These 10 commits are when the Protocol Buffers files have changed:
| Commit: | 3a2dc0e | |
|---|---|---|
| Author: | Matt T. Proud | |
Drop deprecated checkpoint file field. This is no longer used by the in-memory data stores nor the externally- hosted ones.
The documentation is generated from this commit.
| Commit: | 55a39f8 | |
|---|---|---|
| Author: | Oleksandr Sviridenko | |
| Committer: | Oleksandr Sviridenko | |
Output pipeline ID into the ExperimentArg proto that is built.
| Commit: | d22d9c1 | |
|---|---|---|
| Author: | Oleksandr Sviridenko | |
Output pipeline ID into the ExperimentArg proto that is built.
| Commit: | 2564f61 | |
|---|---|---|
| Author: | Oleksandr Sviridenko | |
Now application can request specific pipeline ID. This allows application to have own pipeline ID format and generate an ID from the scratch.
| Commit: | a82c554 | |
|---|---|---|
| Author: | Oleksandr Sviridenko | |
Add support of subjects with default settings in an experiment. The number of default subjects can be set for a cluster or a group of subjects. Fix broken CSV file download.
| Commit: | d92a15d | |
|---|---|---|
| Author: | Mikhail Bushkov | |
Added additional_exec_paths configuration parameter.
| Commit: | a0a47f6 | |
|---|---|---|
| Author: | dave bayer | |
Extend PipelineSynchronizer interface to have methods to allow a caller to sync with the pipeline points and add a synchronizer that will block the pipeline only at the end of the iteration.
| Commit: | e9bb976 | |
|---|---|---|
| Author: | Mikhail Bushkov | |
* Introduced HistoryDatastore interface and 1 implementation: MemoryHistoryDatastore. * Cleaned up ExperimentDb not to use caches - we don't need to cache anything there, as we're using HistoryDatastore. * Changed DisplayMediator to use HistoryDatastore to display historical data. NOTE: that it still maintains the list of all unique subjects (to display all-time best performers) which may leak memory. This will be fixed later.
| Commit: | 8ee98f6 | |
|---|---|---|
| Author: | Mikhail Bushkov | |
* Added PipelineIdGenerator that is sharding-aware, i.e. its' API hash shardForPipelineId() method that determines the corresponding shard for every PipelineId. * Introdcued PipelineState - class that encapsulated all the information about current pipeline state. PipelineState can serialize/deserialize itself. It uses corresponding protobufs to do that. * Refactored ExperimentDb to not depend on any storage layer. Made ExperimentDb a part of PipelineState. * Implemented MemoryDataStore. * Implemented PipelineRestorer, which restores pipelines from their saved states on Groningen server startup.
| Commit: | 0f4d921 | |
|---|---|---|
| Author: | Matt T. Proud | |
Initial import of Groningen into version control.