Search results

FlexAttention + custom score_mod fails under torch.compile (FlexibleLayout / NoValidChoicesError)

Status: Open.

#166 In meta-pytorch/attention-gym;

· v1kstrand opened on Sep 18, 2025

[Question] src/tcmalloc.cc:283] Attempt to free invalid pointer 0x13d801860

Status: Open.

#161 In meta-pytorch/attention-gym;

· PoisonKiller opened on Aug 14, 2025

<code>_compile=True</code> blows up GPU memory when seq length / block_mask change, but is best when both are fixed (FlexAttention)

Status: Open.

#159 In meta-pytorch/attention-gym;

· wileewang opened on Jul 22, 2025

[Question] sliding window attention + dynamic input shapes

Status: Open.

#138 In meta-pytorch/attention-gym;

· pfeatherstone opened on May 6, 2025

[Question] Precision of bfloat16 computations

Status: Open.

#131 In meta-pytorch/attention-gym;

· lukasHoel opened on Apr 4, 2025

RuntimeError in flex_attention: CUDA Device-Side Assertion Failure

Status: Open.

#130 In meta-pytorch/attention-gym;

· NuanBaobao opened on Mar 21, 2025

Would FlexAttention be useful over SDPA for masked language modeling?

Status: Open.

#129 In meta-pytorch/attention-gym;

· abdulfatir opened on Mar 20, 2025

Add Block Diffusion Attention Mask to attn_gym/masks

Status: Open.

#126 In meta-pytorch/attention-gym;

· Skylion007 opened on Mar 13, 2025

Hard-coded max_seq_len in document example

Status: Open.

#124 In meta-pytorch/attention-gym;

· adamoyoung opened on Mar 1, 2025

BlockMask Value Error

Status: Open.

#122 In meta-pytorch/attention-gym;

· Leo-T-Zang opened on Feb 17, 2025

Use case against efficient SDPA backend

Status: Open.

#121 In meta-pytorch/attention-gym;

· TParcollet opened on Feb 13, 2025

FlexAttention customizability for softmax

Status: Open.

#117 In meta-pytorch/attention-gym;

· veritas9872 opened on Feb 9, 2025