X2Paddle tags
https://gitcode.net/paddlepaddle/X2Paddle/-/tags
2021-09-30T11:08:45+08:00
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.4.1
v1.4.1
<h2 data-sourcepos="1:1-1:17" dir="auto">
<a id="user-content-whats-changed" class="anchor" href="#whats-changed" aria-hidden="true"></a>What's Changed</h2>
<ul data-sourcepos="2:1-5:0" dir="auto">
<li data-sourcepos="2:1-2:93">[Bug]Fixed ONNXDecoder by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/939" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/939</a>
</li>
<li data-sourcepos="3:1-3:96">Add ONNX2Paddle for visualdl by @rainyfly in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/954" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/954</a>
</li>
<li data-sourcepos="4:1-5:0">[Bug]Support YOLOv8-cls by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/952" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/952</a>
</li>
</ul>
<h2 data-sourcepos="6:1-6:19" dir="auto">
<a id="user-content-new-contributors" class="anchor" href="#new-contributors" aria-hidden="true"></a>New Contributors</h2>
<ul data-sourcepos="7:1-7:94" dir="auto">
<li data-sourcepos="7:1-7:94">@rainyfly made their first contribution in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/954" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/954</a>
</li>
</ul>
2023-04-02T00:02:07+08:00
wjj19950828
wjjisloser@163.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.4.0
v1.4.0
<h2 data-sourcepos="1:1-1:17" dir="auto">
<a id="user-content-whats-changed" class="anchor" href="#whats-changed" aria-hidden="true"></a>What's Changed</h2>
<ul data-sourcepos="2:1-13:0" dir="auto">
<li data-sourcepos="2:1-2:108">Add input_shape_dict and rm raw_input by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/899" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/899</a>
</li>
<li data-sourcepos="3:1-3:116">Fixed ToPILImage && rm SymbolicShapeInference by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/903" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/903</a>
</li>
<li data-sourcepos="4:1-4:85">Support GFPGAN by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/906" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/906</a>
</li>
<li data-sourcepos="5:1-5:113">Add FastDeploy YOLOv5/v6/v7 demo in readme by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/911" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/911</a>
</li>
<li data-sourcepos="6:1-6:113">Updated torch.nn.utils.clip_grad_value_.md by @PommesPeter in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/918" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/918</a>
</li>
<li data-sourcepos="7:1-7:96">[Bug] Fixed ONNX Gemm bug by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/917" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/917</a>
</li>
<li data-sourcepos="8:1-8:93">[Bug] Fixed typo error by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/920" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/920</a>
</li>
<li data-sourcepos="9:1-9:95">Support fmod=1 in Mod OP by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/921" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/921</a>
</li>
<li data-sourcepos="10:1-10:88">Fixed NonZero bug by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/922" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/922</a>
</li>
<li data-sourcepos="11:1-11:78">add optim by @AndPuQing in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/927" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/927</a>
</li>
<li data-sourcepos="12:1-13:0">Support YOLOv8 for dynamic shape by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/931" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/931</a>
</li>
</ul>
<h2 data-sourcepos="14:1-14:19" dir="auto">
<a id="user-content-new-contributors" class="anchor" href="#new-contributors" aria-hidden="true"></a>New Contributors</h2>
<ul data-sourcepos="15:1-16:95" dir="auto">
<li data-sourcepos="15:1-15:97">@PommesPeter made their first contribution in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/918" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/918</a>
</li>
<li data-sourcepos="16:1-16:95">@AndPuQing made their first contribution in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/927" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/927</a>
</li>
</ul>
2023-01-19T05:16:32+08:00
wjj19950828
wjjisloser@163.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.3.9
v1.3.9
<h2 data-sourcepos="1:1-1:17" dir="auto">
<a id="user-content-whats-changed" class="anchor" href="#whats-changed" aria-hidden="true"></a>What's Changed</h2>
<ul data-sourcepos="2:1-17:109" dir="auto">
<li data-sourcepos="2:1-2:102">Support 7-15 ONNX opset version by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/838" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/838</a>
</li>
<li data-sourcepos="3:1-3:112">Modify opset7~15 inheritance relationship by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/843" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/843</a>
</li>
<li data-sourcepos="4:1-4:88">Add autoscan test by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/836" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/836</a>
</li>
<li data-sourcepos="5:1-5:108">Add autoscan test for Elementwise ops by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/854" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/854</a>
</li>
<li data-sourcepos="6:1-6:85">Add_test_abs by @qqj1130247885 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/859" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/859</a>
</li>
<li data-sourcepos="7:1-7:98">Add logical and compare ops by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/855" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/855</a>
