wehub-resource-sync
115 lines
3.9 KiB
Python
115 lines
3.9 KiB
Python
import sys
|
|
|
|
from cv2 import repeat
|
|
import tvm
|
|
import onnx
|
|
import time
|
|
import numpy as np
|
|
from tvm import relay, autotvm
|
|
from tvm.contrib import ndk
|
|
from tvm.contrib.utils import tempdir
|
|
import tvm.contrib.graph_runtime as runtime
|
|
|
|
def print_progress(msg):
|
|
"""print progress message
|
|
|
|
Parameters
|
|
----------
|
|
msg: str
|
|
The message to print
|
|
"""
|
|
sys.stdout.write(msg + "\r")
|
|
sys.stdout.flush()
|
|
|
|
# host
|
|
host = '30.206.32.132'
|
|
port = 9090
|
|
key = 'android'
|
|
# arch
|
|
arch = "arm64"
|
|
target = "llvm -mtriple=%s-linux-android" % arch
|
|
target_host = None
|
|
# evaluate
|
|
repeat = 3
|
|
|
|
if __name__ == '__main__':
|
|
model = sys.argv[1]
|
|
input_name1 = "unique_ids_raw_output___9:0"
|
|
input_name2 = "segment_ids:0"
|
|
input_name3 = "input_mask:0"
|
|
input_name4 = "input_ids:0"
|
|
dtype = "int64"
|
|
input_shape1 = [1]
|
|
input_shape2 = [1, 256]
|
|
shape_dict = { input_name1 : input_shape1, input_name2 : input_shape2, input_name3 : input_shape2, input_name4 : input_shape2 }
|
|
# load onnx model
|
|
onnx_model = onnx.load(model)
|
|
# relay load
|
|
mod, params = relay.frontend.from_onnx(onnx_model, shape_dict)
|
|
# relay build
|
|
if len(sys.argv) > 2:
|
|
# log_file = "android.mobilenetv2-7-modify.log"
|
|
log_file = sys.argv[2]
|
|
with autotvm.apply_history_best(log_file):
|
|
with tvm.transform.PassContext(opt_level=3):
|
|
lib = relay.build(mod, target=target, target_host=target_host, params=params)
|
|
# save file
|
|
tmp = tempdir()
|
|
filename = "net.so"
|
|
lib.export_library(tmp.relpath(filename), ndk.create_shared)
|
|
|
|
# upload
|
|
print_progress("uploading...")
|
|
tracker = tvm.rpc.connect_tracker(host, port)
|
|
remote = tracker.request(key)
|
|
ctx = remote.context(str(target), 0)
|
|
remote.upload(tmp.relpath(filename))
|
|
rlib = remote.load_module(filename)
|
|
module = runtime.GraphModule(rlib["default"](ctx))
|
|
data_tvm = tvm.nd.array((np.random.uniform(size=input_shape)).astype(dtype))
|
|
|
|
# evaluate
|
|
print_progress("evaluating...")
|
|
module.set_input(input_name, data_tvm)
|
|
ftimer = module.module.time_evaluator("run", ctx, number=1, repeat=repeat)
|
|
prof_res = np.array(ftimer().results) * 1000
|
|
print(
|
|
"avg time: %-19s (%s)" % ("%.2f ms" % np.mean(prof_res), "%.2f ms" % np.std(prof_res))
|
|
)
|
|
else:
|
|
with tvm.transform.PassContext(opt_level=3):
|
|
lib = relay.build(mod, target=target, target_host=target_host, params=params)
|
|
# save file
|
|
tmp = tempdir()
|
|
filename = "net.so"
|
|
#lib.export_library(tmp.relpath(filename), ndk.create_shared)
|
|
lib.export_library(filename, ndk.create_shared)
|
|
|
|
# upload
|
|
tracker = tvm.rpc.connect_tracker(host, port)
|
|
remote = tracker.request(key)
|
|
ctx = remote.context(str(target), 0)
|
|
print("uploading...")
|
|
# remote.upload(tmp.relpath(filename))
|
|
remote.upload(filename)
|
|
print("upload done")
|
|
rlib = remote.load_module(filename)
|
|
module = runtime.GraphModule(rlib["default"](ctx))
|
|
data_1 = tvm.nd.array((np.random.uniform(size=input_shape1)).astype(dtype))
|
|
data_2 = tvm.nd.array((np.random.uniform(size=input_shape2)).astype(dtype))
|
|
|
|
# evaluate
|
|
print("evaluating...")
|
|
module.set_input(input_name1, data_1)
|
|
module.set_input(input_name2, data_2)
|
|
module.set_input(input_name3, data_2)
|
|
module.set_input(input_name4, data_2)
|
|
t1 = time.time()
|
|
ftimer = module.module.time_evaluator("run", ctx, number=1, repeat=repeat)
|
|
t2 = time.time()
|
|
print('evaluator time : {} ms'.format(1000 * (t2 - t1)))
|
|
prof_res = np.array(ftimer().results) * 1000
|
|
print(
|
|
"avg time: %-19s (%s)" % ("%.2f ms" % np.mean(prof_res), "%.2f ms" % np.std(prof_res))
|
|
)
|