [5, 16, 5], 1], ["SP", 3, 12, 400, [25, 2, 1], 1], ["SP", 3, 16, 8, [1, 1, 2], 1], ["SP", 3, 20, 3, [3], 1], ["SP", 3, 22, 7, [7], 1], ["SP", 3, 24, 7, [7], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 2], ["FSP", 6, 3, 2, 2], ["FSP", 6, 6, 3, 2], ["FSP", 6, 9, 4, 2], ["FSP", 6, 12, 5, 2], ["RE", 6, [0, 3, 6, 9, 12, 1, 4, 7, 10, 13, 2, 5, 8, 11, 14]], ["CA", 3, 6, 9], ["CR", 1], ["FU", 1, [0, 1, 2]], ["AN", 1, 0, 3], ["FU", 6, [0, 1, 2, 3, 4, 5, 6, 7, 8]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$0"], ["AN", 1, 2, 2], ["AN", 3, 25, 2], ["AN", 6, 6, 2]]]], "r": [[0.0346851, 0.0347001, 0.0346628, 0.034395, 0.0355137, 0.0347808, 0.0349785, 0.0347191, 0.0347812, 0.0352034], 0, 1.50895, 1630239331], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [4, 2, 1], 1], ["SP", 3, 8, 400, [1, 5, 1], 1], ["SP", 3, 12, 400, [2, 4, 25], 1], ["SP", 3, 16, 8, [2, 4, 1], 1], ["SP", 3, 20, 3, [3], 1], ["SP", 3, 22, 7, [7], 1], ["SP", 3, 24, 7, [1], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["CR", 6], ["CA", 1, 3, 12], ["FU", 3, [0, 1, 2, 3, 4, 5, 6]], ["AN", 3, 0, 3], ["FU", 6, [0, 1, 2]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$0"], ["AN", 1, 4, 2], ["AN", 3, 19, 2], ["AN", 6, 2, 2]]]], "r": [[0.133869, 0.088203, 0.0883493, 0.0884591, 0.106855, 0.093092, 0.223594, 0.0895769, 0.0887379, 0.0887455], 0, 4.6319, 1630239333], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [2, 1, 1], 1], ["SP", 3, 8, 400, [5, 2, 4], 1], ["SP", 3, 12, 400, [50, 2, 2], 1], ["SP", 3, 16, 8, [1, 1, 1], 1], ["SP", 3, 20, 3, [3], 1], ["SP", 3, 22, 7, [1], 1], ["SP", 3, 24, 7, [7], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 1], ["FSP", 6, 2, 2, 1], ["FSP", 6, 4, 3, 1], ["FSP", 6, 6, 4, 1], ["FSP", 6, 8, 5, 1], ["RE", 6, [0, 2, 4, 6, 8, 1, 3, 5, 7, 9]], ["CA", 3, 6, 4], ["CA", 1, 3, 12], ["FU", 6, [0, 1, 2, 3]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$512"], ["AN", 1, 4, 2], ["AN", 3, 25, 2], ["AN", 6, 6, 2]]]], "r": [[0.108386, 0.0913435, 0.0917212, 0.0915141, 0.113009, 0.0916258, 0.0915183, 0.0953672, 0.0939548, 0.0918529], 0, 2.97536, 1630239334], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [1, 1, 1], 1], ["SP", 3, 8, 400, [2, 5, 5], 1], ["SP", 3, 12, 400, [4, 20, 1], 1], ["SP", 3, 16, 8, [1, 2, 4], 1], ["SP", 3, 20, 3, [3], 1], ["SP", 3, 22, 7, [7], 1], ["SP", 3, 24, 7, [1], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 2], ["FSP", 6, 3, 2, 2], ["FSP", 6, 6, 3, 2], ["FSP", 6, 9, 4, 2], ["FSP", 6, 12, 5, 2], ["RE", 6, [0, 3, 6, 9, 12, 1, 4, 7, 10, 13, 2, 5, 8, 11, 14]], ["CA", 3, 6, 9], ["CA", 1, 6, 4], ["FU", 6, [0, 1, 2, 