</li>
<li data-sourcepos="8:1-8:86">Fixed torch pad by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/863" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/863</a>
</li>
<li data-sourcepos="9:1-9:97">Solve the overload problem by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/865" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/865</a>
</li>
<li data-sourcepos="10:1-10:92">Add AveragePool tests by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/872" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/872</a>
</li>
<li data-sourcepos="11:1-11:103">Add Reduce_ops and HardSigmoid by @qqj1130247885 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/867" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/867</a>
</li>
<li data-sourcepos="12:1-12:83">Add isinf by @qqj1130247885 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/868" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/868</a>
</li>
<li data-sourcepos="13:1-13:82">Add IsNaN by @qqj1130247885 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/869" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/869</a>
</li>
<li data-sourcepos="14:1-14:80">Add Sum by @qqj1130247885 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/873" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/873</a>
</li>
<li data-sourcepos="15:1-15:83">Fixed Readme by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/879" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/879</a>
</li>
<li data-sourcepos="16:1-16:93">Fixed reduce tests bug by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/880" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/880</a>
</li>
<li data-sourcepos="17:1-17:109">Set enable_code_optim default to False by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/889" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/889</a>
</li>
</ul>
2022-09-20T13:17:01+08:00
wjj19950828
wjjisloser@163.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.3.8
v1.3.8
<h2 data-sourcepos="1:1-1:17" dir="auto">
<a id="user-content-whats-changed" class="anchor" href="#whats-changed" aria-hidden="true"></a>What's Changed</h2>
<ul data-sourcepos="2:1-11:0" dir="auto">
<li data-sourcepos="2:1-2:112">Add HuggingFace Model Conversion Tutorial by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/823" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/823</a>
</li>
<li data-sourcepos="3:1-3:101">Add aten::upsample_trilinear3d by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/817" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/817</a>
</li>
<li data-sourcepos="4:1-4:85">Fixed misspell by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/821" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/821</a>
</li>
<li data-sourcepos="5:1-5:95">Add sysmpy in requirment by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/824" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/824</a>
</li>
<li data-sourcepos="6:1-6:85">Fixed Misspell by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/829" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/829</a>
</li>
<li data-sourcepos="7:1-7:87">Fixed rename bug by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/820" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/820</a>
</li>
<li data-sourcepos="8:1-8:91">Fix ONNX resize op by @qqj1130247885 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/830" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/830</a>
</li>
<li data-sourcepos="9:1-9:131">Fix bug in docs directory about code combination of save_image by @ZoruaSama in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/833" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/833</a>
</li>
<li data-sourcepos="10:1-11:0">Remove scale for add by @wjj19950828 in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/835" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/835</a>
</li>
</ul>
<h2 data-sourcepos="12:1-12:19" dir="auto">
<a id="user-content-new-contributors" class="anchor" href="#new-contributors" aria-hidden="true"></a>New Contributors</h2>
<ul data-sourcepos="13:1-15:0" dir="auto">
<li data-sourcepos="13:1-13:99">@qqj1130247885 made their first contribution in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/830" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/830</a>
</li>
<li data-sourcepos="14:1-15:0">@ZoruaSama made their first contribution in <a href="https://github.com/PaddlePaddle/X2Paddle/pull/833" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/833</a>
</li>
</ul>
<p data-sourcepos="16:1-16:84" dir="auto"><strong>Full Changelog</strong>: <a href="https://github.com/PaddlePaddle/X2Paddle/compare/v1.3.7...v1.3.8" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/compare/v1.3.7...v1.3.8</a></p>
2022-08-18T17:17:56+08:00
wjj19950828
wjjisloser@163.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.3.7
v1.3.7
<h2 data-sourcepos="1:1-1:15" dir="auto">
<a id="user-content-新增功能" class="anchor" href="#%E6%96%B0%E5%A2%9E%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>新增功能</h2>
<ul data-sourcepos="2:1-5:0" dir="auto">
<li data-sourcepos="2:1-2:147">全面支持 HuggingFace 模型,包括Bert、Albert等热点模型以及对应不同任务类型,共55种 TorchScript 以及40种 ONNX 模型</li>
<li data-sourcepos="3:1-3:129">针对 Bert结构 ONNX模型,新增 Gelu pass,预测速度提升20%以上 <a href="https://github.com/PaddlePaddle/X2Paddle/pull/809" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/809</a>
</li>
<li data-sourcepos="4:1-5:0">针对 Bert结构 ONNX模型,新增 div2scale pass,使其命中PaddleInference multihead_matmul_fuse_pass,性能提升30%以上 <a href="https://github.com/PaddlePaddle/X2Paddle/pull/813" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/813</a>
</li>
</ul>
<h2 data-sourcepos="6:1-6:21" dir="auto">
<a id="user-content-新增模型支持" class="anchor" href="#%E6%96%B0%E5%A2%9E%E6%A8%A1%E5%9E%8B%E6%94%AF%E6%8C%81" aria-hidden="true"></a>新增模型支持</h2>
