[oss] Init gpt-oss bf16 support (#22508)

Jee Jee Li committed over 1 year ago

[V1] [Hybrid] Enable Full CUDA Graph (decode-only) for Mamba layers (#21401)

Thomas Parnell committed over 1 year ago

[FEAT] [Performance] Add triton mrope to replace the torch code path (#22375)

TJian committed over 1 year ago

[Bugfix] Fix basic models tests hanging due to mm processor creation (#22571)

Isotr0py committed over 1 year ago

[Model] Gemma3n MM (#20495)

Nicolò Lucchesi committed over 1 year ago

Move `ParallelConfig` from `config/__init__.py` to `config/parallel.py` (#22565)

Harry Mellor committed over 1 year ago

[Docs] Reduce noise in docs and `--help` from the JSON tip (#22567)

Harry Mellor committed over 1 year ago

[CI] [Hybrid] Speed up hybrid models test by removing large models (#22563)

Thomas Parnell committed over 1 year ago

GLM-4.5V with new class name at transformers (#22520)

Yuxuan Zhang committed over 1 year ago

Update docs for Minimax-Text support (#22562)

Thomas Parnell committed over 1 year ago

[Bugfix] Fix CI moe kernel failure (#22556)

Jee Jee Li committed over 1 year ago

[Bugfix] Fix failing GPT-OSS initialization test (#22557)

Isotr0py committed over 1 year ago

[ROCm][Misc] Rename the context_len to seq_len in ROCm custom paged attention kernel (#22097)

Charlie Fu committed over 1 year ago

[TPU] Add support for online w8a8 quantization (#22425)

Kyuyeun Kim committed over 1 year ago

Fix loading of quantized BigCode models (#22463)

Eldar Kurtić committed over 1 year ago

[Misc] Use config definitions from Transformers library (#21913)

Cyrus Leung committed over 1 year ago

v1: Pass KVConnectorOutput to scheduler-side (#22157)

Or Ozeri committed over 1 year ago

[V1] [Hybrid] Support Minimax-Text-01 in V1 (#22151)

Thomas Parnell committed over 1 year ago

[Log] Add Warning for Deprecation of DeepGEMM old version (#22194)

Wentao Ye committed over 1 year ago

Remove mamba_ssm from vLLM requirements; install inside test container using `--no-build-isolation` (#22541)

Thomas Parnell committed over 1 year ago