3]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$512"], ["AN", 1, 4, 2], ["AN", 3, 25, 2], ["AN", 6, 11, 2]]]], "r": [[0.0147756, 0.0148049, 0.0146872, 0.0147084, 0.0148427, 0.0148288, 0.0148515, 0.0147357, 0.0148098, 0.0147129], 0, 1.39473, 1630239335], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [1, 1, 8], 1], ["SP", 3, 8, 400, [1, 16, 1], 1], ["SP", 3, 12, 400, [10, 1, 10], 1], ["SP", 3, 16, 8, [2, 1, 1], 1], ["SP", 3, 20, 3, [1], 1], ["SP", 3, 22, 7, [1], 1], ["SP", 3, 24, 7, [1], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["CR", 6], ["CA", 1, 3, 5], ["FU", 3, [0, 1, 2, 3, 4]], ["AN", 3, 0, 3], ["FU", 6, [0, 1, 2]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$16"], ["AN", 1, 4, 2], ["AN", 3, 21, 2], ["AN", 6, 2, 2]]]], "r": [[0.140055, 0.136473, 0.136476, 0.140206, 0.136558, 0.136571, 0.136449, 0.136639, 0.14712, 0.136627], 0, 2.42857, 1630239336], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [8, 1, 1], 1], ["SP", 3, 8, 400, [20, 2, 5], 1], ["SP", 3, 12, 400, [4, 5, 5], 1], ["SP", 3, 16, 8, [4, 1, 1], 1], ["SP", 3, 20, 3, [1], 1], ["SP", 3, 22, 7, [7], 1], ["SP", 3, 24, 7, [7], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["CA", 1, 3, 8], ["FU", 3, [0, 1, 2, 3, 4, 5, 6, 7]], ["AN", 3, 0, 3], ["FU", 6, [0, 1, 2]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$0"], ["AN", 1, 4, 2], ["AN", 3, 18, 2], ["AN", 6, 2, 2]]]], "r": [[0.0746363, 0.0745412, 0.0746709, 0.0746361, 0.0745688, 0.0746332, 0.0754652, 0.0745272, 0.0743144, 0.0743876], 0, 3.86081, 1630239338], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [2, 1, 1], 1], ["SP", 3, 8, 400, [4, 1, 4], 1], ["SP", 3, 12, 400, [5, 1, 4], 1], ["SP", 3, 16, 8, [1, 1, 1], 1], ["SP", 3, 20, 3, [1], 1], ["SP", 3, 22, 7, [1], 1], ["SP", 3, 24, 7, [7], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 2], ["FSP", 6, 3, 2, 2], ["FSP", 6, 6, 3, 2], ["FSP", 6, 9, 4, 2], ["FSP", 6, 12, 5, 2], ["RE", 6, [0, 3, 6, 9, 12, 1, 4, 7, 10, 13, 2, 5, 8, 11, 14]], ["CA", 3, 6, 9], ["CA", 1, 3, 8], ["FU", 6, [0, 1, 2, 3]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$0"], ["AN", 1, 4, 2], ["AN", 3, 25, 2], ["AN", 6, 11, 2]]]], "r": [[0.083342, 0.0633061, 0.059164, 0.0590534, 0.0590991, 0.058975, 0.0591572, 0.0593549, 0.0591005, 0.0590778], 0, 1.36869, 1630239339], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [2, 1, 1], 1], ["SP", 3, 8, 400, [1, 8, 10], 1], ["SP", 3, 12, 400, [2, 20, 1], 1], ["SP", 3, 16, 8, [4, 1, 2], 1], ["SP", 3, 20, 3, [3], 1], ["SP", 3, 22, 7, [1], 1], ["SP", 3, 24, 7, [1], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 2], ["FSP", 6, 3, 2, 2], ["FSP", 6, 6, 3, 2], ["FSP", 6, 9, 4, 2], ["FSP", 