<h3 data-sourcepos="7:1-7:19" dir="auto">
<a id="user-content-pytorch-paddle" class="anchor" href="#pytorch-paddle" aria-hidden="true"></a>PyTorch->Paddle</h3>
<ul data-sourcepos="8:1-14:0" dir="auto">
<li data-sourcepos="8:1-8:49"><a href="https://arxiv.org/abs/1905.02244" rel="nofollow noreferrer noopener" target="_blank">MobileNetv3</a></li>
<li data-sourcepos="9:1-9:94"><a href="https://huggingface.co/docs/transformers/model_doc/t5#transformers.T5Model" rel="nofollow noreferrer noopener" target="_blank">HuggingFace T5</a></li>
<li data-sourcepos="10:1-10:106"><a href="https://huggingface.co/docs/transformers/model_doc/marian#transformers.MarianModel" rel="nofollow noreferrer noopener" target="_blank">HuggingFace Marian</a></li>
<li data-sourcepos="11:1-11:100"><a href="https://huggingface.co/docs/transformers/model_doc/bart#transformers.BartModel" rel="nofollow noreferrer noopener" target="_blank">HuggingFace Bart</a></li>
<li data-sourcepos="12:1-12:112"><a href="https://huggingface.co/docs/transformers/model_doc/wav2vec2#transformers.Wav2Vec2Model" rel="nofollow noreferrer noopener" target="_blank">HuggingFace wav2vec2</a></li>
<li data-sourcepos="13:1-14:0"><a href="https://huggingface.co/docs/transformers/model_doc/big_bird#transformers.BigBirdModel" rel="nofollow noreferrer noopener" target="_blank">HuggingFace BigBird</a></li>
</ul>
<h3 data-sourcepos="15:1-15:16" dir="auto">
<a id="user-content-onnx-paddle" class="anchor" href="#onnx-paddle" aria-hidden="true"></a>ONNX->Paddle</h3>
<ul data-sourcepos="16:1-22:0" dir="auto">
<li data-sourcepos="16:1-16:109"><a href="https://huggingface.co/docs/transformers/model_doc/roberta#transformers.RobertaModel" rel="nofollow noreferrer noopener" target="_blank">HuggingFace Roberta</a></li>
<li data-sourcepos="17:1-17:120"><a href="https://huggingface.co/docs/transformers/model_doc/xlm-roberta#transformers.XLMRobertaModel" rel="nofollow noreferrer noopener" target="_blank">HuggingFace XLM-Roberta</a></li>
<li data-sourcepos="18:1-18:94"><a href="https://huggingface.co/docs/transformers/model_doc/t5#transformers.T5Model" rel="nofollow noreferrer noopener" target="_blank">HuggingFace T5</a></li>
<li data-sourcepos="19:1-19:100"><a href="https://huggingface.co/docs/transformers/model_doc/gpt2#transformers.GPT2Model" rel="nofollow noreferrer noopener" target="_blank">HuggingFace GPT2</a></li>
<li data-sourcepos="20:1-22:0"><a href="https://huggingface.co/docs/transformers/model_doc/marian#transformers.MarianModel" rel="nofollow noreferrer noopener" target="_blank">HuggingFace Marian</a></li>
</ul>
<h2 data-sourcepos="23:1-23:11" dir="auto">
<a id="user-content-新增op" class="anchor" href="#%E6%96%B0%E5%A2%9Eop" aria-hidden="true"></a>新增op</h2>
<ul data-sourcepos="24:1-25:161" dir="auto">
<li data-sourcepos="24:1-24:486">新增PyTorch op(7个):<code>aten::rsqrt</code>、<code>aten::replication_pad1d</code>、<code>aten::full</code>、 <code>aten::group_norm</code>、<code>aten::argmax</code>、<code>aten::copy</code>、<code>aten::linear</code> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/783" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/783</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/781" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/781</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/786" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/786</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/791" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/791</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/789" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/789</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/769" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/769</a>
</li>
<li data-sourcepos="25:1-25:161">新增ONNX op(2个):<code>CumSum </code>、<code>LessOrEqual</code> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/793" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/793</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/794" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/794</a>
</li>
</ul>
<h2 data-sourcepos="26:1-26:11" dir="auto">
<a id="user-content-修复op" class="anchor" href="#%E4%BF%AE%E5%A4%8Dop" aria-hidden="true"></a>修复op</h2>
<ul data-sourcepos="27:1-29:0" dir="auto">
<li data-sourcepos="27:1-27:174">修复PyTorch op(2个):<code>aten::masked_fill</code>、<code>aten::index</code> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/786" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/786</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/787" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/787</a>
</li>
<li data-sourcepos="28:1-29:0">修复ONNX op(12个):<code>ConvTranspose </code>、<code>Unsqueeze </code>、<code>MatMul </code>、<code>Gather </code>、<code>Slice </code>、<code>Squeeze </code>、<code>Flatten</code>、<code>Constant </code>、<code>ConstantOfShape </code>、<code>Where </code>、<code>Topk</code>、<code>NonZero </code> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/792" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/792</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/794" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/794</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/796" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/796</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/797" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/797</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/805" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/805</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/808" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/808</a>
</li>
</ul>
<h2 data-sourcepos="30:1-30:15" dir="auto">
<a id="user-content-修复功能" class="anchor" href="#%E4%BF%AE%E5%A4%8D%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>修复功能</h2>
