-
Notifications
You must be signed in to change notification settings - Fork 10
Pull requests: google/tpu-raiden
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
The WeightSynchronizer FFI derived the host staging slot as
shard_idx % num_shards, which only works if a slice's device ordinals are contiguous and aligned to a multiple of num_shards. Device ordinals are not guaranteed to be, so pass the staging slot explicitly as a range-checked logical_shard_idx input instead of deriving it from the device ordinal.
#386
opened Jul 21, 2026 by
copybara-service
Bot
Loading…
The WeightSynchronizer FFI derived the host staging slot as
shard_idx % num_shards, which only works if a slice's device ordinals are contiguous and aligned to a multiple of num_shards. Device ordinals are not guaranteed to be, so pass the staging slot explicitly as a range-checked logical_shard_idx input instead of deriving it from the device ordinal.
#385
opened Jul 21, 2026 by
copybara-service
Bot
Loading…
[TPU Raiden] Test BAP regression gate skipping (Expected to pass)
#384
opened Jul 21, 2026 by
copybara-service
Bot
Loading…
Deprecate legacy raw-offset TransferBuffers overloads in RaidenController.
#382
opened Jul 20, 2026 by
copybara-service
Bot
Loading…
[TPU Raiden] Test BAP regression gate (Expected to fail)
#381
opened Jul 20, 2026 by
copybara-service
Bot
Loading…
[TPU Raiden] Test gating without skip tag to verify performance gate fails
#380
opened Jul 20, 2026 by
copybara-service
Bot
Loading…
Implement nested structural transfer payloads for multi-NUMA Raiden host mapping.
#379
opened Jul 20, 2026 by
copybara-service
Bot
Loading…
Refactor TPU Raiden API endpoint terminology and connection handling:
#377
opened Jul 20, 2026 by
copybara-service
Bot
Loading…
Fixed layout permutation and packing issues in weight synchronizer tiling.
#371
opened Jul 20, 2026 by
copybara-service
Bot
Loading…
[TPU Raiden] Inject latency into benchmark D2h for testing
#370
opened Jul 20, 2026 by
copybara-service
Bot
Loading…
[TPU Raiden] Test BAP regression gate skipping (Expected to pass)
#369
opened Jul 20, 2026 by
copybara-service
Bot
Loading…
kv_cache: reject non-row-addressable layouts and mixed-slice managers
#363
opened Jul 18, 2026 by
yinlin09
Collaborator
Loading…
[TPU Raiden] Inject latency into KVCacheManagerBase::D2h for testing
#354
opened Jul 17, 2026 by
copybara-service
Bot
Loading…
[tpu_raiden] refactor connection pool as a separate class
#352
opened Jul 16, 2026 by
copybara-service
Bot
Loading…
Update personal agent configurations for tpu_raiden migration.
#330
opened Jul 15, 2026 by
copybara-service
Bot
Loading…
feat: implement Raiden OTEL design and telemetry
#321
opened Jul 14, 2026 by
copybara-service
Bot
Loading…
Add pre-submit gating test (H2D/D2H/H2H)
#290
opened Jul 9, 2026 by
haochengd-google
Collaborator
Loading…
Add Global Prefix Caching design documentation for TPU Raiden.
#279
opened Jul 8, 2026 by
copybara-service
Bot
Loading…
Add KVCacheManager.read_host_bytes and write_host_bytes: length-bounded
#278
opened Jul 8, 2026 by
copybara-service
Bot
Loading…
Expose the optional
dst_block_ids parameter in the JAX KVCacheManager.h2h_write Python binding to align with the underlying C++ signature.
#274
opened Jul 7, 2026 by
copybara-service
Bot
Loading…
Implement remote-fetch destination auto-allocation.
#269
opened Jul 6, 2026 by
copybara-service
Bot
Loading…
Implement KVCacheStore::Evict (DRAM Cleanup) API.
#267
opened Jul 6, 2026 by
copybara-service
Bot
Loading…
Previous Next
ProTip!
Updated in the last three days: updated:>2026-07-17.