6, 12, 5, 2], ["RE", 6, [0, 3, 6, 9, 12, 1, 4, 7, 10, 13, 2, 5, 8, 11, 14]], ["CA", 3, 6, 9], ["CA", 1, 3, 8], ["FU", 6, [0, 1, 2, 3]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$16"], ["AN", 1, 4, 2], ["AN", 3, 25, 2], ["AN", 6, 11, 2]]]], "r": [[0.0329612, 0.032938, 0.0328995, 0.0329447, 0.0329467, 0.0334307, 0.0329345, 0.0329786, 0.0329294, 0.0329546], 0, 2.62262, 1630239339], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [2, 2, 2], 1], ["SP", 3, 8, 400, [10, 8, 5], 1], ["SP", 3, 12, 400, [1, 2, 4], 1], ["SP", 3, 16, 8, [1, 4, 1], 1], ["SP", 3, 20, 3, [1], 1], ["SP", 3, 22, 7, [1], 1], ["SP", 3, 24, 7, [1], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["CR", 6], ["CA", 1, 3, 5], ["FU", 3, [0, 1, 2, 3, 4, 5]], ["AN", 3, 0, 3], ["FU", 6, [0, 1, 2]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$512"], ["AN", 1, 4, 2], ["AN", 3, 20, 2], ["AN", 6, 2, 2]]]], "r": [[0.117331, 0.117277, 0.117283, 0.117122, 0.117265, 0.117214, 0.117242, 0.117132, 0.117225, 0.117247], 0, 4.81655, 1630239341], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [4, 1, 2], 1], ["SP", 3, 8, 400, [25, 1, 8], 1], ["SP", 3, 12, 400, [5, 8, 2], 1], ["SP", 3, 16, 8, [2, 1, 1], 1], ["SP", 3, 20, 3, [3], 1], ["SP", 3, 22, 7, [1], 1], ["SP", 3, 24, 7, [7], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 2], ["FSP", 6, 3, 2, 2], ["FSP", 6, 6, 3, 2], ["FSP", 6, 9, 4, 2], ["FSP", 6, 12, 5, 2], ["RE", 6, [0, 3, 6, 9, 12, 1, 4, 7, 10, 13, 2, 5, 8, 11, 14]], ["CA", 3, 6, 9], ["CA", 1, 6, 2], ["FU", 6, [0, 1, 2]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$64"], ["AN", 1, 4, 2], ["AN", 3, 25, 2], ["AN", 6, 12, 2]]]], "r": [[0.258307, 0.258151, 0.25922, 0.260498, 0.258435, 0.258232, 0.258002, 0.258331, 0.258371, 0.260911], 0, 7.12184, 1630239344], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [1, 2, 2], 1], ["SP", 3, 8, 400, [5, 1, 1], 1], ["SP", 3, 12, 400, [5, 2, 8], 1], ["SP", 3, 16, 8, [4, 2, 1], 1], ["SP", 3, 20, 3, [1], 1], ["SP", 3, 22, 7, [7], 1], ["SP", 3, 24, 7, [7], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 2], ["FSP", 6, 3, 2, 2], ["FSP", 6, 6, 3, 2], ["FSP", 6, 9, 4, 2], ["FSP", 6, 12, 5, 2], ["RE", 6, [0, 3, 6, 9, 12, 1, 4, 7, 10, 13, 2, 5, 8, 11, 14]], ["CA", 3, 6, 9], ["CA", 1, 6, 7], ["FU", 6, [0, 1, 2, 3]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$64"], ["AN", 1, 4, 2], ["AN", 3, 25, 2], ["AN", 6, 11, 2]]]], "r": [[0.0545458, 0.054685, 0.055003, 0.0548082, 0.0545325, 0.0545613, 0.0545352, 0.0545277, 0.0546259, 0.0547964], 0, 1.75536, 1630239345], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [1, 1, 8], 1], ["SP", 3, 8, 400, [8, 1, 2], 