updated flash_attn GIT_TAG to point to torch Stable ABI FA3 commit (#47995)

Chris Leonard committed 4 months ago

[CI] BugFix Eval Small Models Distributed test for DiffusionGemma (#47980)

Ilya Markov committed 4 months ago

[Bugfix] Re-enable benchmarking of librispeech dataset. (#47033)

almayne committed 4 months ago

[Bugfix][DCP] Cast LSE to fp32 in a2a combine to fix bf16 bitcast crash (#47801)

shawn committed 4 months ago

Fix FlashAttention MLA prefill V unpadding (#42642)

Martin Vit committed 4 months ago

[Spec Decode] Support hybrid (SWA + full attention) DFlash drafters (#47914)

Michael Goin committed 4 months ago

Fix NVML capability lookup for visible devices (#47892)

Tyler Michael Smith committed 4 months ago

[Bug] Fix Batched DeepGEMM (#47884)

Robert Shaw committed 4 months ago

Improvement of Docker image build for IBM Power using prebuilt wheels from IBM published devpi index (#46017)

Vivek Sharma committed 4 months ago

[Bugfix][V1] Free out-of-window blocks on the processed-token basis under async scheduling (#47728)

Saddss committed 4 months ago

[CI/Build][BugFix][The Rock] Fix get_ssm_device_name to return sanitized, usable filename (#47781)

rasmith committed 4 months ago

[Rust Frontend] Handle `continue_final_message` with renderer sentinel (#47844)

Bugen Zhao committed 4 months ago

Add tuned selective_state_update float32 config for AMD Instinct MI355 (#47943)

vanshbhatia-amd committed 4 months ago

[Bugfix] Allocate HY V3 expert_bias in float32 to prevent silent downcasting (#47797)

aoright committed 4 months ago

Remove router weight upcast for DSv2-related models (#47970)

Thien Tran committed 4 months ago

[ROCm] Add tuned selective_state_update float32 config for AMD Instinct MI300X (#47947)

vanshbhatia-amd committed 4 months ago

[Model] Support MOSS-Transcribe-Diarize (#47729)

Canlin Guo committed 4 months ago

[ROCm] Add tuned selective_state_update float16 config for AMD Instinct MI300X (#47945)

vanshbhatia-amd committed 4 months ago

[Frontend] Support bad_words in the /v1/completions endpoint (#46793)

sungbin1015 committed 4 months ago

[Misc][Docs] Add human-readable integer support for more cli-args (#47608)

Nicolò Lucchesi committed 4 months ago