Releases: NVIDIA/FasterTransformer
Releases · NVIDIA/FasterTransformer
v5.3 release
release/v5.3_tag Update CMakeLists.txt
release/v5.2.1
Fix some bugs of v5.2
release/v5.2_bug_fix
release/v5.2_bug_fix_tag fix: fix fmha kernel assert bug
v5.2 release
release/v5.2_tag fix: add cutlass submodule
v5.1.1 bug fix
- fix stop criterion.
- fix bug of attention mask chosen when enabling shared context opt
- fix swin qk scale
- fix bug of repetition penalty of t5 under beam search
- fix bug of gpt_guide.md
- fix bug of decoder_masked_multihead_attention_template
v5.1 T5 triton bug fix
Fix the bug of model parallelism setting of T5 on v5.1
v5.1 release
release/v5.1_tag feat: update v5.1 (#281)
v5.0 release
release/v5.0_tag feat: update v5.0
release/v1.0_tag: Merge pull request #123 from pmixer/del_v1_duplicated_files
Del v1 duplicated files
release/v4.0_tag: Merge pull request #54 from NVIDIA/main
Update the v4.0 with new modification