1], ["SP", 3, 12, 400, [5, 8, 10], 1], ["SP", 3, 16, 8, [2, 2, 2], 1], ["SP", 3, 20, 3, [1], 1], ["SP", 3, 22, 7, [7], 1], ["SP", 3, 24, 7, [1], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 2], ["FSP", 6, 3, 2, 2], ["FSP", 6, 6, 3, 2], ["FSP", 6, 9, 4, 2], ["FSP", 6, 12, 5, 2], ["RE", 6, [0, 3, 6, 9, 12, 1, 4, 7, 10, 13, 2, 5, 8, 11, 14]], ["CA", 3, 6, 9], ["CA", 1, 3, 5], ["FU", 6, [0, 1]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$16"], ["AN", 1, 4, 2], ["AN", 3, 25, 2], ["AN", 6, 13, 2]]]], "r": [[0.235892, 0.234964, 0.236793, 0.235233, 0.235298, 0.236097, 0.237922, 0.235726, 0.237895, 0.235831], 0, 4.65153, 1630239348], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [1, 2, 4], 1], ["SP", 3, 8, 400, [2, 10, 10], 1], ["SP", 3, 12, 400, [10, 2, 1], 1], ["SP", 3, 16, 8, [2, 2, 2], 1], ["SP", 3, 20, 3, [1], 1], ["SP", 3, 22, 7, [7], 1], ["SP", 3, 24, 7, [1], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 2], ["FSP", 6, 3, 2, 2], ["FSP", 6, 6, 3, 2], ["FSP", 6, 9, 4, 2], ["FSP", 6, 12, 5, 2], ["RE", 6, [0, 3, 6, 9, 12, 1, 4, 7, 10, 13, 2, 5, 8, 11, 14]], ["CA", 3, 6, 9], ["CA", 1, 6, 3], ["FU", 6, [0, 1, 2, 3]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$512"], ["AN", 1, 4, 2], ["AN", 3, 25, 2], ["AN", 6, 11, 2]]]], "r": [[0.047996, 0.0481619, 0.0480023, 0.0479662, 0.0779532, 0.0479476, 0.0480034, 0.0478164, 0.0480926, 0.0479665], 0, 2.52389, 1630239349], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [1, 1, 2], 1], ["SP", 3, 8, 400, [1, 2, 1], 1], ["SP", 3, 12, 400, [5, 40, 1], 1], ["SP", 3, 16, 8, [1, 4, 2], 1], ["SP", 3, 20, 3, [1], 1], ["SP", 3, 22, 7, [1], 1], ["SP", 3, 24, 7, [1], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 2], ["FSP", 6, 3, 2, 2], ["FSP", 6, 6, 3, 2], ["FSP", 6, 9, 4, 2], ["FSP", 6, 12, 5, 2], ["RE", 6, [0, 3, 6, 9, 12, 1, 4, 7, 10, 13, 2, 5, 8, 11, 14]], ["CA", 3, 6, 9], ["CA", 1, 3, 6], ["FU", 6, [0, 1, 2]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$512"], ["AN", 1, 4, 2], ["AN", 3, 25, 2], ["AN", 6, 12, 2]]]], "r": [[0.0410378, 0.0412974, 0.0412714, 0.0411679, 0.0412848, 0.0412572, 0.0410735, 0.0412144, 0.0412247, 0.0412169], 0, 2.41664, 1630239350], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [1, 1, 1], 1], ["SP", 3, 8, 400, [5, 2, 20], 1], ["SP", 3, 12, 400, [10, 10, 2], 1], ["SP", 3, 16, 8, [1, 1, 1], 1], ["SP", 3, 20, 3, [3], 1], ["SP", 3, 22, 7, [7], 1], ["SP", 3, 24, 7, [1], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 2], ["FSP", 6, 3, 2, 2], ["FSP", 6, 6, 3, 2], ["FSP", 6, 9, 4, 2], ["FSP", 6, 12, 5, 2], ["RE", 6, [0, 3, 6, 9, 12, 1, 4, 7, 10, 13, 2, 5, 8, 11, 14]], ["CA", 3, 6, 