<ul data-sourcepos="31:1-33:0" dir="auto">
<li data-sourcepos="31:1-31:71">修复LN fuser pass <a href="https://github.com/PaddlePaddle/X2Paddle/pull/777" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/777</a>
</li>
<li data-sourcepos="32:1-33:0">修复Linear fuser pass <a href="https://github.com/PaddlePaddle/X2Paddle/pull/800" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/800</a>
</li>
</ul>
<h2 data-sourcepos="34:1-34:9" dir="auto">
<a id="user-content-其他" class="anchor" href="#%E5%85%B6%E4%BB%96" aria-hidden="true"></a>其他</h2>
<ul data-sourcepos="35:1-36:82" dir="auto">
<li data-sourcepos="35:1-35:91">在转换完成之后添加问卷链接 <a href="https://github.com/PaddlePaddle/X2Paddle/pull/801" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/801</a>
</li>
<li data-sourcepos="36:1-36:82">清除<code>paddle.fluid</code>接口 <a href="https://github.com/PaddlePaddle/X2Paddle/pull/806" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/806</a>
</li>
</ul>
2022-07-16T06:01:39+08:00
wjj19950828
wjjisloser@163.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.3.6
v1.3.6
<h2 dir="auto">
<a id="user-content-新增功能" class="anchor" href="#%E6%96%B0%E5%A2%9E%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>新增功能</h2>
<ul dir="auto">
<li>新增转换记录统计功能 <a href="https://github.com/PaddlePaddle/X2Paddle/pull/766" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/766</a>
</li>
<li>针对 ONNX Transformer 模型,新增 LayerNorm pass,优化预测速度 <a href="https://github.com/PaddlePaddle/X2Paddle/pull/761" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/761</a>
</li>
<li>HuggingFace 参数迁移功能 <a href="https://github.com/PaddlePaddle/X2Paddle/pull/748" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/748</a>
</li>
</ul>
<h2 dir="auto">
<a id="user-content-新增模型支持" class="anchor" href="#%E6%96%B0%E5%A2%9E%E6%A8%A1%E5%9E%8B%E6%94%AF%E6%8C%81" aria-hidden="true"></a>新增模型支持</h2>
<h3 dir="auto">
<a id="user-content-pytorch-paddle" class="anchor" href="#pytorch-paddle" aria-hidden="true"></a>PyTorch->Paddle</h3>
<ul dir="auto">
<li>图像修复算法:<a href="https://github.com/saic-mdal/lama" rel="nofollow noreferrer noopener" target="_blank">Saicinpainting_LaMa</a>
</li>
</ul>
<h3 dir="auto">
<a id="user-content-onnx-paddle" class="anchor" href="#onnx-paddle" aria-hidden="true"></a>ONNX->Paddle</h3>
<ul dir="auto">
<li>SOTA backbone:<a href="https://github.com/open-mmlab/mmdetection/tree/master/configs/swin" rel="nofollow noreferrer noopener" target="_blank">SwinTransformer</a>
</li>
<li>关键点检测算法:<a href="https://arxiv.org/abs/2006.10204" rel="nofollow noreferrer noopener" target="_blank">Blazepose</a>
</li>
</ul>
<h3 dir="auto">
<a id="user-content-caffe-paddle" class="anchor" href="#caffe-paddle" aria-hidden="true"></a>Caffe->Paddle</h3>
<ul dir="auto">
<li>视频分割算法:<a href="https://github.com/JingchunCheng/SegFlow" rel="nofollow noreferrer noopener" target="_blank">SegFlow</a>
</li>
</ul>
<h3 dir="auto">
<a id="user-content-tensorflow-paddle" class="anchor" href="#tensorflow-paddle" aria-hidden="true"></a>TensorFlow->Paddle</h3>
<ul dir="auto">
<li>图像生成算法:<a href="https://github.com/affinelayer/pix2pix-tensorflow" rel="nofollow noreferrer noopener" target="_blank">Pix2Pix</a>
</li>
</ul>
<h2 dir="auto">
<a id="user-content-新增op" class="anchor" href="#%E6%96%B0%E5%A2%9Eop" aria-hidden="true"></a>新增op</h2>
<ul dir="auto">
<li>新增PyTorch op(5个):<code>aten::complex</code>、<code>aten::real</code>、<code>aten::imag</code>、<code>aten::fft_rfftn</code>、<code>aten::fft_irfftn</code> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/729" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/729</a>
</li>
<li>新增ONNX op(5个):<code>Sin</code>、<code>Cos</code>、<code>Neg</code>、<code>SpaceToDepth</code>、<code>GatherElement</code> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/723" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/723</a>
</li>
<li>新增PyTorch-Paddle api(2个):<code>gather</code>、<code>scatter</code> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/726" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/726</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/730" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/730</a>
</li>
</ul>
<h2 dir="auto">
<a id="user-content-修复op" class="anchor" href="#%E4%BF%AE%E5%A4%8Dop" aria-hidden="true"></a>修复op</h2>
<ul dir="auto">
<li>修复PyTorch op(1个):<code>aten::_convolution</code> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/725" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/725</a>
</li>
<li>修复ONNX op(7个):<code>pad</code>、<code>resize</code>、<code>Upsample </code>、<code>Gather </code>、<code>Slice </code> 、<code>convtranspose</code> 、<code>Split</code> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/716" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/716</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/720" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/720</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/722" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/722</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/723" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/723</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/749" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/749</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/751" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/751</a>
</li>
<li>修复Caffe op(3个):<code>scale</code>、<code>crop</code>、<code>Pow</code> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/721" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/721</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/757" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/757</a>
</li>
