ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1
1603 commits 1 branch 0 tags 37 MiB
Commit graph

1603 commits

This branch
This branch
All branches
Author SHA1 Message Date
Xianyan
de2a9e8dac Fix SmoothL1Criterion and SoftmaxWithCriterion (#1830) 2017-11-03 16:47:38 +08:00
Yang Wang
9b24d126c3 Add some operations (#1816) 
* add some ops

* add SquareDifference

* squaredifference

* add SquaredDifference test

* add tests

* add abs

* fix style

* fix style

* fix style

* fix tests

* fix style

* add div

* meet code review

* meet code review

* meet code review
 2017-11-03 09:55:51 +08:00
Hawkwood
7fb2dadba7 Fix cmul (#1822) 
* fix CMul

* 1697 issue

*  add exception capture and modify unit test
 2017-11-03 08:43:55 +08:00
Ian Wong
0bbcd78854 Support more tensorflow operations (#1818) 
* add more operations

* add more ops

* fix style issue

* fix style issue

* remove some jdk8 pai
 2017-11-02 15:01:19 +08:00
zhangxiaoli73
debbc93d92 optimize MSECriterion and Adam (#1701) 2017-11-02 10:28:48 +08:00
Jerry Wu
fd60a000f6 Caffe layer wise unit test (#1810) 
* add run caffe jobs

* fix style

* fix typo

* chmod

* add test info

* refinement

* fix import

* refinement

* refinement to get exit status

* fix typo

* fix path

* add layers unit tests

* add python interface

* fix style
 2017-11-01 15:37:25 +08:00
Yang Wang
a9a48ef47f support argmax operation (#1714) 2017-10-31 10:01:25 +08:00
Xin Qiu
6656112ec8 MSRA filler (#1802) 2017-10-31 10:00:11 +08:00
Xianyan
19c068e3c2 MapTable: Add module back in clearstate (#1778) 
* MapTable: Add module back in clearstate

* add unit test
 2017-10-27 14:19:38 +08:00
Xianyan
2d593c46a2 Clone graph keep their original prevs nodes and next nodes sequence (#1763) 
* Clone graph keep their original prevs nodes and next nodes sequence

* add comments

* update

* simplify table merge
 2017-10-27 11:24:40 +08:00
Ian Wong
f2018c6614 fix shape output tensor type from T to Int (#1737) 2017-10-25 20:36:48 +08:00
zhangxiaoli73
a5ae91c6d2 add toGraph (#1189) 
* add toGraph

* refactor toGraph

* add getEndNodes for ParallelTable, MapTable

* add unit test
 2017-10-23 10:45:34 +08:00
Yanzhang Wang
e8244efb6c fix: make the length of weight and gradWeight in quantized conv same length (#1692) 
* fix: the different length of weight and gradWeight in quantized conv is
very confused. So make them same length.

* fix: return grad
 2017-10-20 04:33:06 -04:00
Ian Wong
0cebe9d5f0 Support load more tensorflow operations (#1656) 
* Support load more tf operation

* refine session

* fix style issue

* meet code reivew
 2017-10-19 14:35:17 +08:00
zhangxiaoli73
c58cf4b348 add ptb model example (#1672) 
* add ptb model example

* meet review
 2017-10-19 14:24:36 +08:00
Yang Wang
a1508bc7f6 fix decode raw and StridedSlice and Transpose serilization issue (#1673) 2017-10-19 10:08:24 +08:00
Ian Wong
e72e2fae82 add more layers and operations (#1678) 
* add more layers and operations

* fix compile error

* fix compile error
 2017-10-19 09:56:14 +08:00
Jerry Wu
d48acf9ab9 Fix duplicated msg proto (#1659) 
* fix protobuf duplicate msgs

* fix test cases
 2017-10-18 09:56:11 +08:00
Yao Zhang
ec031c72a0 Add BroadcastGradientArgs (#1658) 
* finish BroadcastGradientArgs

* remove breeze

* pass the tests

* refine the function

* fix scala style

* while instead of for
 2017-10-16 17:29:30 +08:00
Xianyan
f4ccc9f754 move model freeze to AbstractModule (#1647) 
* move model freeze to container

* update

* update freeze API

* refactor freeze API

* meet code review and add unit test

* fix doc

* meet code review
 2017-10-15 22:28:35 -05:00
Xin Qiu
cd9c7b6cf7 SparseLinear SparseJoinTable DenseToSparse (#1652) 
* SparseLinear SparseJoinTable DenseToSparse

* Python api

* add DenseToSparseSpec

* update to upstream

* add some method

* meet code review

* fix python unit test

* fix python unit test
 2017-10-13 16:44:26 +08:00
zhangxiaoli73
7cbfdb88d4 throw clear exception in lookuptable (#1586) 2017-10-13 12:51:24 +08:00
dding3
e3b1518c41 RecurrentDecoder enhancement (#1619) 
* RecurrentDecoder refactor
 2017-10-12 22:03:22 -04:00
Yang Wang
5c66b437be serveral fix for preprocessing (#1609) 
* serveral fix for preprocessing

* load all

* fix tests

* more fixes

* meet code review

* fix style

* fix style

* fix tests

* fix tests

* rebase and fix api

* refine test case

* fix tests

* add transpose

* fix tests

* fix expandims

* fix serializer

* fix doc
 2017-10-12 15:49:35 +08:00
zhangxiaoli73
39113e9a2a add Variational Auto Encoder (#1432) 
* vae

* add python api and unit tests

* add docs
 2017-10-12 13:16:43 +08:00
Yao Zhang
2c4b9d83db Add tile operation (#1646) 
* Add tile operation

