tag:github.com,2008:https://github.com/aobannong/pytorch/releases Release notes from pytorch 2019-08-08T12:54:09Z tag:github.com,2008:Repository/208687989/v1.2.0 2019-08-08T12:54:09Z v1.2.0 <p>delete C_CONTIGUOUS assertions to be compatible with particular build…</p> soumith tag:github.com,2008:Repository/208687989/v1.1.0 2019-04-30T23:22:19Z v1.1.0: Fix version handler in 1.1.0 docs. (#19977) <p>Update the find &amp; replace to be less restrictive. Will port this change<br> to master to avoid problems in the future.</p> zou3519 tag:github.com,2008:Repository/208687989/v1.0.1 2019-02-07T07:24:24Z v1.0.1: Remove unnecessary typing dependency. (#16776) <p>Signed-off-by: Edward Z. Yang <a href="mailto:ezyang@fb.com">ezyang@fb.com</a></p> ezyang tag:github.com,2008:Repository/208687989/v1.0.0 2018-12-06T23:44:56Z v1.0.0 <p>add fix for CUDA 10</p> soumith tag:github.com,2008:Repository/208687989/v1.0rc1 2018-10-02T04:54:52Z v1.0rc1 <p>Back out "Revert D10123245: Back out "codemod cuda_gpu_id to device_i…</p> bddppq tag:github.com,2008:Repository/208687989/v1.0rc0 2018-10-02T04:54:52Z v1.0rc0 <p>Back out "Revert D10123245: Back out "codemod cuda_gpu_id to device_i…</p> bddppq tag:github.com,2008:Repository/208687989/v0.4.1 2018-07-26T00:28:04Z v0.4.1 <p>fix lint</p> soumith tag:github.com,2008:Repository/208687989/v0.4.0 2018-05-30T21:37:47Z v0.4.0 <p>move to eigenteam github for eigen submodule</p> soumith tag:github.com,2008:Repository/208687989/v0.3.1 2018-02-09T17:07:43Z v0.3.1: Scopes 0.3.1 backport (#5153) <ul> <li> <p>Introduce scopes during tracing (<a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="263595888" data-permission-text="Title is private" data-url="https://github.com/pytorch/pytorch/issues/3016" data-hovercard-type="pull_request" data-hovercard-url="/pytorch/pytorch/pull/3016/hovercard" href="https://github.com/pytorch/pytorch/pull/3016">pytorch#3016</a>)</p> </li> <li> <p>Fix segfault during ONNX export</p> </li> <li> <p>Further fix to tracing scope (<a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="287133511" data-permission-text="Title is private" data-url="https://github.com/pytorch/pytorch/issues/4558" data-hovercard-type="pull_request" data-hovercard-url="/pytorch/pytorch/pull/4558/hovercard" href="https://github.com/pytorch/pytorch/pull/4558">pytorch#4558</a>)</p> </li> <li> <p>Set missing temporary scope in callPySymbolicMethod</p> </li> <li> <p>Use expected traces in all scope tests</p> </li> <li> <p>Fix tracking of tracing scopes during ONNX pass (<a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="286570426" data-permission-text="Title is private" data-url="https://github.com/pytorch/pytorch/issues/4524" data-hovercard-type="pull_request" data-hovercard-url="/pytorch/pytorch/pull/4524/hovercard" href="https://github.com/pytorch/pytorch/pull/4524">pytorch#4524</a>)</p> </li> <li> <p>Fix tracking of tracing scopes during ONNX pass</p> </li> <li> <p>Use ResourceGuard to manage setting a temporary current scope in Graph</p> </li> <li> <p>Add tests for ONNX pass scopes</p> </li> <li> <p>Remove unused num_classes argument</p> </li> <li> <p>Expose node scopeName to python (<a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="282550193" data-permission-text="Title is private" data-url="https://github.com/pytorch/pytorch/issues/4200" data-hovercard-type="pull_request" data-hovercard-url="/pytorch/pytorch/pull/4200/hovercard" href="https://github.com/pytorch/pytorch/pull/4200">pytorch#4200</a>)</p> </li> <li> <p>Inherit JIT scopes when cloning only when it's correct</p> </li> </ul> <p>It's correct only when the new graph owns the same scope tree<br> as the original one. We can end up with dangling pointers otherwise.</p> <ul> <li> <p>Fixes after cherry-picking, still one test to go</p> </li> <li> <p>Fix for last failing test after scope cherry-pick</p> </li> <li> <p>Fix linting issue</p> </li> </ul> lantiga tag:github.com,2008:Repository/208687989/v0.3.0 2017-12-04T08:00:43Z Backport transposes optimization to v0.3.0 (#3994) <ul> <li> <p>Optimizer: optimize transposes in variety of circumstances (<a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="271554935" data-permission-text="Title is private" data-url="https://github.com/pytorch/pytorch/issues/3509" data-hovercard-type="pull_request" data-hovercard-url="/pytorch/pytorch/pull/3509/hovercard" href="https://github.com/pytorch/pytorch/pull/3509">pytorch#3509</a>)</p> </li> <li> <p>Optimizer: Optimize transposes in variety of circumstances</p> </li> </ul> <ul> <li>No-op transposes</li> <li>Consecutive transposes (fuse them)</li> <li>Transposes into Gemm (fuse them into transA/transB parameter)</li> </ul> <ul> <li> <p>touch up out of date comment</p> </li> <li> <p>Backporting optimizer changes</p> </li> </ul> dzhulgakov