{"about":{"non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","site":"https://codewithpapers.app","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page","syntology":{"site":"https://syntology.ai","developers":"https://syntology.ai/developers","mcp":{"server":"https://syntology.ai/mcp","transport":"streamable-http","server_card":"https://syntology.ai/.well-known/mcp/server-card.json","auth":{"type":"trial token, no account","trial_token":"https://syntology.ai/api/oauth/trial/token","method":"POST","docs":"https://syntology.ai/developers"}},"have":"https://syntology.ai/api/graph/have?x=<method, arXiv id or title> (free, answers coverage only)","paper_base":"https://syntology.ai/paper/","atlas_base":"https://app.syntology.ai/?focus="},"machine_readable":[{"url":"https://codewithpapers.app/llms.txt","what":"the machine catalog: every machine-readable file, counted"},{"url":"https://codewithpapers.app/index/manifest.json","what":"paper-to-code index by arXiv id, with Syntology's counts"},{"url":"https://codewithpapers.app/search/manifest.json","what":"site search index (titles, authors) and its files"},{"url":"https://codewithpapers.app/download","what":"bulk files: Syntology's layer, described there"},{"url":"https://codewithpapers.app/build_manifest.json","what":"the build record: inputs, counts, exclusions, probes"}]},"url":"/paper/flashattention-2-faster-attention-with-better","title":"FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning","arxiv_id":"2307.08691","date":"2023-07-17","proceeding":null,"authors":["Tri Dao"],"abstract":"Scaling Transformers to longer sequence lengths has been a major problem in the last several years, promising to improve performance in language modeling and high-resolution image understanding, as well as to unlock new applications in code, audio, and video generation. The attention layer is the main bottleneck in scaling to longer sequences, as its runtime and memory increase quadratically in the sequence length. FlashAttention exploits the asymmetric GPU memory hierarchy to bring significant memory saving (linear instead of quadratic) and runtime speedup (2-4$\\times$ compared to optimized baselines), with no approximation. However, FlashAttention is still not nearly as fast as optimized matrix-multiply (GEMM) operations, reaching only 25-40\\% of the theoretical maximum FLOPs/s. We observe that the inefficiency is due to suboptimal work partitioning between different thread blocks and warps on the GPU, causing either low-occupancy or unnecessary shared memory reads/writes. We propose FlashAttention-2, with better work partitioning to address these issues. In particular, we (1) tweak the algorithm to reduce the number of non-matmul FLOPs (2) parallelize the attention computation, even for a single head, across different thread blocks to increase occupancy, and (3) within each thread block, distribute the work between warps to reduce communication through shared memory. These yield around 2$\\times$ speedup compared to FlashAttention, reaching 50-73\\% of the theoretical maximum FLOPs/s on A100 and getting close to the efficiency of GEMM operations. We empirically validate that when used end-to-end to train GPT-style models, FlashAttention-2 reaches training speed of up to 225 TFLOPs/s per A100 GPU (72\\% model FLOPs utilization).","url_abs":"https://arxiv.org/abs/2307.08691v1","url_pdf":"https://arxiv.org/pdf/2307.08691v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"flashattention-2-faster-attention-with-better","repo_url":"https://github.com/dao-ailab/flash-attention","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"BSD-3-Clause"}},{"paper_slug":"flashattention-2-faster-attention-with-better","repo_url":"https://github.com/ROCm/flash-attention","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"BSD-3-Clause"}},{"paper_slug":"flashattention-2-faster-attention-with-better","repo_url":"https://github.com/facebookresearch/xformers/blob/main/xformers/ops/fmha/flash.py","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null},{"paper_slug":"flashattention-2-faster-attention-with-better","repo_url":"https://github.com/jundaf2/INT8-Flash-Attention-FMHA-Quantization","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"flashattention-2-faster-attention-with-better","repo_url":"https://github.com/pytorch/pytorch/blob/main/torch/nn/functional.py","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"jax","reach":null},{"paper_slug":"flashattention-2-faster-attention-with-better","repo_url":"https://github.com/tlc-pack/libflash_attn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok","spdx":"BSD-3-Clause"}}],"tasks":[{"task_slug":null,"task_name":"GPU"},{"task_slug":"language-modeling","task_name":"Language