<li>修复TensorFlow(3个):<code>FusedBatchNorm</code>、<code>SpaceToBatchND</code>、<code>BatchToSpaceND</code> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/738" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/738</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/741" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/741</a>
</li>
</ul>
<h2 dir="auto">
<a id="user-content-修复功能" class="anchor" href="#%E4%BF%AE%E5%A4%8D%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>修复功能</h2>
<ul dir="auto">
<li>优化转换过程中的 log 提示 <a href="https://github.com/PaddlePaddle/X2Paddle/pull/764" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/764</a>
</li>
<li>更新外部框架一键转换为 Paddle Lite 模型文档,新增<code>lite_valid_places</code>选择类型 <a href="https://github.com/PaddlePaddle/X2Paddle/pull/758" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/758</a>
</li>
<li>PyTorch2Paddle 动态shape转换指引 <a href="https://github.com/PaddlePaddle/X2Paddle/pull/732" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/732</a>
</li>
</ul>
<h2 dir="auto">
<a id="user-content-contributors" class="anchor" href="#contributors" aria-hidden="true"></a>Contributors</h2>
<ul dir="auto">
<li>Thanks to external developers for their contributions to open source projects!@geoyee <a href="/HUSTLX" data-user="662687" data-reference-type="user" data-container="body" data-placement="top" class="gfm gfm-project_member js-user-link" title="hustlx">@HUSTLX</a>
</li>
</ul>
2022-04-11T06:33:05+08:00
wjj19950828
wjjisloser@163.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.3.5
v1.3.5
<h2 data-sourcepos="1:1-1:15" dir="auto">
<a id="user-content-新增功能" class="anchor" href="#%E6%96%B0%E5%A2%9E%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>新增功能</h2>
<ul data-sourcepos="2:1-2:150" dir="auto">
<li data-sourcepos="2:1-2:150">添加enable_code_optim开关,解决部分场景下代码优化可能导致转换出错的问题 <a href="https://github.com/PaddlePaddle/X2Paddle/pull/696" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/696</a>
</li>
</ul>
<h2 data-sourcepos="3:1-3:21" dir="auto">
<a id="user-content-新增模型支持" class="anchor" href="#%E6%96%B0%E5%A2%9E%E6%A8%A1%E5%9E%8B%E6%94%AF%E6%8C%81" aria-hidden="true"></a>新增模型支持</h2>
<h3 data-sourcepos="4:1-4:19" dir="auto">
<a id="user-content-pytorch-paddle" class="anchor" href="#pytorch-paddle" aria-hidden="true"></a>PyTorch->Paddle</h3>
<ul data-sourcepos="5:1-6:80" dir="auto">
<li data-sourcepos="5:1-5:88">首次支持声音克隆模型:<a href="https://github.com/babysor/MockingBird" rel="nofollow noreferrer noopener" target="_blank">MockingBird</a>
</li>
<li data-sourcepos="6:1-6:80">轻量级检测算法:<a href="https://github.com/Megvii-BaseDetection/YOLOX" rel="nofollow noreferrer noopener" target="_blank">YOLOX</a>
</li>
</ul>
<h2 data-sourcepos="7:1-7:11" dir="auto">
<a id="user-content-新增op" class="anchor" href="#%E6%96%B0%E5%A2%9Eop" aria-hidden="true"></a>新增op</h2>
<ul data-sourcepos="8:1-9:86" dir="auto">
<li data-sourcepos="8:1-8:223">新增PyTorch op(5个):aten::repeat_interleave、aten::maxpool1d、aten::frobenius_norm、aten::silu、aten::format <a href="https://github.com/PaddlePaddle/X2Paddle/pull/691" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/691</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/696" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/696</a>
</li>
<li data-sourcepos="9:1-9:86">新增ONNX op(1个):GatherND <a href="https://github.com/PaddlePaddle/X2Paddle/pull/645" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/645</a>
</li>
</ul>
<h2 data-sourcepos="10:1-10:11" dir="auto">
<a id="user-content-修复op" class="anchor" href="#%E4%BF%AE%E5%A4%8Dop" aria-hidden="true"></a>修复op</h2>
<ul data-sourcepos="11:1-13:164" dir="auto">
<li data-sourcepos="11:1-11:225">修复PyTorch op(5个):aten::div、aten::index_select、aten::select、aten::upsample_nearest2d、aten::_convolution <a href="https://github.com/PaddlePaddle/X2Paddle/pull/691" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/691</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/689" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/689</a>
</li>
<li data-sourcepos="12:1-12:196">修复ONNX op(3个):Split、Conv、Pad <a href="https://github.com/PaddlePaddle/X2Paddle/pull/645" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/645</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/707" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/707</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/687" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/687</a>
</li>
<li data-sourcepos="13:1-13:164">修复PyTorch训练代码转换(2个):ToPILImage、Split <a href="https://github.com/PaddlePaddle/X2Paddle/pull/690" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/690</a> <a href="https://github.com/PaddlePaddle/X2Paddle/pull/694" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/694</a>
</li>
</ul>
<h2 data-sourcepos="14:1-14:15" dir="auto">
<a id="user-content-修复功能" class="anchor" href="#%E4%BF%AE%E5%A4%8D%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>修复功能</h2>
<ul data-sourcepos="15:1-17:112" dir="auto">
<li data-sourcepos="15:1-15:83">Windows环境生成代码问题 <a href="https://github.com/PaddlePaddle/X2Paddle/pull/706" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/706</a>
</li>
<li data-sourcepos="16:1-16:84">修复ONNX和PyTorch版本问题 <a href="https://github.com/PaddlePaddle/X2Paddle/pull/708" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/708</a>
</li>
<li data-sourcepos="17:1-17:112">去除代码优化部分to_tensor操作,优化预测速度 <a href="https://github.com/PaddlePaddle/X2Paddle/pull/696" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/696</a>
</li>
</ul>
<h2 data-sourcepos="18:1-18:15" dir="auto">
<a id="user-content-contributors" class="anchor" href="#contributors" aria-hidden="true"></a>Contributors</h2>
<ul data-sourcepos="19:1-19:132" dir="auto">
<li data-sourcepos="19:1-19:132">Thanks to external developers for their contributions to open source projects!@littletomatodonkey @zhoucz97 @firedent @dawn1206</li>
</ul>
