Paddle

Commit Graph

Author	SHA1	Message	Date
Aurelius84	def27bc801	[Dy2stat]Fix bug with static_convert_var_shape in locals scope (#31556 ) * Fix bug with static_convert_var_shape * replace dot with dash	5 years ago
Aurelius84	f3959e9ddc	[save/load] Fix bug with input_spec=dict[InputSpec] in jit.save (#31517 ) * fix bug with jit.save * refine code	5 years ago
liym27	522c91ec67	[Dy2Stat] Remove gast.Index for compatibility of gast 0.4.0 (#31358 )	5 years ago
Huihuang Zheng	c40b98e068	Fix comment (#31424 ) Fix wrong code comment	5 years ago
Huihuang Zheng	6bf02a1261	[Dy2stat] Fix Read-Only Attribute as while_loop Output (#31415 ) Fix Read-Only Attribute as while_loop Output: Usually, our convert_while_loop will be like: ``` [a, b, c] = paddle.jit.dy2static.convert_while_loop( condition_name, body_name, [a, b, c]) ``` where a, b, c are in loop_var_names. However, if loop_var_names contains property such as foo.x, we cannot assign the attribute as output of convert_while_loop because Python property is a kind of read-only attribute. To handle the case, we replace the attributes which are output of convert_while_loop with generated variables, then if we know the attribute is not read-only at runtime, we assign the attribute. The created statements are like: ``` [a, b, __attribute_variable_1] = paddle.jit.dy2static.convert_while_loop( condition_name, body_name, [a, b, foo.x]) if not isinstance(getattr(type(foo), x, None), property): foo.x = __attribute_variable_1 ```	5 years ago
Aurelius84	1dd40870fc	[Dy2Stat] Fix eval_if_exist_else_none bug (#31261 ) * fix eval_if_exist_else_none bug * fix typo * fix typo * fix test_op_num unittest	5 years ago
chentianyu03	70131b475f	add warning message when dtypes of operator are not same (#31136 ) * add error msg when dtypes of operator are not same * add error msg when dtypes of operator are not same * change error msg to warning msg when dtypes of operator are not same * modify test case to fit for python2	5 years ago
Huihuang Zheng	cf43a321a8	[Dy2stat] Refactoring tensor_shape_transformer.py to Fix Change after Assign Bug (#31082 ) Problem In our old shape transformer logic, if user write: ``` s = tensor.shape ... y = paddle.some_api(s) ``` Dy2stat will change it to ``` ... y = paddle.some_api(convert_var_shape(tensor)) ``` However it will cause fatal bug if user changes the shape of `x` after assign. For example: ``` s = tensor.shape ... tensor = paddle.some_change_shape_api(tensor) ... y = paddle.some_api(s) ``` Then the Dy2stat will get wrong result because the code is translated into: ``` tensor = paddle.some_change_shape_api(tensor) ... y = paddle.some_api(convert_var_shape(tensor)) # tensor shape has been changed, not origin `s` value ``` Solution Logic It can not be solved in the old logic, so I refactoring tensor_shape_transformer logic. Now we will use `s` to store shape attribute and generate a var `s__STATIC_CONVERT_VAR_SHAPE_SUFFIX` to store static shape API `shape(tensor)` ``` s = tensor.shape ... y = paddle.some_api(s) ``` Dy2stat will change it to ``` s = tensor.shape s__STATIC_CONVERT_VAR_SHAPE_SUFFIX = shape(tensor) ... y = paddle.some_api(choose_shape_attr_or_api(s, s__STATIC_CONVERT_VAR_SHAPE_SUFFIX )) ``` In this case, the code is consistent with origin dygraph meaning and it fixed the change after assign bug. Code Key Note To help reviewers, the key change of this PR is changing `self.name_to_var_shape` from "mapping name to shape node" to "mapping name to its STATIC_CONVERT_VAR_SHAPE_SUFFIX name", then if a variable name has the SUFFIX, we can choose to use attribute shape or shape api. Other changes go with the key change. Consideration The issue of this PR is that we store extra static `shape` API result, will it harms the speed of Dy2stat? In some cases it will, but we argue that the benefit would be greater than the cost. 1. The extra calling to static `shape` API will happen when coder assign among shape variables. Take the following dygraph code as an instance: ``` s1 = tensor.shape s2 = s1 s3 = s2 ... ``` Then we called extra static `shape` APIs again and again, however users seldom write code like this. 2. If the shape variable is used a lot, for example: ``` s = tensor.shape y1 = paddle.some_api1(s) y2 = paddle.some_api2(s) y3 = paddle.some_api3(s) ``` Our old logic will create 3 shape APIs but now just 1. This is more common user code pattern. In fact, if reviewers take a look at the current unit test in this PR, you could see the op numbers decrease after this PR. So we argue that this PR can also improve speed in this code pattern.	5 years ago
