github tensorflow/serving 2.12.0

latest releases: 2.17.0, 2.16.1, 2.15.1...
17 months ago

Release 2.12.0

  • NOTE: 2.12.0 has been identified as bad release. Please use 2.12.1 or later instead.

Major Features and Improvements

  • No major features or improvements.

Breaking Changes

  • Users of remote_predict_py should stay on v2.11 and avoid v2.12.

Bug Fixes and Other Changes

  • Update TF Serving to Bazel 5.3.0, to match with TF. (commit: cda26f6)
  • Update TF Text to v2.11.0. (commit: 1624fb2)
  • Add pyclif_proto_library for get_model_metadata and session_service (commit: bedf391)
  • Raise the vlog level about aspired versions (commit: 73746fb)
  • Add PredictStreamed to PredictionLog. It represents a logged stream of PredictRequests and PredictResponses. (commit: 557f68a)
  • Update rules_pkg to 0.7.1. (commit: 2b5ad9a)
  • Track additional metadata for request logs. (commit: e0a3b5f)
  • Replace usage of the tsl::Status constructor with a tsl::{error, errors}::Code. (commit: cfb9fb2)
  • Replace usage of the tsl::Status constructor with a tsl::{error, errors}::Code. (commit: 19345a6)
  • Update Dockerfile.devel* with py3.8 installed. (commit: 68d92ff)
  • Update TF Text to v2.12.0. (commit: dbe9339)
  • Stop depend on 'tensorflow-gpu' for tensorflow-serving GPU build on master branch. (commit: 85ff9c0)
  • Upgrade cuda from 11.2 to 11.8. (commit: 20f51c9)
  • Upgrade to decision-forests-1.3.0 and yggdrasil-decision-forests-1.4.0 (commit: 45d1574)
  • Add tool to send predict (grpc) requests to ModelServer. (commit: 25c5125)
  • Ignore remote_predict from :all for 2.12 release due to upstream TF breakage. (commit: 5830714)

Don't miss a new serving release

NewReleases is sending notifications on new releases.