2021-12-07T06:37:25+08:00
wjj19950828
wjjisloser@163.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.3.4
v1.3.4
<h2 data-sourcepos="1:1-1:15" dir="auto">
<a id="user-content-修复功能" class="anchor" href="#%E4%BF%AE%E5%A4%8D%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>修复功能</h2>
<ul data-sourcepos="2:1-2:106" dir="auto">
<li data-sourcepos="2:1-2:106">修复X2Paddle调用ResNet以及Vgg预训练模型bug(<a href="https://github.com/PaddlePaddle/X2Paddle/pull/684" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/684</a>)</li>
</ul>
2021-10-15T00:10:08+08:00
wjj19950828
wjjisloser@163.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.3.3
v1.3.3
<h2 data-sourcepos="1:1-1:15" dir="auto">
<a id="user-content-新增功能" class="anchor" href="#%E6%96%B0%E5%A2%9E%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>新增功能</h2>
<ul data-sourcepos="2:1-2:104" dir="auto">
<li data-sourcepos="2:1-2:104">集成Paddle-Lite opt工具,提供一键转换API(<a href="https://github.com/PaddlePaddle/X2Paddle/pull/681" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/681</a>)</li>
</ul>
<h2 data-sourcepos="3:1-3:11" dir="auto">
<a id="user-content-新增op" class="anchor" href="#%E6%96%B0%E5%A2%9Eop" aria-hidden="true"></a>新增op</h2>
<ul data-sourcepos="4:1-5:134" dir="auto">
<li data-sourcepos="4:1-4:151">新增PyTorch op转换支持(3个):aten::bitwise_not、aten::bitwise_xor、aten::bitwise_and(<a href="https://github.com/PaddlePaddle/X2Paddle/pull/677" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/677</a>)</li>
<li data-sourcepos="5:1-5:134">新增ONNX op转换支持(5个):Max、Min、GreaterOrEqual、GatherND、And(<a href="https://github.com/PaddlePaddle/X2Paddle/pull/682" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/682</a>)</li>
</ul>
<h2 data-sourcepos="6:1-6:11" dir="auto">
<a id="user-content-修复op" class="anchor" href="#%E4%BF%AE%E5%A4%8Dop" aria-hidden="true"></a>修复op</h2>
<ul data-sourcepos="7:1-8:99" dir="auto">
<li data-sourcepos="7:1-7:177">修复ONNX op转换支持(4个):ReduceSum、Unsqueeze(<a href="https://github.com/PaddlePaddle/X2Paddle/pull/682" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/682</a>) Clip、PRelu(<a href="https://github.com/PaddlePaddle/X2Paddle/pull/680" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/680</a>)</li>
<li data-sourcepos="8:1-8:99">修复Caffe op转换支持(1个):pooling(<a href="https://github.com/PaddlePaddle/X2Paddle/pull/681" rel="nofollow noreferrer noopener" target="_blank">https://github.com/PaddlePaddle/X2Paddle/pull/681</a>)</li>
</ul>
2021-10-14T00:08:36+08:00
wjj19950828
wjjisloser@163.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.3.2
v1.3.2
<h2 data-sourcepos="1:1-1:15" dir="auto">
<a id="user-content-新增功能" class="anchor" href="#%E6%96%B0%E5%A2%9E%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>新增功能</h2>
<ul data-sourcepos="2:1-2:46" dir="auto">
<li data-sourcepos="2:1-2:46">新增PyTorch op映射(1个):aten::sum</li>
</ul>
<h2 data-sourcepos="3:1-3:15" dir="auto">
<a id="user-content-修复功能" class="anchor" href="#%E4%BF%AE%E5%A4%8D%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>修复功能</h2>
<ul data-sourcepos="4:1-6:27" dir="auto">
<li data-sourcepos="4:1-4:70">修复PyTorch op映射(2个):aten::instance_norm、aten::split</li>
<li data-sourcepos="5:1-5:65">修复Caffe op映射(3个):directly_map, Eltwise and crop</li>
<li data-sourcepos="6:1-6:27">修复ONNX版本判断bug</li>
</ul>
2021-09-30T11:08:45+08:00
wjj19950828
wjjisloser@163.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.3.1
v1.3.1
<h2 data-sourcepos="1:1-1:28" dir="auto">
<a id="user-content-新增pytorch模型支持" class="anchor" href="#%E6%96%B0%E5%A2%9Epytorch%E6%A8%A1%E5%9E%8B%E6%94%AF%E6%8C%81" aria-hidden="true"></a>新增PyTorch模型支持</h2>
<ul data-sourcepos="2:1-2:43" dir="auto">
<li data-sourcepos="2:1-2:43"><a href="https://arxiv.org/abs/2001.08972" rel="nofollow noreferrer noopener" target="_blank">SOLAR</a></li>
</ul>
<h2 data-sourcepos="3:1-3:15" dir="auto">
<a id="user-content-修复功能" class="anchor" href="#%E4%BF%AE%E5%A4%8D%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>修复功能</h2>
<ul data-sourcepos="4:1-4:57" dir="auto">
<li data-sourcepos="4:1-4:57">修复PyTorch op映射(2个):expand_as、Upsample</li>
</ul>
2021-09-30T11:08:45+08:00
wjj19950828
wjjisloser@163.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.3.0
v1.3.0
<ol data-sourcepos="1:1-5:80" dir="auto">
<li data-sourcepos="1:1-1:241">新增MMDetection模型库支持,包括YOLO-V3、FCOS、RetinaNet、SSD、Faster R-CNN以及FSAF,有相关AP精度对比,具体参考<a href="/paddlepaddle/X2Paddle/-/blob/develop/./docs/inference_model_convertor/toolkits/MMDetection2paddle.md">MMDetection模型转换指南</a>。</li>
<li data-sourcepos="2:1-2:164">新增PyTorch训练代码转换对<a href="https://github.com/clovaai/CRAFT-pytorch" rel="nofollow noreferrer noopener" target="_blank">CRAFT</a>的支持,新增PyTorch预测模型转换对Seg-Swin-Transformer的支持。</li>
<li data-sourcepos="3:1-3:73">优化模型预测速度,去除forward函数开头to_tensor操作。</li>
<li data-sourcepos="4:1-4:48">新增Tensorflow op映射(1个):Sign。</li>
<li data-sourcepos="5:1-5:80">新增ONNX op映射(4个):NMS、ReduceL1、ReduceL2、3D Interpolate。</li>
</ol>
2021-09-30T11:08:45+08:00
WJJ1995
wjjisloser@163.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.2.3
v1.2.3
<h2 data-sourcepos="1:1-1:21" dir="auto">
<a id="user-content-新增模型转换" class="anchor" href="#%E6%96%B0%E5%A2%9E%E6%A8%A1%E5%9E%8B%E8%BD%AC%E6%8D%A2" aria-hidden="true"></a>新增模型转换</h2>
<p data-sourcepos="2:1-2:459" dir="auto">ONNX2Paddle新增模型转换:<a href="https://github.com/open-mmlab/mmdetection/blob/master/configs/retinanet/retinanet_r50_fpn_1x_coco.py" rel="nofollow noreferrer noopener" target="_blank">RetinaNet</a>、<a href="https://github.com/open-mmlab/mmdetection/blob/master/configs/fsaf/fsaf_r50_fpn_1x_coco.py" rel="nofollow noreferrer noopener" target="_blank">FSAF</a>、<a href="https://github.com/open-mmlab/mmdetection/blob/master/configs/ssd/ssd300_coco.py" rel="nofollow noreferrer noopener" target="_blank">SSD</a>、<a href="https://github.com/open-mmlab/mmdetection/blob/master/configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py" rel="nofollow noreferrer noopener" target="_blank">Faster R-CNN</a></p>
