gaspersic/triton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fixed bug

Browse Source
This commit is contained in:
Philippe Tillet
2019-07-05 15:07:20 -07:00
parent 88ebdddf3d
commit c666f71fd6
3 changed files with 31 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
examples/python/tensorflow/run.py
View File

@@ -61,30 +61,30 @@ def run_shift():
np.random.seed(2)
a = tf.placeholder(tf.float32, shape=[C, H, W, B])
b = tf.placeholder(tf.float32, shape=[C, F])
hshift_h = np.random.randint(- (R//2), R//2 + 1, size=C, dtype=np.int32)
hshift_w = np.random.randint(- (S//2), R//2 + 1, size=C, dtype=np.int32)
#hshift_h = np.random.randint(- (R//2), R//2 + 1, size=C, dtype=np.int32)
#hshift_w = np.random.randint(- (S//2), R//2 + 1, size=C, dtype=np.int32)
hshift_h = -1*np.ones(C, dtype=np.int32)
hshift_w = -1*np.ones(C, dtype=np.int32)
print(hshift_h)
print(hshift_w)
#hshift_h = np.ones(C, dtype=np.int32)
#hshift_w = np.ones(C, dtype=np.int32)
c = module.shift_conv(a, b, shift_h=tf.make_tensor_proto(hshift_h), shift_w=tf.make_tensor_proto(hshift_w))
c = tf.math.reduce_sum(c)
# Reference
ha = np.random.rand(C, H, W, B)
hb = np.random.rand(C, F)
ha = np.ones((C, H, W, B), dtype=np.float32)
hb = np.ones((C, F), dtype=np.float32)
#ha = np.ones((C, H, W, B), dtype=np.int32)
#hb = np.ones((C, F), dtype=np.int32)
sess = tf.InteractiveSession()
grads = tf.test.compute_gradient([a, b], [(C, H, W, B), (C, F)], c, (C, H, W, B),
grads = tf.test.compute_gradient([a, b], [(C, H, W, B), (C, F)], c, (1,),
extra_feed_dict={a: ha, b: hb})
dx_t, dx_n = grads[0]
dw_t, dw_n = grads[1]
print(dx_t)
print(dx_n)
#print(np.max(dw_t - dw_n))
#print(dw_t - dw_n)
#np.savetxt('diff.dat', dw_t - dw_n, fmt='%2.4f')
#np.savetxt('theoretical.dat', dw_t, fmt='%2.4f')
#np.savetxt('numerical.dat', dw_n, fmt='%2.4f')
print(np.max(dx_t - dx_n))
print(np.max(np.abs(dw_t - dw_n)))
print(np.max(np.abs(dx_t - dx_n)))
np.savetxt('diff.dat', dx_t - dx_n, fmt='%2.4f')
np.savetxt('theoretical.dat', dx_t, fmt='%2.4f')
np.savetxt('numerical.dat', dx_n, fmt='%2.4f')