diff --git a/tutorials/autotvm/tune_conv2d_cuda.py b/tutorials/autotvm/tune_conv2d_cuda.py
index 09b56045edaf..ca2a93c34f18 100644
--- a/tutorials/autotvm/tune_conv2d_cuda.py
+++ b/tutorials/autotvm/tune_conv2d_cuda.py
@@ -45,6 +45,7 @@
 # Now return to python code. Import packages.
 
 import logging
+import os
 import sys
 import numpy as np
 
@@ -195,23 +196,32 @@ def conv2d_no_batching(N, H, W, CO, CI, KH, KW, stride, padding):
 # Begin tuning, log records to file `conv2d.log`
 # During tuning we will also try many invalid configs, so you are expected to
 # see many error reports. As long as you can see non-zero GFLOPS, it is okay.
+try:
+    curdir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "../../")
+except NameError:
+    # in sphinx
+    curdir = os.path.join(os.getcwd(), "../../")
+logdir = os.path.join(curdir, "docs/tutorials/autotvm")
+if not os.path.isdir(logdir):
+    os.makedirs(logdir)
+logfile = os.path.join(logdir, "conv2d.log")
 tuner = autotvm.tuner.XGBTuner(task)
 tuner.tune(n_trial=20,
            measure_option=measure_option,
-           callbacks=[autotvm.callback.log_to_file('conv2d.log')])
+           callbacks=[autotvm.callback.log_to_file(logfile)])
 
 #########################################################################
 # Finally we can inspect the best config from log file, check correctness,
 # and measure running time.
 
 # inspect the best config
-dispatch_context = autotvm.apply_history_best("conv2d.log")
+dispatch_context = autotvm.apply_history_best(logfile)
 best_config = dispatch_context.query(task.target, task.workload)
 print("\nBest config:")
 print(best_config)
 
 # apply history best from log file
-with autotvm.apply_history_best('conv2d.log'):
+with autotvm.apply_history_best(logfile):
     with tvm.target.create("cuda"):
         s, arg_bufs = conv2d_no_batching(N, H, W, CO, CI, KH, KW, strides, padding)
         func = tvm.build(s, arg_bufs)
diff --git a/tutorials/autotvm/tune_simple_template.py b/tutorials/autotvm/tune_simple_template.py
index b6ad7e94f883..ccc5d6ff0575 100644
--- a/tutorials/autotvm/tune_simple_template.py
+++ b/tutorials/autotvm/tune_simple_template.py
@@ -52,6 +52,7 @@
 
 import logging
 import sys
+import os
 
 import numpy as np
 import tvm
@@ -297,10 +298,19 @@ def matmul(N, L, M, dtype):
 
 # Begin tuning with RandomTuner, log records to file `matmul.log`
 # You can use alternatives like XGBTuner.
+try:
+    curdir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "../../")
+except NameError:
+    # in sphinx
+    curdir = os.path.join(os.getcwd(), "../../")
+logdir = os.path.join(curdir, "docs/tutorials/autotvm")
+if not os.path.isdir(logdir):
+    os.makedirs(logdir)
+logfile = os.path.join(logdir, "matmul.log")
 tuner = autotvm.tuner.RandomTuner(task)
 tuner.tune(n_trial=10,
            measure_option=measure_option,
-           callbacks=[autotvm.callback.log_to_file('matmul.log')])
+           callbacks=[autotvm.callback.log_to_file(logfile)])
 
 #########################################################################
 # Finally we apply history best from the cache file and check its correctness.
@@ -310,7 +320,7 @@ def matmul(N, L, M, dtype):
 # with the same argument.
 
 # apply history best from log file
-with autotvm.apply_history_best('matmul.log'):
+with autotvm.apply_history_best(logfile):
     with tvm.target.create("llvm"):
         s, arg_bufs = matmul(N, L, M, 'float32')
         func = tvm.build(s, arg_bufs)