Revert "[PERF] Use faster way of decode in tokenizer: avoid useless list-to-list conversion (#20000)" (#23396)

Cyrus Leung committed about 1 year ago

Add glm4.5v tp2,4 fp8 config on H100_80GB (#23443)

Chenxi Yang committed about 1 year ago

Quantization: support FP4 quantized models on AMD CDNA2/CDNA3 GPUs (#22527)

Daifeng Li committed about 1 year ago

[UX] Move Dockerfile DeepGEMM install to tools/install_deepgemm.sh (#23360)

Michael Goin committed about 1 year ago

[Model] Support DP for ViT on MiniCPM-V-4 (#23327)

WeiQing Chen committed about 1 year ago

[BugFix] Fix `MinPLogitsProcessor.update_states()` (#23401)

Nick Hill committed about 1 year ago

[BugFix] Fix batch updates for pooling models (#23398)

Nick Hill committed about 1 year ago

[NVIDIA][torch.compile] Support Flashinfer TRTLLM FP8-q/kv NVFP4-out Attention Kernel (#22703)

elvischenv committed about 1 year ago

[BugFix][AMD][Quantization] Fix torch.compile issue where wvSplitKQ not being called when it should when using quantized FP8 model (#22281)

rasmith committed about 1 year ago

[PERF] PyTorch Symmetric Memory All-Reduce (#20759)

Ilya Markov committed about 1 year ago

[CI/Build] add EP dependencies to docker (#21976)

Zhewen Li committed about 1 year ago

add an env var for path to pre-downloaded flashinfer cubin files (#22675)

Shiyan Deng committed about 1 year ago

[Quantization] Allow GGUF quantization to skip unquantized layer (#23188)

Isotr0py committed about 1 year ago

[Bugfix] Fix broken Florence-2 model (#23426)

Isotr0py committed about 1 year ago

[Model] Add Ovis2.5 PP support (#23405)

Isotr0py committed about 1 year ago

[Doc]: fix various typos in multiple files (#23179)

Didier Durand committed about 1 year ago

Add unit tests for batched guided and non-guided requests (#23389)

Yong Hoon Shin committed about 1 year ago

fix(tests): Ensure reliable CUDA cache clearing in MoE test (#23416)

Aziz committed about 1 year ago

[BugFix] Fix the issue where image embeddings were incorrectly split.… (#23366)

bppps committed about 1 year ago

[Speculators][Speculative Decoding] Fix Qwen 2 Eagle3 Support (#23337)

PapaGoose committed about 1 year ago