9], ["CA", 1, 3, 9], ["FU", 6, [0, 1, 2, 3, 4, 5, 6, 7, 8]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$16"], ["AN", 1, 4, 2], ["AN", 3, 25, 2], ["AN", 6, 6, 2]]]], "r": [[0.110274, 0.109006, 0.110994, 0.107778, 0.10959, 0.108713, 0.107351, 0.109196, 0.117567, 0.110073], 0, 2.58545, 1630239351], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [2, 1, 1], 1], ["SP", 3, 8, 400, [4, 1, 5], 1], ["SP", 3, 12, 400, [5, 10, 4], 1], ["SP", 3, 16, 8, [4, 1, 2], 1], ["SP", 3, 20, 3, [1], 1], ["SP", 3, 22, 7, [7], 1], ["SP", 3, 24, 7, [1], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["FSP", 6, 0, 1, 1], ["FSP", 6, 2, 2, 1], ["FSP", 6, 4, 3, 1], ["FSP", 6, 6, 4, 1], ["FSP", 6, 8, 5, 1], ["RE", 6, [0, 2, 4, 6, 8, 1, 3, 5, 7, 9]], ["CA", 3, 6, 4], ["CA", 1, 3, 11], ["FU", 6, [0, 1, 2, 3, 4]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$16"], ["AN", 1, 4, 2], ["AN", 3, 25, 2], ["AN", 6, 5, 2]]]], "r": [[0.0541485, 0.0376101, 0.037688, 0.0379611, 0.0377433, 0.0376278, 0.0376677, 0.0376754, 0.0377531, 0.0384117], 0, 1.20223, 1630239352], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [2, 1, 1], 1], ["SP", 3, 8, 400, [8, 5, 1], 1], ["SP", 3, 12, 400, [1, 40, 5], 1], ["SP", 3, 16, 8, [1, 4, 2], 1], ["SP", 3, 20, 3, [3], 1], ["SP", 3, 22, 7, [1], 1], ["SP", 3, 24, 7, [1], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["CR", 6], ["CA", 1, 3, 4], ["FU", 3, [0, 1, 2, 3, 4]], ["AN", 3, 0, 3], ["FU", 6, [0, 1, 2]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$64"], ["AN", 1, 4, 2], ["AN", 3, 21, 2], ["AN", 6, 2, 2]]]], "r": [[0.0371947, 0.0371288, 0.037179, 0.0380376, 0.0371198, 0.0371033, 0.0371491, 0.0370496, 0.03719, 0.0372351], 0, 1.67623, 1630239353], "v": "v0.6"}
{"i": [["[\"8de29ce92833b18f317071115f796b28\", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]", "llvm -keys=cpu -link-params=0 -mcpu=broadwell", [12, 64, 64, 0, 0, 0, 0, 0], "", 2, []], [[], [["CI", 5], ["SP", 3, 0, 1, [1, 1, 1], 1], ["SP", 3, 4, 8, [4, 2, 1], 1], ["SP", 3, 8, 400, [1, 5, 1], 1], ["SP", 3, 12, 400, [2, 4, 25], 1], ["SP", 3, 16, 8, [2, 4, 1], 1], ["SP", 3, 20, 3, [1], 1], ["SP", 3, 22, 7, [1], 1], ["SP", 3, 24, 7, [1], 1], ["RE", 3, [0, 4, 8, 12, 16, 1, 5, 9, 13, 17, 20, 22, 24, 2, 6, 10, 14, 18, 21, 23, 25, 3, 7, 11, 15, 19]], ["CR", 6], ["CA", 1, 3, 12], ["FU", 3, [0, 1, 2, 3, 4, 5, 6]], ["AN", 3, 0, 3], ["FU", 6, [0, 1, 2]], ["AN", 6, 0, 3], ["PR", 3, 0, "auto_unroll_max_step$0"], ["AN", 1, 4, 2], ["AN", 3, 19, 2], ["AN", 6, 2, 2]]]], "r": [[0.160069, 0.161015, 0.167056, 0.160802, 0.247418, 0.272026, 0.228201, 0.183938, 0.195762, 0.287887], 0, 4.5092, 1630239356], "v": "v0.6"}

