Figure 5d reports forward attention throughput on A100 80GB SXM4; PyTorch is marked OOM at sequence length 16k. The panel covers causal attention with head dimension 128.

Figure 5d reports forward attention throughput on A100 80GB SXM4; PyTorch is marked OOM at sequence length 16k. The panel covers causal attention with head dimension 128. · CiteArk

Measurement results

0/29 assessed

0 supported0 disputed or conflicting29 not assessed

The complete result group stays with its claim. Assessed rows come first; their judgments do not replace the whole-claim conclusion. Expand a row for conditions, sources, and evidence history.

Each row is a measurement with its reported value, latest available observation, and cumulative evidence state.
Metric and conditions	Reported	Latest observation	Evidence judgment
method: PyTorch · sequenceLength: 512	15 TFLOPs/s	—	Not assessed
method: FlashAttention · sequenceLength: 512	49 TFLOPs/s	—	Not assessed
method: xFormers · sequenceLength: 512	79 TFLOPs/s	—	Not assessed
method: FlashAttention Triton · sequenceLength: 512	89 TFLOPs/s	—	Not assessed
method: FlashAttention-2 · sequenceLength: 512	132 TFLOPs/s	—	Not assessed
method: PyTorch · sequenceLength: 1k	18 TFLOPs/s	—	Not assessed
method: FlashAttention · sequenceLength: 1k	59 TFLOPs/s	—	Not assessed
method: xFormers · sequenceLength: 1k	95 TFLOPs/s	—	Not assessed
method: FlashAttention Triton · sequenceLength: 1k	108 TFLOPs/s	—	Not assessed
method: FlashAttention-2 · sequenceLength: 1k	168 TFLOPs/s	—	Not assessed
method: PyTorch · sequenceLength: 2k	19 TFLOPs/s	—	Not assessed
method: FlashAttention · sequenceLength: 2k	65 TFLOPs/s	—	Not assessed

Figure 5d reports forward attention throughput on A100 80GB SXM4; PyTorch is marked OOM at sequence length 16k. The panel covers causal attention with head dimension 128.

Reported and observed measurements

Assessments (0)

Runs (0)

Measurement results