- 25 Jun, 2019 2 commits
-
-
Summary: In multi-threaded applications where we have multiple inferences on the same model in parallel (consider e.g. a TTS system handling multiple requests), it can be useful to share the parameters of a model amongst these multiple instances. This improves the cache utilization behaviour of the system, as multiple cores can use the same set of weights instead of evicting the identical copies of weights in a shared cache. As the underlying `NDArray` instances in `data_entry_` implement a ref-counted based sharing system, this is a simple modification of the `GraphRuntime::LoadParams` logic to instead copy parameters from an existing GraphRuntime instance. This is a little ugly in that we need both the pre-existing GraphRuntime instance, as well as the 'serialized' params (since we need to know the set of names we should copy), but without imposing additional assumptions (i.e. storing the set of param names in GraphRuntime, and enforcing that shared param names are identical to the parameters set in the preceding `LoadParams` call), this seems unavoidable. Test Plan: Unit test added.
Andrew Tulloch committed -
Sammy committed
-
- 24 Jun, 2019 2 commits
-
-
雾雨魔理沙 committed
-
Alexander Pivovarov committed
-
- 23 Jun, 2019 1 commit
-
-
* Support bidirectional RNN layer * tweak * tweak
Haichen Shen committed
-
- 22 Jun, 2019 5 commits
-
-
雾雨魔理沙 committed
-
* [QUANTIZE] Support for clip operator * [QUANTIZE] Memorizing the quantize node mapping. * [QUANTIZE] Remove use_stop_fusion and skip_k_conv in qconfig * update * update * update * update
ziheng committed -
Wei Chen committed
-
Wei Chen committed
-
Jessica Davies committed
-
- 21 Jun, 2019 2 commits
-
-
* update README * fix typo
Luis Vega committed -
Lianmin Zheng committed
-
- 20 Jun, 2019 5 commits
-
-
* Add EtaExpand to transform API * Add test case
Wei Chen committed -
henrywu2019 committed
-
Wuwei Lin committed
-
Zhi committed
-
Issue: when using vivado compile vta.cc with top function 'vta', vivado report deadlock error like '...with default size is used in a non -dataflow region, which may result in deadlock Please consider to resize the stream using the directive ‘set_directive_stream’ or the ‘HL S stream’ pragma.' Solution: give the queue a default size as 8.
Hua Jiang committed
-
- 19 Jun, 2019 2 commits
- 18 Jun, 2019 4 commits
-
-
* [CI] Update ci-gpu to v0.52 * update nodejs
Tianqi Chen committed -
Alexander Pivovarov committed
-
Tianqi Chen committed
-
Zhi committed
-
- 17 Jun, 2019 8 commits
-
-
This reverts commit df6957a5.
Tianqi Chen committed -
Alexander Pivovarov committed
-
Jared Roesch committed
-
Howave committed
-
Wuwei Lin committed
-
Zhi committed
-
Tianqi Chen committed
-
Sheng Zha committed
-
- 15 Jun, 2019 2 commits
-
-
save save save upstream lint remove bad changes fix build save save please the ci god Update src/relay/pass/partial_eval.cc Co-Authored-By: Wei Chen <ipondering.weic@gmail.com> save fix test ci is ANGRY fix rebase problem fix rebase add test save save comment
雾雨魔理沙 committed -
Alexander Pivovarov committed
-
- 14 Jun, 2019 5 commits
-
-
* Update vm print & add AllocTensor instruction * patch * fix invoke packed * update cmake * tweak move * update invoke_closure * lint * add doc * tweak
Haichen Shen committed -
Alexander Pivovarov committed
-
Tianqi Chen committed
-
Luis Vega committed
-
* fix flaky test * fix flaky quantize pass
Haichen Shen committed
-
- 13 Jun, 2019 2 commits
-
-
* add support to event counters in VTA * fix comment * fix event-counter interface parameter * no longer needed * add sim back * add docs to event counters * fix docs * add more details about event counting * make dpi-module docs more accurate
Luis Vega committed -
Marcelo Duarte Trevisani committed
-