namduc · August 31, 2021, 2:10am

Hmm… I got a lot of warnings, almost on all tasks, a part of my output:

WARNING:autotvm:Cannot find config for target=llvm -keys=cpu -link-params=0 -mcpu=broadwell, workload=('conv2d_NCHWc.x86', ('TENSOR', (1, 256, 25, 25), 'float32'), ('TENSOR', (3, 256, 1, 1), 'float32'), (1, 1), (0, 0, 0, 0), (1, 1), 'NCHW', 'NCHW', 'float32'). A fallback configuration is used, which may bring great performance regression.
WARNING:autotvm:Cannot find config for target=llvm -keys=cpu -link-params=0 -mcpu=broadwell, workload=('conv2d_NCHWc.x86', ('TENSOR', (1, 256, 13, 13), 'float32'), ('TENSOR', (3, 256, 1, 1), 'float32'), (1, 1), (0, 0, 0, 0), (1, 1), 'NCHW', 'NCHW', 'float32'). A fallback configuration is used, which may bring great performance regression.
WARNING:autotvm:Cannot find config for target=llvm -keys=cpu -link-params=0 -mcpu=broadwell, workload=('conv2d_NCHWc.x86', ('TENSOR', (?, 256, 14, 14), 'float32'), ('TENSOR', (256, 256, 3, 3), 'float32'), (1, 1), (1, 1, 1, 1), (1, 1), 'NCHW', 'NCHW', 'float32'). A fallback configuration is used, which may bring great performance regression.
WARNING:autotvm:Cannot find config for target=llvm -keys=cpu -link-params=0 -mcpu=broadwell, workload=('conv2d_NCHWc.x86', ('TENSOR', (?, 256, 28, 28), 'float32'), ('TENSOR', (2, 256, 1, 1), 'float32'), (1, 1), (0, 0, 0, 0), (1, 1), 'NCHW', 'NCHW', 'float32'). A fallback configuration is used, which may bring great performance regression.
WARNING:auto_scheduler:-----------------------------------
fused_nn.contrib_conv2d_NCHWc_multiply_add_nn.relu
Cannot find tuned schedules for target=llvm -keys=cpu -link-params=0 -mcpu=broadwell, workload_key=["819ea812b87c43178150fe4a63b6733e", [1, 1, 800, 800, 3], [8, 1, 7, 7, 3, 8], [1, 8, 1, 1, 8], [1, 8, 1, 1, 8], [1, 8, 400, 400, 8]]. A fallback TOPI schedule is used, which may bring great performance regression or even compilation failure. Compute DAG info:
placeholder = PLACEHOLDER [1, 1, 800, 800, 3]
data_pad(i0, i1, i2, i3, i4) = tir.if_then_else(((((i2 >= 3) && (i2 < 803)) && (i3 >= 3)) && (i3 < 803)), placeholder[i0, i1, (i2 - 3), (i3 - 3), i4], 0f)
placeholder = PLACEHOLDER [8, 1, 7, 7, 3, 8]
conv2d_NCHWc(n, oc_chunk, oh, ow, oc_block) += (data_pad[n, floordiv(ic, 3), ((oh*2) + kh), ((ow*2) + kw), floormod(ic, 3)]*placeholder[oc_chunk, floordiv(ic, 3), kh, kw, floormod(ic, 3), oc_block])
placeholder = PLACEHOLDER [1, 8, 1, 1, 8]
T_multiply(ax0, ax1, ax2, ax3, ax4) = (conv2d_NCHWc[ax0, ax1, ax2, ax3, ax4]*placeholder[ax0, ax1, 0, 0, ax4])
placeholder = PLACEHOLDER [1, 8, 1, 1, 8]
T_add(ax0, ax1, ax2, ax3, ax4) = (T_multiply[ax0, ax1, ax2, ax3, ax4] + placeholder[ax0, ax1, 0, 0, ax4])
T_relu(ax0, ax1, ax2, ax3, ax4) = max(T_add[ax0, ax1, ax2, ax3, ax4], 0f)

