Pull requests: google/tunix
Pull requests list
[Tunix] Detach incompatible vLLM cleanup finalizers for non-torch models.
#2199
opened Sep 11, 2026 by
copybara-service
Bot
Loading…
[Tunix] Update vLLM mapping config in run_rollout_node.py.
#2198
opened Sep 11, 2026 by
copybara-service
Bot
Loading…
[Tunix] Preload Raiden in run_rollout_node.
#2197
opened Sep 11, 2026 by
copybara-service
Bot
Loading…
Support sequence packing in sub-batch checkpointing for the agentic RL learner.
#2195
opened Sep 11, 2026 by
copybara-service
Bot
Loading…
Introduce kv-cache coordinator for experimental v2 sampler
ready-to-submit
#2193
opened Sep 10, 2026 by
copybara-service
Bot
Loading…
[Tunix] Add sampler-trainer agreement metrics and importance sampling to the RL orchestrator.
#2192
opened Sep 10, 2026 by
copybara-service
Bot
Loading…
[Tunix] Add mock rollout and trainer worker nodes for distributed orchestrator testing.
#2190
opened Sep 10, 2026 by
copybara-service
Bot
Loading…
[Tunix DeepSWE] Optimize and rebuild Agentic Sandbox dynamic pre-warming iterator.
#2189
opened Sep 10, 2026 by
copybara-service
Bot
Loading…
Automated g4 rollback of changelist 979014468.
#2187
opened Sep 10, 2026 by
copybara-service
Bot
Loading…
[Tunix] Add support for computing per-token log probabilities and sampler-trainer agreement.
#2186
opened Sep 10, 2026 by
copybara-service
Bot
Loading…
Add Qwen3 model support and distributed RL fixes for Kubernetes runs.
#2183
opened Sep 10, 2026 by
copybara-service
Bot
Loading…
Add a multi-turn, stateful household task exploration environment and reinforcement learning training recipe for Tunix Agentic RL.
#2180
opened Sep 10, 2026 by
copybara-service
Bot
Loading…
[Tunix] Check on the
skip_jit and cache_nnx_graph params from kwargs. cache_nnx_graph can save the cpu host time.
#2178
opened Sep 10, 2026 by
copybara-service
Bot
Loading…
Add OpenTelemetry metrics export setup for experimental orchestrator:
#2174
opened Sep 10, 2026 by
copybara-service
Bot
Loading…
Forward max_response_length to GRPOAdapter in deepswe_dist example.
#2172
opened Sep 10, 2026 by
copybara-service
Bot
Loading…
[Tunix] add aux_metrics to peft_trainer_v2 the metrics logger
#2171
opened Sep 10, 2026 by
copybara-service
Bot
Loading…
Implement a new profiler for multi-host RL topology (Tunix).
#2170
opened Sep 9, 2026 by
copybara-service
Bot
Loading…
Add various improvements to train_maxtext_nb.py.
#2169
opened Sep 9, 2026 by
niting
Contributor
Loading…
6 tasks done
Add ML Diagnostics scalar backend for Tunix RL.
#2165
opened Sep 9, 2026 by
copybara-service
Bot
Loading…
Support custom
kv_cache_size in base RL pipeline.
#2163
opened Sep 9, 2026 by
copybara-service
Bot
Loading…
Reuse DeepSWE recipe agent, env, and dataset loader in distributed DeepSWE.
#2161
opened Sep 9, 2026 by
copybara-service
Bot
Loading…
ProTip!
Find all pull requests that aren't related to any open issues with -linked:issue.