Update flash-linear-attention requirement from ==0.4.* to ==0.5.* in /requirements/full

[dependabot]

Updates the requirements on flash-linear-attention to permit the latest version.

Release notes

Sourced from flash-linear-attention's releases.

v0.5.0

✨ Highlights

The main new model this release is MoBA / FlashMoBA (#840, #845), which brings Moonshot's Mixture of Block Attention into the fla family.

Beyond new models, fla is also moving from a Triton-only stack to a multi-backend one: FlashKDA is added as a new backend for KDA (#852), and TileLang is introduced for GDN, KDA, and parallel attention kernels (#827, #846, #854), with more backends to come.

What's Changed

• [CP] Fix missing bos and i_h offsets in backward gk loads by @​zhiyuan1i in fla-org/flash-linear-attention#781
• [KDA] Clarify gate input tracking in chunk backward by @​zhiyuan1i in fla-org/flash-linear-attention#785
• [GDN] Fuse kkt + solve_tril kernel & unified benchmark infrastructure by @​yzhangcs in fla-org/flash-linear-attention#789
• [Conv] Fix int32 overflow in conv kernel pointer arithmetic for large tensors by @​tmct in fla-org/flash-linear-attention#783
• [GDN] Add exp2 support across chunk kernels for improved performance by @​yzhangcs in fla-org/flash-linear-attention#791
• Fix parameter initialization for FSDP meta device compatibility by @​yzhangcs in fla-org/flash-linear-attention#793
• [GDN] Fix missing mask on off-diagonal blocks in fused kkt+s… by @​yzhangcs in fla-org/flash-linear-attention#794
• Fix layer_norm_bwd_kernel OOB access on high-SM GPUs by @​mpurland in fla-org/flash-linear-attention#795
• [Misc] Upgrade minimum PyTorch requirement to 2.7.0 by @​zhiyuan1i in fla-org/flash-linear-attention#801
• [Conv] Fix int32 overflow in varlen conv kernel pointer arithmetic by @​tmct in fla-org/flash-linear-attention#803
• [GDN] Add GVA support by @​zhiyuan1i in fla-org/flash-linear-attention#799
• [BugFix] Fix illegal memory access in KDA backward by dropping buggy autotune configs on Hopper by @​zhiyuan1i in fla-org/flash-linear-attention#807
• [CE] Add logit softcapping support to fused cross entropy by @​yzhangcs in fla-org/flash-linear-attention#810
• [Mamba] Remove unused arguments and update to align with mamba_ssm by @​PuR3Luck in fla-org/flash-linear-attention#782
• [GDN] Native GVA support: remove redundant Q/K repeat and unify head naming by @​yzhangcs in fla-org/flash-linear-attention#812
• [KDA] Add safe_gate/lower_bound support and improve docstrings by @​yzhangcs in fla-org/flash-linear-attention#814
• [GDN] Add fused gate kernel with use_gate_in_kernel support by @​yzhangcs in fla-org/flash-linear-attention#813
• chore: add AUTHORS, unify copyright headers, and add CI workflows by @​yzhangcs in fla-org/flash-linear-attention#816
• [CI] Improve benchmark outputs by @​yzhangcs in fla-org/flash-linear-attention#817
• [CI] Fix skip-test check failing on fork PRs by @​zhiyuan1i in fla-org/flash-linear-attention#821
• [CP] Enable KCP for DPLR by @​zhiyuan1i in fla-org/flash-linear-attention#822
• [Fix] Guard checkpoint weight re-initialization in RWKV-7, Mamba, Mamba2, and LogLinearMamba2 by @​puigde in fla-org/flash-linear-attention#820
• fix: register default global_scratch allocator on Blackwell GPUs by @​ssubbotin in fla-org/flash-linear-attention#825
• [Attn] Add sliding window attention support by @​yzhangcs in fla-org/flash-linear-attention#824
• [Docs] Add CONTRIBUTING.md by @​yzhangcs in fla-org/flash-linear-attention#830
• allow neg eigvals for delta-net by @​hoedt in fla-org/flash-linear-attention#832
• chore: add standalone isort config by @​yzhangcs in fla-org/flash-linear-attention#834
• Add autotune for causal conv update by @​MARD1NO in fla-org/flash-linear-attention#828
• [GDN] Add TileLang backend for chunk_bwd_dqkwg kernel by @​zhiyuan1i in fla-org/flash-linear-attention#827
• [Refactor] Simplify TileLang backend directory structure by @​yzhangcs in fla-org/flash-linear-attention#835
• [CI] Post benchmark comment via workflow_run for fork-safe PRs by @​yzhangcs in fla-org/flash-linear-attention#841
• [KDA] Add Grouped Value Attention (GVA) support by @​yzhangcs in fla-org/flash-linear-attention#833
• [Attn] Add GPT-OSS-style attention sink support by @​Shomvel in fla-org/flash-linear-attention#831
• [Fix] respect user-provided cu_seqlens when attention_mask is present by @​yzhangcs in fla-org/flash-linear-attention#842
• [Fix] flatten batched qkv in varlen cu_seqlens path by @​lxr-tech in fla-org/flash-linear-attention#839
• [Fix] Enforce batch-size check for varlen mode across multiple ops by @​zhiyuan1i in fla-org/flash-linear-attention#844
• [MoBA] Integrate MOBA and FlashMOBA by @​ReyJerry in fla-org/flash-linear-attention#840
• [MoBA] Follow-up: fix broken import, rename layer, add modeling, tests & docs by @​yzhangcs in fla-org/flash-linear-attention#845
• [GDN/KDA] Fuse gate activation into fused_recurrent kernels by @​yzhangcs in fla-org/flash-linear-attention#848
• [TileLang] Add fwd/bwd kernel for parallel attention by @​zhiyuan1i in fla-org/flash-linear-attention#846
• [CI] Add issue/PR pytest command workflow by @​zhiyuan1i in fla-org/flash-linear-attention#843
• [GDN] Optimize b_dg computation in chunk_bwd_kernel_dqkwg#USE_G by @​MzeroMiko in fla-org/flash-linear-attention#823

... (truncated)

Commits

• 3a9ce1c [KDA][TileLang] Add TileLang backend for chunk_kda_bwd_wy_dqkg_fused (#854)
• 97aaf76 [KDA] Support FLASHKDA backend skip test
• 1d4d88f [Cleanup] Remove deprecated head_first parameter from public ops (#853)
• 12e2a70 [Bug] Fix incorrect cumsum dim for naive_chunk_linear_attn normalize (fixes #...
• 0b7fb85 [Linear Attention] Update fused_recurrent.py for inference with normalization...
• 00b5a0e [GDN][TileLang] Optimize b_dg computation in chunk_bwd_kernel_dqkwg (#849)
• cf8d1dd [GDN] Optimize b_dg computation in chunk_bwd_kernel_dqkwg (#823)
• e4a7874 [README] Polish wording, fix broken tip block, add MLA entry
• 265da34 [CI] Add issue/PR pytest command workflow (#843)
• d1a88fc [TileLang] Add fwd/bwd kernel for parallel attention (#846)
• Additional commits viewable in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

---

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)