Skip to content

Pull requests: vllm-project/tpu-inference

New pull request New
123 Open 1,459 Closed
123 Open 1,459 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

[Jax] Add support for forward pass of channel-wise requantized Fp8FusedMoEMethod for GMM_TP/GMM_EP
#1655 opened Feb 7, 2026 by jrplatin • Draft
4
Remove _reqs_being_saved check from status clean up for the finished …
#1654 opened Feb 7, 2026 by dannawang0221 Loading…
1
[Multi-host] Generalize device_put for single and multi-host sharding ready ONLY add when PR is ready to merge/full CI is needed
#1653 opened Feb 7, 2026 by Lumosis Loading…
2
Update Dockerfile to allow automatically starting vLLM server ready ONLY add when PR is ready to merge/full CI is needed
#1651 opened Feb 6, 2026 by wenxindongwork Loading…
1
[wip] [do not review] batch load for tpu kv host offload connector
#1649 opened Feb 6, 2026 by saikat-royc Loading…
[AWQ][MoE] Support AWQ MoE
#1644 opened Feb 6, 2026 by catswe • Draft
Fix the vllm upstream break changes ready ONLY add when PR is ready to merge/full CI is needed
#1642 opened Feb 5, 2026 by mrjunwan-lang Loading…
2
[Jax/Flax Version Change] Update Jax to 0.8.1 and flax to 0.12.3 ready ONLY add when PR is ready to merge/full CI is needed
#1641 opened Feb 5, 2026 by patrickji2014 Loading…
1
[Qwen][FP8] Add test for pre-quantized FP8 qwen3 model ready ONLY add when PR is ready to merge/full CI is needed
#1640 opened Feb 5, 2026 by lk-chen Loading…
1
1
JAX FP8 Linear Infrastructure - Tensor Quantized ready ONLY add when PR is ready to merge/full CI is needed
#1638 opened Feb 5, 2026 by patemotter Loading…
13
[JAX][MoE] fix bug for SparseMoE
#1625 opened Feb 4, 2026 by bzgoogle Loading…
1
Massage some parameters to balance prefill/decode and improve disagg performance
#1622 opened Feb 4, 2026 by mrjunwan-lang Loading…
[Log] resolves confusion for ep test log ready ONLY add when PR is ready to merge/full CI is needed
#1619 opened Feb 4, 2026 by syhuang22 Loading…
3
[TPU host offload] batched save bulk transfer for tpu host offload connector
#1612 opened Feb 3, 2026 by saikat-royc Loading…
8
Split JAX unit tests into two parts with coverage ready ONLY add when PR is ready to merge/full CI is needed
#1609 opened Feb 3, 2026 by ernie-chang • Draft
Bump Jax to 0.9.0, flax to 0.12.3 and qwix to 0.1.4
#1598 opened Feb 2, 2026 by patrickji2014 Loading…
18
MLA Support for GLM 4.7 Flash
#1595 opened Feb 1, 2026 by mkurman Loading…
39
Support DeepSeek-V2-Lite-Chat-FP8
#1593 opened Feb 1, 2026 by catswe • Draft
[bug] resolve compatibility issues with vllm trunk
#1592 opened Feb 1, 2026 by bangshengtang Loading…
4
[Parallelism Support Matrix Tests] fix ep test with sharding additional_config ready ONLY add when PR is ready to merge/full CI is needed
#1589 opened Jan 31, 2026 by syhuang22 Loading…
override vllm ray worker wrapper
#1579 opened Jan 30, 2026 by Chenyaaang • Draft
1
Standardize Tuned Block Sizes via JSON ready ONLY add when PR is ready to merge/full CI is needed
#1578 opened Jan 30, 2026 by RobMulla Loading…
1 task done
1
Pin the vllm hash to a vllm hash to minitgate test issue. ready ONLY add when PR is ready to merge/full CI is needed
#1574 opened Jan 30, 2026 by QiliangCui Loading…
2
Fix/structured output batch alignment ready ONLY add when PR is ready to merge/full CI is needed
#1563 opened Jan 29, 2026 by XMaster96 Loading…
1
3
Update contributor guidelines to prefer Torchax implementation of models
#1558 opened Jan 29, 2026 by muskansh-google Loading…
ProTip! Add no:assignee to see everything that’s not assigned.