-
Notifications
You must be signed in to change notification settings - Fork 249
Pull requests: google/tunix
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Supported mixed precision training in qwen2
#1199
opened Mar 5, 2026 by
copybara-service
bot
Loading…
Refactor BackendMappingMixin to use explicit BACKEND_PACKAGE_PATH
#1198
opened Mar 5, 2026 by
copybara-service
bot
Loading…
Harden calculate_reward: replace unsafe eval() with safe arithmetic evaluator
#1197
opened Mar 5, 2026 by
0xmrma
Loading…
make perf engine (v1 or v2) and export function selectable.
#1191
opened Mar 3, 2026 by
copybara-service
bot
Loading…
Extending scanned to unscanned support for GPT-OSS like models.
#1189
opened Mar 3, 2026 by
NicoGrande
Loading…
6 tasks done
Fix sharding rank mismatch for LoRA weights and optimizer states in
qlora_gemma.ipynb notebook
#1186
opened Mar 3, 2026 by
rajasekharporeddy
Loading…
4 tasks done
Fix sharding rank mismatch for LoRA weights and optimizer states in
grpo_gemma.ipynb notebook
#1180
opened Mar 2, 2026 by
rajasekharporeddy
Loading…
4 tasks done
fix(rl): flush final partial micro-batch to avoid sample drop
#1179
opened Mar 1, 2026 by
pjo256
Loading…
6 tasks done
[Tunix] Clear and reinitialize vLLM KV cache when loading weights.
#1167
opened Feb 26, 2026 by
copybara-service
bot
Loading…
fix: add thread safety locks to perf tracing module
#1166
opened Feb 26, 2026 by
kbhujbal
Loading…
6 tasks
feat: add DFT loss and refactor SFT losses into submodule
#1153
opened Feb 25, 2026 by
kryvokhyzha
Loading…
5 of 6 tasks
Previous Next
ProTip!
Find all pull requests that aren't related to any open issues with -linked:issue.