Huihuang Zheng	ef627ac5b9	Fix that convert_var_shape doesn't support slice like [0:], test=develop (#31051 ) As the title, when slice_node like 1:3 being passed to idx of convert_var_shape, it will cause syntax error because a function cannot take this as argument. This PR fixed it.	5 years ago
ShenLiang	9401173e3a	Remove scale loss before reduce in dygraph (#30807 )	5 years ago
Huihuang Zheng	c137578341	Add Support for Tuple in for Loop (#30998 ) Dy2stat didn't support tuple as iteration variable in the past. This PR added there main cases: 1). Non-enumerate case: for var1, var2 in var\|var.numpy() will be re-written as: for FOR_ITER_TUPLE_PREFIX_x in var \| var.numpy(): var1 = FOR_ITER_TUPLE_PREFIX_x[0] var2 = FOR_ITER_TUPLE_PREFIX_x[1] 2). Enumerate out tuple case: for t in enumerate(var\|var.numpy) will be rewritten as: for FOR_ITER_TUPLE_INDEX_PREFIX_x, FOR_ITER_TUPLE_PREFIX_x in enumerate(var\|var.numpy): t = (FOR_ITER_TUPLE_INDEX_PREFIX_x, FOR_ITER_TUPLE_PREFIX_x) 3). Enumerate inner tuple case: for i, (var1, (var2, va3)) in enumerate(var\|var.numpy()) will be re-written as: for i, FOR_ITER_TUPLE_PREFIX_x in var \| var.numpy(): var1 = FOR_ITER_TUPLE_PREFIX_x[0] var2 = FOR_ITER_TUPLE_PREFIX_x[1][0] var3 = FOR_ITER_TUPLE_PREFIX_x[1][1]	5 years ago
wanghuancoder	823f499a8a	fix a bug of Sequential::__getitem__ (#30899 ) * fix a bug of Sequential::__getitem__, test=develop * add testcase, test=develop	5 years ago
wanghuancoder	90d92111cf	let LayerList could add [None], test=develop (#30911 )	5 years ago
AshburnLee	666efc2336	Call new cudnn batch norm API regardless of data type and data layout (#30157 )	5 years ago
WangXi	b1026f64af	【kunlun】dygraph supports multi xpu card training (#30671 )	5 years ago
liym27	13ef444fa6	[Dy2Stat] Fix error message when the message has more than one lines. (#30714 )	5 years ago
Leo Chen	1a13626f5f	polish printing dtype (#30682 ) * polish printing dtype * fix special case	5 years ago
wanghuancoder	27a5c0cff6	fix layers train eval bug (#30580 ) * delete empty line of pybing.cc, test=develop * fix layers train eval bug, test=develop	5 years ago
Aurelius84	621bc4f771	[Dy2static]Fix paddle prefix in is_paddle_api (#30569 ) * add paddle. * add unittest	5 years ago
Aurelius84	5067e3a8d2	[Dy2Static]Enhance check of TracedLayers out vars (#30576 )	5 years ago
Chen Weihang	ae1f32091a	fix prune input bug (#30384 )	5 years ago
Leo Chen	3d015f1cf5	Set expected place in child thread for dataloader to avoid costing cuda memory on other card (#30338 ) * set expected place in child thread for dataloader * set device id when set tensor from numpy * revert tensor_py change * add compile guard * fix ci * fix bug	5 years ago
ShenLiang	a60f17b89d	Support unused parameters in dynamic graph distributed (#30224 )	5 years ago
WeiXin	edafb5465a	Fix bug for 'save mutiple method' (#30218 ) * Fix bug for 'save mutiple method' * To pass coverage. * edit code to pass coverage. * edit code to pass coverage. * add unittest for coverage. * change for coverage. * edit for coverage.	5 years ago
Huihuang Zheng	c372a76303	Add Static Variable Clone (#30208 ) Add clone method for static Variable so that this interface will be same as dygraph. It fixed some bugs in dy2stat	5 years ago
XiaoguangHu	6bfdef727e	clean redundant API alias in 2.0 - part 2 (#30013 ) * delete paddle.nn.functional.assign * fix dynamic to static error	5 years ago