* reset core

* remove unnecessary comments
 2017-10-11 11:36:14 -05:00
Yanzhang Wang
1ddcaa63fc feat: quantize a whole graph/modules (#1618) 
* feat: quantize a whole graph/modules

* feat: python supports

* fix: delete unusage
 2017-10-11 03:16:48 -04:00
Yanzhang Wang
05283d3b60 feat: quantized dilated conv (#1617) 
* feat: quantized dilated conv

* fix: serialization api changes
 2017-10-10 07:43:28 -04:00
Yanzhang Wang
1f5c0a5701 feat: quantized linear (#1615) 
* feat: quantized linear

* fix: serialization api changes
 2017-10-10 06:45:17 -04:00
Yanzhang Wang
3470d5e205 feat: quantized convolution (#1614) 
* feat: quantized convolution

* fix: serialization api changes
 2017-10-10 06:02:05 -04:00
Yanzhang Wang
8bf29dc119 feat: serilization for quantized modules (#1613) 
* feat: serilization for quantized modules

All quantized modules extending QuantModule, which has an empty Tensor
for gradient. And the object mixes QuantSerializer for protobuf
 supporting.

* refactor: serialization api changes
 2017-10-10 04:21:36 -04:00
Jerry Wu
af0aaccf68 support graph (#1636) 2017-10-10 10:26:06 +08:00
Yanzhang Wang
7a5dfa350b feat: quantized tensor support (#1611) 
* feat: quantized tensor support

QuantizedTensor inherits from Tensor with supporting part of methods.
The memory is allocated from native code, so it only matains a
pointer `getNativeStorage`.

Note,

We don't support automatically `free` method. Anyone using this method
should think about the memory lacking, especially for memory
optimization.

test: quantized tensor tests

* fix: test case failed on spark 1.6
 2017-10-09 08:20:01 -04:00
Yanzhang Wang
2a23278691 feat: quantization methods (#1610) 
Quantization object support:

1. quantize a value with max and min.
2. quantize an array
3. quantize a Tensor[Float]

And for test, there're relative dequantize methods.
 2017-10-09 04:33:18 -04:00
Jerry Wu
99da09f0ed Tensor types support (#1630) 
* refinemnt to support more tensor types

* refinemnt

* refinemnt

* support object sharing

* support shared memory

* remove unused type

* refinement
 2017-10-09 15:14:31 +08:00
jenniew
b63bea22a6 fix MapTable issue 2017-10-05 17:59:03 -07:00
Ian Wong
34a3b0f5fa Refine scheduler (#1607) 
don't execute relevant nodes in forward/backward
fix a bug when handle const node
 2017-09-29 13:23:31 +08:00
dding3
da17fa7586 Support feed prior prediction back into recurrent (#1476) 
* Implement RecurrentDecoder
* add ut for lstm with recurrent decoder
 2017-09-27 13:42:18 -04:00
Ian Wong
46a7195bbe 1. throw exception when ConcatTable contains no submodule (#1598) 
2. rescursive remove stopGradient nodes
 2017-09-27 17:13:42 +08:00
Ian Wong
ce4f777a39 Refine tensorflow loader code strcuture (#1593) 
* revise tensorflow loaders

* refine more layers

* fix sytle issue

* fix style issue

* more ops

* meet code review

* fix style issue
 2017-09-27 16:54:11 +08:00
zhangxiaoli73
1e1466b33d refactor predictor (#1582) 
* refactor predictor

* predictClass share model output memory

* refactor repeatMemory to shareBuffer
 2017-09-26 14:48:56 +08:00
Yang Wang
423027c89d Support Tensorflow Preprocessing (#1572) 
* fix tensor bug

* preprocessing

* test

* fix tests

* refine

* fix tests

* fix style

* fix tests

* meet code review

* meet code review

* meet code review

* add doc
 2017-09-25 16:23:09 +08:00
Yao Zhang
e66d7f6e8f Add Realdiv, BiasAdd and Pad (#1579) 
* Add Pad

* refactor biasAdd
 2017-09-25 15:33:33 +08:00
Yao Zhang
08161f96ed Add Assign, Cast, ExpandDims, MaxPool operations (#1562) 
* add assign

* Add Cast and ExpandDims

* meet code review

* meet code reiew

* meet code review

* meet code review

* meet code review

* add comment to NumericWildcard type

* try to fix an AbstractMethod error in Cast method

* fix test failure
 2017-09-25 15:22:15 +08:00
Xin Qiu
aa31da4797 fix Logsoftmax infinity error (#1576) 2017-09-21 16:02:03 +08:00
Ian Wong
f7ea6315e5 Support load tf unet (#1575) 
* add deconv2d, resizebilinear and support broadcast in add

* fix test

* fix style error and meet code review

* fix python failure

* fix unit test
 2017-09-21 11:53:21 +08:00
Jerry Wu
956593ad02 Handle null array (#1571) 
* fix type comparision

* handle null value array

* fix type diff
 2017-09-21 10:16:16 +08:00
Yao Zhang
cc21eaf0cf Add slice and onehot (#1547) 
* Add slice and onehot

* refactor the code

* fit new code

* fix a typo

* meet code review

* fix test failed
 2017-09-20 21:47:37 +08:00
Yang Wang
4cb832c141 Add assert, image decoders, parseExample operation (#1559) 
* add some ops

* add parse example

* style

* add variable

* add png decoder

* add raw decoder

* remove resources

* fix ops

* test sig

* remove useless code

* remove import

* fix style

* fix conflict

* fix api
 2017-09-20 10:02:05 +08:00
Yang Wang
42fafc3ba0 support different type in tensorflow loader (#1569) 
* support different type in tensorflow

* meet code review

* meet code review

* fix tests
 2017-09-19 18:51:59 +08:00