name | about | labels |
---|---|---|
Bug Report | Use this template for reporting a bug | kind/bug |
GPU环境下图模式偶现1d场景,float16类型出现精度问题
Ascend
/GPU
/CPU
) / 硬件环境:Please delete the backend not involved / 请删除不涉及的后端:
GPU
Software Environment / 软件环境 (Mandatory / 必填):
-- MindSpore version (e.g., 1.7.0.Bxxx) :
-- Python version (e.g., Python 3.7.5) :
-- OS platform and distribution (e.g., Linux Ubuntu 16.04):
-- GCC/Compiler version (if compiled from source):
Excute Mode / 执行模式 (Mandatory / 必填)(PyNative
/Graph
):
Please delete the mode not involved / 请删除不涉及的模式:
/mode graph
test_p_applyftrl_1d_float16 __________________________________________________________________________________________
@Level2
@SKIP_ENV_DAVINCI_EXECUTOR(reason='issue=I5O1EN,算子在ascend端,发现必现问题,报错显示标杆期望值显示为0,但是实际是存在的')
def test_p_applyftrl_1d_float16():
var = Parameter(Tensor(np.random.randn(8, ).astype(np.float16)), name="var")
accum = Parameter(Tensor(np.random.randn(8, ).astype(np.float16)), name="accum")
linear = Parameter(Tensor(np.random.randn(8, ).astype(np.float16)), name="linear")
grad = Tensor(np.random.randn(8, ).astype(np.float16))
lr = Tensor(0.001).astype(np.float16)
l1 = Tensor(0.0).astype(np.float16)
l2 = Tensor(0.0).astype(np.float16)
lr_power = Tensor(-0.5).astype(np.float16)
fact = ApplyFtrlMock(attributes={'use_locking': False},
inputs=[var, accum, linear, grad, lr, l1, l2, lr_power])
if os.environ['CONTEXT_DEVICE_TARGET'] == 'CPU':
fact.forward_cmp()
else:
> fact.forward_cmp_1()
testcase pass
../share/ops/primitive/applyftrl_ops.py:160: in forward_cmp_1
allclose_nparray(py_real, ms_real, self.loss, self.loss)
../share/utils.py:31: in allclose_nparray
_count_unequal_element(data_expected, data_me, rtol, atol)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
data_expected = array([ 0.1814 , -0.006058, 0. , 0.02211 , -0.1737 , -0.10864 ,
-0.597 , 1.598 ], dtype=float16)
data_me = array([ 0.1816 , -0.006058, 0. , 0.02211 , -0.1736 , -0.10864 ,
-0.597 , 1.601 ], dtype=float16), rtol = 0.001, atol = 0.001
def _count_unequal_element(data_expected, data_me, rtol, atol):
assert data_expected.shape == data_me.shape
total_count = len(data_expected.flatten())
error = np.abs(data_expected - data_me)
greater = np.greater(error, atol + np.abs(data_me) * rtol)
loss_count = np.count_nonzero(greater)
assert (loss_count / total_count) < rtol, \
"\ndata_expected_std:{0}\ndata_me_error:{1}\nloss:{2}". \
> format(data_expected[greater], data_me[greater], error[greater])
E AssertionError:
E data_expected_std:[1.598]
E data_me_error:[1.601]
E loss:[0.00293]
../share/utils.py:24: AssertionError
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。
Please add labels (comp or sig), also you can visit https://gitee.com/mindspore/community/blob/master/sigs/dx/docs/labels.md to find more.
为了让代码尽快被审核,请您为Pull Request打上 组件(comp)或兴趣组(sig) 标签,打上标签的PR可直接推送给责任人进行审核。
更多的标签可以查看https://gitee.com/mindspore/community/blob/master/sigs/dx/docs/labels.md
以组件相关代码提交为例,如果你提交的是data组件代码,你可以这样评论:
//comp/data
当然你也可以邀请data SIG组来审核代码,可以这样写:
//sig/data
另外你还可以给这个PR标记类型,例如是bugfix或者是特性需求:
//kind/bug or //kind/feature
恭喜你,你已经学会了使用命令来打标签,接下来就在下面的评论里打上标签吧!
ApplyFtrl算子改造前就有float16类型的,实测改造前已存在精度问题。频率为每1000次3次精度误差。
ApplyFtrl 算子GPU上的历史问题。
最新master上的包未复现问题,待复现问题,若持续观察未复现按正常流程回归。
未复现该场景问题,目前所有用例已pass
登录 后才可以发表评论