Aurelius84	03e072736e	Skip convert tensor shape while using Paddle.shape (#30223 ) * fix tensor shape bug * fix op_num * clean code	5 years ago
liym27	31ed9a5ed3	[Dy2Stat] Use Paddle2.0 api paddle.tensor.array_* (#30156 )	5 years ago
liym27	ad55f609d5	[Dy2Stat] Don't convert to paddle.shape if var_x.shape is not negetive (#29965 ) 1. When x is Variable, call nn.shape(x) only in following cases: 1）The shape of x is used in control flow condition. 2）The dim to be used is negetive 2. When x is Variable, but x.shape or x.shape[idx] doesn't contain negetive value, don't convert to paddle.shape()	5 years ago
wangchaochaohu	7dd551e08b	refine the paddle place support using str (#28769 )	5 years ago
Zhou Wei	30888ca343	Polish and Optimize the print/repr information of Layer (#29998 ) * Polish and Optimize the print/repr message of all layer * fix some code format	5 years ago
lilong12	2bc5121da8	add the paddle.distributed.split api (#29970 ) * add distributed.split, test=develop	5 years ago
wawltor	587b67ef62	fix the state_dict bug for the xpu (#29888 ) fix the state_dict bug for the xpu	5 years ago
ShenLiang	01e2874a0e	Support multi-stream communication for dynamic graph distributed (#29525 ) * fix fleet for multi-stream * fix memcpy for ncclid * use sync to solve move operation	5 years ago
liym27	41a7b07159	[Dy2Stat] Fix bug for loop: a variable is used and created in loop, but used before created (#29769 )	5 years ago
Huihuang Zheng	dfffee8a5d	[Dy2stat] Enable jit.save to Save Without Running (#29579 ) Enable jit.save to Save Without Running.	5 years ago
liym27	a0b60716f1	[Dy2Stat] Support grammar: for ele in var[idx] (#29541 ) Support to transformfor ele in var stms in which var is a slice of Tensor.	5 years ago
liym27	0cad1152f4	[Dy2Stat] 1. Fix bug of for-range stmts. 2. Support that step value is negative in for-range stmts (#29519 ) 1. Fix error in _build_cond_stmt of for-range stmts. 2. Support that step value is negative in for-range stmts 3. Fix code because of the diff between Py2 and Py3	5 years ago
Chen Weihang	c1a26e2a05	fix train eval set error in static mode (#29540 )	5 years ago
Leo Chen	0fdd365665	Add fast path for dropout when p == 0 (#29553 ) * add fast path for p == 0 in dropout * add ut	5 years ago
huangxu96	4001979309	Add ReserveSpace in dygraph batch_norm. (#29221 ) * Add ReserveSpace in dygraph batch_norm. * Add unittest for reservespace	5 years ago
Zhou Wei	e74e1a226c	support deepcopy for Layer/Tensor/Paramerbase (#29387 ) * support deepcopy for Layer/Tensor/Paramerbase * fix some code	5 years ago
ShenLiang	2ef9e0e23c	Rebuild group automatically in dynamic graph distributed (#29255 ) * add tensor_indices in AssignGroupBySize * add rebuild group in reducer	5 years ago
ShenLiang	311b3b44fc	Fix the bug where embedding can‘t be processed correctly in reducer (#29485 ) * fix the bug of reducer in embedding * add comment	5 years ago
chentianyu03	64e4e17f0c	remove complexvariable (#29390 ) * rm complexvariable * modify test_var_base unittest * remove duplicated codes	5 years ago
ShenLiang	4064354a01	support dp run single card (#29358 )	5 years ago
Huihuang Zheng	8f7627907c	[Dy2stat] Reduce Exception Type for Better Error Message (#29268 ) Reduce exception type so that if covert_to_static failed, it reports right error message.	5 years ago
liym27	61a8f2874f	[Dy2Stat] Fix bug: Do not use gast.Subscript to replace gast.Name in when transforming for_enumerate_loop (#29310 )	5 years ago
Chen Weihang	9ad800ebb2	Support type promote for basic math ops (quantum required) (#29265 ) * basic impl of type promote * add comment & another testcase * fix complex bugs & support python op promote type * fix failed unittests & polish code * add unittest for coverage * change to only promote complex type * polish code details * polish several comments	5 years ago
Leo Chen	b58cfff89d	use has_grad instead of train_mode (#29309 ) * use has_grad instead of train_mode * add vlog for debug * fix ut * fix ut	5 years ago

1 2 3 4 5 ...

504 Commits (test_benchmark_ci)