2021-09-30T11:08:45+08:00
SunAhong1993
sunyanfang01@baidu.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.2.2
v1.2.2
<h2 data-sourcepos="1:1-1:15" dir="auto">
<a id="user-content-优化功能" class="anchor" href="#%E4%BC%98%E5%8C%96%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>优化功能</h2>
<ol data-sourcepos="2:1-4:48" dir="auto">
<li data-sourcepos="2:1-2:50">Caffe Pooling:支持kernel_size=0的情况。</li>
<li data-sourcepos="3:1-3:44">ONNX Resize:支持3d Tensor的转换。</li>
<li data-sourcepos="4:1-4:48">支持PaddlePaddle 2.0.x 中的paddle type。</li>
</ol>
2021-09-30T11:08:45+08:00
channingss
chen_lingchi@163.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.2.1
v1.2.1
<h2 data-sourcepos="1:1-1:15" dir="auto">
<a id="user-content-新增功能" class="anchor" href="#%E6%96%B0%E5%A2%9E%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>新增功能</h2>
<ul data-sourcepos="2:1-2:47" dir="auto">
<li data-sourcepos="2:1-2:47">支持使用PaddlePaddle 2.1.0进行转换。</li>
</ul>
<h2 data-sourcepos="3:1-3:15" dir="auto">
<a id="user-content-使用优化" class="anchor" href="#%E4%BD%BF%E7%94%A8%E4%BC%98%E5%8C%96" aria-hidden="true"></a>使用优化</h2>
<ul data-sourcepos="4:1-4:86" dir="auto">
<li data-sourcepos="4:1-4:86">移除静态图代码模式,全面切换为基于Paddle动态图代码的转换。</li>
</ul>
<h2 data-sourcepos="5:1-5:31" dir="auto">
<a id="user-content-新增可转换的onnx模型" class="anchor" href="#%E6%96%B0%E5%A2%9E%E5%8F%AF%E8%BD%AC%E6%8D%A2%E7%9A%84onnx%E6%A8%A1%E5%9E%8B" aria-hidden="true"></a>新增可转换的ONNX模型</h2>
<ul data-sourcepos="6:1-7:126" dir="auto">
<li data-sourcepos="6:1-6:126"><a href="https://github.com/open-mmlab/mmdetection/blob/master/configs/fcos/fcos_r50_caffe_fpn_gn-head_1x_coco.py" rel="nofollow noreferrer noopener" target="_blank">MMDetection Fcos</a></li>
<li data-sourcepos="7:1-7:126"><a href="https://github.com/open-mmlab/mmdetection/blob/master/configs/yolo/yolov3_d53_mstrain-608_273e_coco.py" rel="nofollow noreferrer noopener" target="_blank">MMDetection Yolov3</a></li>
</ul>
2021-09-30T11:08:45+08:00
Jason
jiangjiajun@baidu.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.2.0
v1.2.0
<h2 data-sourcepos="1:1-1:15" dir="auto">
<a id="user-content-新增功能" class="anchor" href="#%E6%96%B0%E5%A2%9E%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>新增功能</h2>
<ul data-sourcepos="3:1-4:142" dir="auto">
<li data-sourcepos="3:1-3:158">支持PyTorch训练项目转换,使用方式详见<a href="https://github.com/PaddlePaddle/X2Paddle/tree/develop/docs/pytorch_project_convertor#readme" rel="nofollow noreferrer noopener" target="_blank">文档</a>。</li>
<li data-sourcepos="4:1-4:142">新增PyTorch-PaddlePaddle API <a href="https://github.com/PaddlePaddle/X2Paddle/tree/develop/docs/pytorch_project_convertor/API_docs" rel="nofollow noreferrer noopener" target="_blank">映射表</a>。</li>
</ul>
<h2 data-sourcepos="5:1-5:34" dir="auto">
<a id="user-content-新增可转换的pytorch项目" class="anchor" href="#%E6%96%B0%E5%A2%9E%E5%8F%AF%E8%BD%AC%E6%8D%A2%E7%9A%84pytorch%E9%A1%B9%E7%9B%AE" aria-hidden="true"></a>新增可转换的PyTorch项目</h2>
<ul data-sourcepos="6:1-6:191" dir="auto">
<li data-sourcepos="6:1-6:191">支持转换的项目:<a href="https://github.com/yunjey/stargan" rel="nofollow noreferrer noopener" target="_blank">StarGAN</a>、<a href="https://github.com/Linzaer/Ultra-Light-Fast-Generic-Face-Detector-1MB" rel="nofollow noreferrer noopener" target="_blank">Ultra-Light-Fast-Generic-Face-Detector-1MB</a>。</li>
</ul>
2021-09-30T11:08:45+08:00
SunAhong1993
48579383+SunAhong1993@users.noreply.github.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.1.0
v1.1.0
<h2 data-sourcepos="1:1-1:27" dir="auto">
<a id="user-content-新增支持转的模型" class="anchor" href="#%E6%96%B0%E5%A2%9E%E6%94%AF%E6%8C%81%E8%BD%AC%E7%9A%84%E6%A8%A1%E5%9E%8B" aria-hidden="true"></a>新增支持转的模型</h2>
<p data-sourcepos="2:1-2:308" dir="auto"><a href="https://github.com/microsoft/Swin-Transformer/" rel="nofollow noreferrer noopener" target="_blank">SwinTransformer</a>、<a href="https://github.com/xuebinqin/BASNet" rel="nofollow noreferrer noopener" target="_blank">BASNet</a>、<a href="https://github.com/dlunion/DBFace" rel="nofollow noreferrer noopener" target="_blank">DBFace</a>、<a href="https://github.com/JaidedAI/EasyOCR" rel="nofollow noreferrer noopener" target="_blank">EasyOCR</a>、<a href="https://github.com/tensorflow/models/blob/master/research/slim/nets/cifarnet.py" rel="nofollow noreferrer noopener" target="_blank">CifarNet</a>等。</p>
<h2 data-sourcepos="3:1-3:15" dir="auto">
<a id="user-content-新增功能" class="anchor" href="#%E6%96%B0%E5%A2%9E%E5%8A%9F%E8%83%BD" aria-hidden="true"></a>新增功能</h2>
<ul data-sourcepos="4:1-8:48" dir="auto">
<li data-sourcepos="4:1-4:42">修复Windows上使用本工具的bug。</li>
<li data-sourcepos="5:1-5:95">新增TensorFlow op映射(4个):SplitV、ReverseV2、BatchToSpaceND、SpaceToBatchND。</li>
<li data-sourcepos="6:1-6:246">新增PyTorch op映射(11个):aten::index、aten::roll、aten::adaptive_avg_pool1d、aten::reflection_pad2d、aten::reflection_pad1d、aten::instance_norm、aten::gru、aten::norm、aten::clamp_min、aten::prelu、aten:split_with_sizes。</li>
<li data-sourcepos="7:1-7:49">新增ONNX op映射(1个):DepthToSpace。</li>
<li data-sourcepos="8:1-8:48">新增Caffe op映射(1个):MemoryData。</li>
</ul>
2021-09-30T11:08:45+08:00
channingss
chen_lingchi@163.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.0.2
v1.0.2
<h2 data-sourcepos="1:1-1:15" dir="auto">
<a id="user-content-功能更新" class="anchor" href="#%E5%8A%9F%E8%83%BD%E6%9B%B4%E6%96%B0" aria-hidden="true"></a>功能更新</h2>
<h3 data-sourcepos="2:1-2:11" dir="auto">
<a id="user-content-pytorch" class="anchor" href="#pytorch" aria-hidden="true"></a>PyTorch</h3>
<ul data-sourcepos="3:1-4:243" dir="auto">
<li data-sourcepos="3:1-3:226">新增支持的op:aten::index、aten::roll、aten::adaptive_avg_pool1d、aten::reflection_pad2d、aten::reflection_pad1d、aten::instance_norm、aten::gru、aten::norm、aten::clamp_min、aten:prelu、aten:split_with_sizes</li>