WARNING:auto_scheduler:-----------------------------------
fused_nn.contrib_conv2d_NCHWc_multiply_add_nn.relu
Cannot find tuned schedules for target=llvm -keys=cpu -link-params=0 -mcpu=broadwell, workload_key=["4974391240cf2f772e7c729c89d682e9", [1, 8, 200, 200, 8], [8, 8, 1, 1, 8, 8], [1, 8, 1, 1, 8], [1, 8, 1, 1, 8], [1, 8, 200, 200, 8]]. A fallback TOPI schedule is used, which may bring great performance regression or even compilation failure. Compute DAG info:
placeholder = PLACEHOLDER [1, 8, 200, 200, 8]
placeholder = PLACEHOLDER [8, 8, 1, 1, 8, 8]
conv2d_NCHWc(n, oc_chunk, oh, ow, oc_block) += (placeholder[n, floordiv(ic, 8), (oh + kh), (ow + kw), floormod(ic, 8)]*placeholder[oc_chunk, floordiv(ic, 8), kh, kw, floormod(ic, 8), oc_block])
placeholder = PLACEHOLDER [1, 8, 1, 1, 8]
T_multiply(ax0, ax1, ax2, ax3, ax4) = (conv2d_NCHWc[ax0, ax1, ax2, ax3, ax4]*placeholder[ax0, ax1, 0, 0, ax4])
placeholder = PLACEHOLDER [1, 8, 1, 1, 8]
T_add(ax0, ax1, ax2, ax3, ax4) = (T_multiply[ax0, ax1, ax2, ax3, ax4] + placeholder[ax0, ax1, 0, 0, ax4])
T_relu(ax0, ax1, ax2, ax3, ax4) = max(T_add[ax0, ax1, ax2, ax3, ax4], 0f)

namduc · August 31, 2021, 2:44am

Yepp
I agree, seem using the VM on autoscheduling logs hits the autotvm instead of the autoscheduling, warnings in my output proves it:

WARNING:autotvm:Cannot find config for target=llvm -keys=cpu -link-params=0 -mcpu=broadwell, workload=('conv2d_NCHWc.x86', ('TENSOR', (1, 3, 800, 800), 'float32'), ('TENSOR', (64, 3, 7, 7), 'float32'), (2, 2), (3, 3, 3, 3), (1, 1), 'NCHW', 'NCHW', 'float32'). A fallback configuration is used, which may bring great performance regression.

My version tvm: tvm-0.8.dev1577+gf1ca91d4e-py3.7-linux-x86_64.egg

AndrewZhaoLuo · August 31, 2021, 4:05am

I would checkout commit 7214f5239dbb8da4585d4d10fbc8c65c8f155b12 and rebuild and reinstall TVM (Install from Source — tvm 0.8.dev0 documentation). I can confirm my script above does not have the error anymore in that environment. Don’t checkout commits after this since I made a change to the log structure for autoscheduler and you’ll need to retune to use that.

namduc · August 31, 2021, 4:16am

I have tested your tune_network file in my tvm environment. No errors are thrown and compiling vm success.
I noticed there are some other points in your tuning function:

 measure_ctx = auto_scheduler.LocalRPCMeasureContext(repeat=1, timeout=10)

and here:

 runner=measure_ctx.runner,

Does it makes difference !?

AndrewZhaoLuo · August 31, 2021, 6:19am

I believe this is important with CUDA since CUDA is finicky unless you isolate the process using it. If you are using cpu target then you can just use LocalRunner. It should not affect logs.