These 30 commits are when the Protocol Buffers files have changed:
| Commit: | 12270cd | |
|---|---|---|
| Author: | Lequn Chen | |
Hack: transfer filename instead of real image Hopefully this could mitigate the backend load imbalance
The documentation is generated from this commit.
| Commit: | 4c2d41e | |
|---|---|---|
| Author: | Lequn Chen | |
per-gpu profile use `-generic_profile` to use the same profile for GPUs of the same model.
| Commit: | 2ef205a | |
|---|---|---|
| Author: | Lequn Chen | |
complex query checkpoint
| Commit: | c623092 | |
|---|---|---|
| Author: | Lequn Chen | |
Revert "Merge remote-tracking branch 'bingyu/develop' into lqchen" This reverts commit 3f1b7658d03393bb52c34fb79c9638a058eda243, reversing changes made to f2541b539f103491c4ee3e61639455f2fab82fc5.
| Commit: | 3f1b765 | |
|---|---|---|
| Author: | Lequn Chen | |
Merge remote-tracking branch 'bingyu/develop' into lqchen (not compiled yet)
| Commit: | 85cd3db | |
|---|---|---|
| Author: | Lequn Chen | |
cmake support (not fully finished yet but TF compiles)
| Commit: | 2f3dbe2 | |
|---|---|---|
| Author: | crystalrem | |
add complex query analysis to scheduler
| Commit: | 69bf9ee | |
|---|---|---|
| Author: | crystalrem | |
bug fix
| Commit: | b7f2e9e | |
|---|---|---|
| Author: | crystalrem | |
fix some bugs in complex query analysis
| Commit: | 8f4ea0e | |
|---|---|---|
| Author: | crystalrem | |
add complex query analysis to scheduler
| Commit: | 0881668 | |
|---|---|---|
| Author: | Haichen Shen | |
More debugging info
| Commit: | 83b2fc9 | |
|---|---|---|
| Author: | Haichen Shen | |
Performance tuning
| Commit: | 6ef4041 | |
|---|---|---|
| Author: | Haichen Shen | |
Switch to use frontend to report workload request rate
| Commit: | 8a89435 | |
|---|---|---|
| Author: | Haichen Shen | |
Support customize weight in workload config
| Commit: | 0f1fa00 | |
|---|---|---|
| Author: | Haichen Shen | |
| Committer: | GitHub | |
Support tensorflow SSD model (#15)
| Commit: | 000e3ec | |
|---|---|---|
| Author: | haichen | |
| Committer: | Haichen Shen | |
Add breakdown latency for diagnosis in the user reply
| Commit: | 9d69e87 | |
|---|---|---|
| Author: | Haichen Shen | |
| Committer: | Haichen Shen | |
Change frontend to event driven mode
| Commit: | 13b4f7b | |
|---|---|---|
| Author: | Haichen Shen | |
| Committer: | Haichen Shen | |
Support to load backup models in backends, and forward requests to backups when queue is full
| Commit: | 19e82b6 | |
|---|---|---|
| Author: | Haichen Shen | |
| Committer: | Haichen Shen | |
Tweak share prefix model impl for better performance
| Commit: | cfbb481 | |
|---|---|---|
| Author: | Haichen Shen | |
Include latency details in backend reply
| Commit: | fa05d66 | |
|---|---|---|
| Author: | Haichen Shen | |
Add prefix model and add caffe2 support for prefix batching
| Commit: | 7f3a220 | |
|---|---|---|
| Author: | Haichen Shen | |
Update scheduling
| Commit: | adff9ab | |
|---|---|---|
| Author: | Haichen Shen | |
Bug fix
| Commit: | 94d8df9 | |
|---|---|---|
| Author: | Haichen Shen | |
Support multi-batching in scheduler and backend server
| Commit: | 27fd028 | |
|---|---|---|
| Author: | Haichen Shen | |
Add support for multi-batching in backend rpc client; add cpp test case
| Commit: | efbe9cb | |
|---|---|---|
| Author: | Haichen Shen | |
Add more function in global scheduler
| Commit: | 79a9453 | |
|---|---|---|
| Author: | Haichen Shen | |
Update control logic && backend pool
| Commit: | 5f2ceab | |
|---|---|---|
| Author: | Haichen Shen | |
Update model database and model ID
| Commit: | a243c4a | |
|---|---|---|
| Author: | Haichen Shen | |
Update docker file
| Commit: | de5cc11 | |
|---|---|---|
| Author: | Haichen Shen | |
init