Modeling"},{"task_slug":"language-modelling","task_name":"Language Modelling"}],"methods":[{"method_slug":"speed","method_name":"SPEED"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2307.08691","atlas_url":"https://app.syntology.ai/?focus=2307.08691","mcp":null,"developers":"https://syntology.ai/developers","held":true,"reason":"arxiv_id_in_atlas","checked_against":{"atlas_ids":549449,"atlas_ids_sha256":"afd72399dafd9906e9062927f70655952385a3d75488558e1f0484e950ab09b7","read_at":"2026-10-05T01:46:21Z"},"agent_calls":[{"tool":"get_citation_path","arguments":{"paper_1":"2307.08691"},"arguments_in":{"paper_2":"another paper's arXiv id or title"},"call":"get_citation_path(paper_1=\"2307.08691\", paper_2=\"…\")"},{"tool":"get_concepts_for_paper","arguments":{"arxiv_id":"2307.08691"},"call":"get_concepts_for_paper(arxiv_id=\"2307.08691\")"}],"by_repo":[{"repo":"dao-ailab/flash-attention","kind":"official","samples":0,"ran":0,"instrument":0,"unverified":0,"harvested_for_other_papers":2,"harvested_for":["2407.08608","2604.23798"],"harvested_for_calls":[{"arxiv_id":"2407.08608","tool":"get_harvested_code_for_paper","arguments":{"arxiv_id":"2407.08608"},"call":"get_harvested_code_for_paper(arxiv_id=\"2407.08608\")"},{"arxiv_id":"2604.23798","tool":"get_harvested_code_for_paper","arguments":{"arxiv_id":"2604.23798"},"call":"get_harvested_code_for_paper(arxiv_id=\"2604.23798\")"}]},{"repo":"ROCm/flash-attention","kind":"listed","samples":0,"ran":0,"instrument":0,"unverified":0,"harvested_for_other_papers":1,"harvested_for":["2309.06180"],"harvested_for_calls":[{"arxiv_id":"2309.06180","tool":"get_harvested_code_for_paper","arguments":{"arxiv_id":"2309.06180"},"call":"get_harvested_code_for_paper(arxiv_id=\"2309.06180\")"}]},{"repo":"facebookresearch/xformers","kind":"listed","samples":0,"ran":0,"instrument":0,"unverified":0,"harvested_for_other_papers":2,"harvested_for":["2004.05150","2112.05682"],"harvested_for_calls":[{"arxiv_id":"2004.05150","tool":"get_harvested_code_for_paper","arguments":{"arxiv_id":"2004.05150"},"call":"get_harvested_code_for_paper(arxiv_id=\"2004.05150\")"},{"arxiv_id":"2112.05682","tool":"get_harvested_code_for_paper","arguments":{"arxiv_id":"2112.05682"},"call":"get_harvested_code_for_paper(arxiv_id=\"2112.05682\")"}]},{"repo":"jundaf2/INT8-Flash-Attention-FMHA-Quantization","kind":"listed","samples":0,"ran":0,"instrument":0,"unverified":0},{"repo":"pytorch/pytorch","kind":"listed","samples":0,"ran":0,"instrument":0,"unverified":0,"harvested_for_other_papers":1,"harvested_for":["1912.01703"],"harvested_for_calls":[{"arxiv_id":"1912.01703","tool":"get_harvested_code_for_paper","arguments":{"arxiv_id":"1912.01703"},"call":"get_harvested_code_for_paper(arxiv_id=\"1912.01703\")"}]},{"repo":"tlc-pack/libflash_attn","kind":"listed","samples":0,"ran":0,"instrument":0,"unverified":0}],"by_repo_harvested_for_is":"harvested_for: the arXiv ids (as Syntology's graph holds them) of every paper whose samples Syntology's graph links to this repository, in arXiv-id order, complete; the page names the first 3 and says how many more. A link only: it does not say the papers are the same work, that one copied the other, or that the code implements this paper's method; harvested_for_calls gives the get_harvested_code_for_paper call for each arXiv-shaped id among them","by_repo_is":"one row per repository in the archive's code links, Syntology's graph links or the samples; samples 0 means no sample is linked to this paper, a statement about Syntology's coverage, not about the repository: with harvested_for_other_papers present, Syntology harvested the repository for that many other papers, and without it, Syntology harvested nothing from it; placed_by_identical_code counts samples whose link records no repository, placed here because identical code was harvested from this repository for another paper; ran counts executions on a synthesized input and instrument counts failures of Syntology's instrument, not of the code"},"code_links_note":null,"arxiv_metadata":null,"syntology_extracted_results":null,"record_sha256":"3bc0c86baa21261ce44775090bd019ef56dbd768895edca2daca83ccf7c4e581","record_changed_at":"2026-09-28","record_changed_at_basis":"first_hashed"}