<li data-sourcepos="4:1-4:243">新增支持转换的模型:<a href="https://github.com/microsoft/Swin-Transformer/" rel="nofollow noreferrer noopener" target="_blank">SwinTransformer</a>、<a href="https://github.com/xuebinqin/BASNet" rel="nofollow noreferrer noopener" target="_blank">BASNet</a>、<a href="https://github.com/dlunion/DBFace" rel="nofollow noreferrer noopener" target="_blank">DBFace</a>、<a href="https://github.com/JaidedAI/EasyOCR" rel="nofollow noreferrer noopener" target="_blank">EasyOCR</a>等</li>
</ul>
<h3 data-sourcepos="5:1-5:14" dir="auto">
<a id="user-content-tensorflow" class="anchor" href="#tensorflow" aria-hidden="true"></a>TensorFlow</h3>
<ul data-sourcepos="6:1-6:40" dir="auto">
<li data-sourcepos="6:1-6:40">新增支持的op:SplitV、ReverseV2</li>
</ul>
<h3 data-sourcepos="7:1-7:8" dir="auto">
<a id="user-content-onnx" class="anchor" href="#onnx" aria-hidden="true"></a>ONNX</h3>
<ul data-sourcepos="8:1-9:126" dir="auto">
<li data-sourcepos="8:1-8:34">新增支持的op:DepthToSpace</li>
<li data-sourcepos="9:1-9:126">新增支持转换的模型:<a href="https://github.com/tensorflow/models/blob/master/research/slim/nets/cifarnet.py" rel="nofollow noreferrer noopener" target="_blank">CifarNet</a>等</li>
</ul>
<h3 data-sourcepos="10:1-10:9" dir="auto">
<a id="user-content-caffe" class="anchor" href="#caffe" aria-hidden="true"></a>Caffe</h3>
<ul data-sourcepos="11:1-12:0" dir="auto">
<li data-sourcepos="11:1-12:0">新增支持的op:MemoryData</li>
</ul>
<h2 data-sourcepos="13:1-13:12" dir="auto">
<a id="user-content-bug修复" class="anchor" href="#bug%E4%BF%AE%E5%A4%8D" aria-hidden="true"></a>Bug修复</h2>
<h3 data-sourcepos="14:1-14:10" dir="auto">
<a id="user-content-全局" class="anchor" href="#%E5%85%A8%E5%B1%80" aria-hidden="true"></a>全局</h3>
<ul data-sourcepos="15:1-16:63" dir="auto">
<li data-sourcepos="15:1-15:29">修复中文编码问题。</li>
<li data-sourcepos="16:1-16:63">修复路径设置问题,支持Windows上使用本工具。</li>
</ul>
<h3 data-sourcepos="17:1-17:9" dir="auto">
<a id="user-content-caffe-1" class="anchor" href="#caffe-1" aria-hidden="true"></a>Caffe</h3>
<ul data-sourcepos="18:1-18:56" dir="auto">
<li data-sourcepos="18:1-18:56">BatchNorm映射不支持输入shape为2维的情况。</li>
</ul>
<h3 data-sourcepos="19:1-19:14" dir="auto">
<a id="user-content-tensorflow-1" class="anchor" href="#tensorflow-1" aria-hidden="true"></a>TensorFlow</h3>
<ul data-sourcepos="20:1-21:45" dir="auto">
<li data-sourcepos="20:1-20:44">Transpose映射<code>perm</code>为Tensor的情况。</li>
<li data-sourcepos="21:1-21:45">Pad映射无法获取<code>pad_mode</code>的情况。</li>
</ul>
<h3 data-sourcepos="22:1-22:8" dir="auto">
<a id="user-content-onnx-1" class="anchor" href="#onnx-1" aria-hidden="true"></a>ONNX</h3>
<ul data-sourcepos="23:1-23:48" dir="auto">
<li data-sourcepos="23:1-23:48">Pad映射不支持<code>pad</code>长度为10的情况。</li>
</ul>
2021-09-30T11:08:45+08:00
SunAhong1993
48579383+SunAhong1993@users.noreply.github.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.0.1
v1.0.1
<h2 data-sourcepos="1:1-1:15" dir="auto">
<a id="user-content-功能更新" class="anchor" href="#%E5%8A%9F%E8%83%BD%E6%9B%B4%E6%96%B0" aria-hidden="true"></a>功能更新</h2>
<ul data-sourcepos="2:1-5:132" dir="auto">
<li data-sourcepos="2:1-2:238">
<ol data-sourcepos="2:3-2:238">
<li data-sourcepos="2:3-2:238">新增PyTorch模型转换,支持Tracing和Scripting两种方式进行转换,转换方式和使用说明参考文档<a href="https://github.com/PaddlePaddle/X2Paddle/blob/develop/docs/user_guides/pytorch2paddle.md" rel="nofollow noreferrer noopener" target="_blank">PyTorch模型转换</a>;</li>
</ol>
</li>
<li data-sourcepos="3:1-3:172">
<ol start="2" data-sourcepos="3:3-3:172">
<li data-sourcepos="3:3-3:172">新增Caffe/ONNX/Tensorflow到Paddle2.0 动态图的转换支持,使用方式参考<a href="https://github.com/PaddlePaddle/X2Paddle/blob/develop/README.md" rel="nofollow noreferrer noopener" target="_blank">说明文档</a>;</li>
</ol>
</li>
<li data-sourcepos="4:1-4:180">
<ol start="3" data-sourcepos="4:3-4:180">
<li data-sourcepos="4:3-4:180">TensorFlow新增14个OP算子转换支持,包含Neg、Greater、FloorMod、LogicalAdd、Prod、Equal、Conv3D、Ceil、AddN、DivNoNan、Where、MirrorPad、Size、TopKv2;</li>
</ol>
</li>
<li data-sourcepos="5:1-5:132">
<ol start="4" data-sourcepos="5:3-5:132">
<li data-sourcepos="5:3-5:132">ONNX新增9个OP算子转换支持,包括ArgMax、Sign、Reciprocal、Size、OneHot、ReduceProd、LogSoftmax、LSTM、LRN。</li>
</ol>
</li>
</ul>
<h2 data-sourcepos="6:1-6:18" dir="auto">
<a id="user-content-易用性更新" class="anchor" href="#%E6%98%93%E7%94%A8%E6%80%A7%E6%9B%B4%E6%96%B0" aria-hidden="true"></a>易用性更新</h2>
<ul data-sourcepos="7:1-9:0" dir="auto">
<li data-sourcepos="7:1-7:138">
<ol data-sourcepos="7:3-7:138">
<li data-sourcepos="7:3-7:138">新增Optimizer模块,主要包括op融合、op消除功能,提升转换后模型代码的可读性以及模型的预测性能;</li>
</ol>
</li>
<li data-sourcepos="8:1-9:0">
<ol start="2" data-sourcepos="8:3-9:0">
<li data-sourcepos="8:3-9:0">新增8个op算子融合优化,包括PyTorch:adativate_pool细粒度op融合、batchnorm细粒度op融合、常数赋值融合、trace fc细粒度融合、script fc细粒度融合、双线性细粒度融合、reshape细粒度融合;Tensorflow:prelu融合</li>
</ol>
</li>
</ul>
2021-09-30T11:08:45+08:00
SunAhong1993
sunyanfang01@baidu.com
https://gitcode.net/paddlepaddle/X2Paddle/-/tags/v1.0.0rc1
v1.0.0rc1
<h2 data-sourcepos="1:1-1:15" dir="auto">
<a id="user-content-功能更新" class="anchor" href="#%E5%8A%9F%E8%83%BD%E6%9B%B4%E6%96%B0" aria-hidden="true"></a>功能更新</h2>
<ul data-sourcepos="2:1-4:110" dir="auto">
<li data-sourcepos="2:1-2:61">
<ol data-sourcepos="2:3-2:61">
<li data-sourcepos="2:3-2:61">新增PyTorch Scripting方式转换后代码的优化;</li>
</ol>
</li>
<li data-sourcepos="3:1-3:51">
<ol start="2" data-sourcepos="3:3-3:51">
<li data-sourcepos="3:3-3:51">ONNX新增1个OP算子转换支持:ArgMax;</li>
</ol>
</li>
<li data-sourcepos="4:1-4:110">
<ol start="3" data-sourcepos="4:3-4:110">
<li data-sourcepos="4:3-4:110">PyTorch新增5个OP算子转换支持:aten::bmm、aten::sub_ 、aten:erf、aten::lstm、aten::gather。</li>
</ol>
</li>
</ul>
2021-09-30T11:08:46+08:00
channingss
chen_lingchi@163.com