Ecosyste.ms: Timeline
Browse the timeline of events for every public repo on GitHub. Data updated hourly from GH Archive.
shuttie closed an issue on nixiesearch/nixiesearch
composite onnx models are not fetched from HF
So when a model is split between `model.onnx` and `model_data.onnx`, we only fetch the first split - which fails the loading
shuttie closed a pull request on nixiesearch/nixiesearch
add support for composite ONNX models (like bge-m3)
fixes https://github.com/nixiesearch/nixiesearch/issues/362
shuttie opened a pull request on nixiesearch/nixiesearch
switch to buildinfo for guessing CUDA support
Before that we used a set of heuristics based on classpath to guess it - now we can make it simpler and more reliable.
shuttie created a branch on nixiesearch/nixiesearch
fix/build-vars - Hybrid search engine, combining best features of text and semantic search worlds
shuttie opened a pull request on nixiesearch/nixiesearch
add support for composite ONNX models (like bge-m3)
fixes https://github.com/nixiesearch/nixiesearch/issues/362
shuttie created a branch on nixiesearch/nixiesearch
fix/onnx-data - Hybrid search engine, combining best features of text and semantic search worlds
shuttie published a release on nixiesearch/nixiesearch
0.4.0
## Major changes in 0.4.0 * we now use an embedded llamacpp-server (instead of the JNI wrapper), which allows much smoother llamacpp updates. * [date/datetime/geopoint](https://www.nixiesearch....
shuttie created a tag on nixiesearch/nixiesearch
0.4.0 - Hybrid search engine, combining best features of text and semantic search worlds
shuttie pushed 2 commits to master nixiesearch/nixiesearch
shuttie closed a pull request on nixiesearch/nixiesearch
migrate to scalafmt 3.8.4 - now with named tuples and scala 3.6 dialect
shuttie opened a pull request on nixiesearch/nixiesearch
migrate to scalafmt 3.8.4 - now with named tuples and scala 3.6 dialect
shuttie created a branch on nixiesearch/nixiesearch
fix/scalafmt384 - Hybrid search engine, combining best features of text and semantic search worlds
shuttie pushed 2 commits to master nixiesearch/nixiesearch
shuttie closed a pull request on nixiesearch/nixiesearch
enable ONNX verbose logs when global loglevel is verbose
Should help to debug graph allocation problems on CUDA.
shuttie pushed 3 commits to master nixiesearch/nixiesearch
shuttie opened a pull request on nixiesearch/nixiesearch
enable ONNX verbose logs when global loglevel is verbose
Should help to debug graph allocation problems on CUDA.
shuttie created a branch on nixiesearch/nixiesearch
fix/onnx-verbose-logs - Hybrid search engine, combining best features of text and semantic search worlds
shuttie opened an issue on nixiesearch/nixiesearch
Expose in config GPU allocation
Right now llamacpp uses all GPUs, and onnx - the first one by default. In practice we should have GPU related options in config: * which GPUs to use * there should be an option to force model j...
shuttie pushed 1 commit to fix/llamacpp-expwait nixiesearch/nixiesearch
- add log entry when llamacpp started 76eb80e
shuttie created a branch on nixiesearch/nixiesearch
fix/llamacpp-expwait - Hybrid search engine, combining best features of text and semantic search worlds
shuttie pushed 2 commits to master nixiesearch/nixiesearch
shuttie closed a pull request on nixiesearch/nixiesearch
fix crash in readiness probe when llamacpp takes some time to spin up
Trace: ``` 17:51:53.251 WARN o.h.ember.client.EmberClientBuilder - timeout (120 seconds) is >= idleConnectionTime (60 seconds). It is recommended to configure timeout < idleConnectionTime, or ...