#bf16
Live, measured metrics for the hashtag #bf16 from the open social web. Every number carries a named source and the time it was fetched. Nothing is estimated.
Own #bf16
This #name is available to claim. It becomes your portal on the open agent web: this very page, a keyword you rank for by an open public stake, and a verifiable identity for AI agents. Nobody else sells a page like this for every #name.
Day-by-day usage
measured · fosstodon.org (Mastodon public tags API) · fetched 2026-07-30 02:42 UTC0 uses by 0 unique accounts across the window. Real per-day counts, not estimates. Newest bar is today so far.
Related hashtags
measured · fosstodon.org (Mastodon public search API) · fetched 2026-07-30 02:42 UTCLive pulse
measured · fosstodon.org (Mastodon tag timeline) · fetched 2026-07-30 02:42 UTCEverything below is measured over the latest 4 public posts (spanning ~14610 hours).
Posting hours (UTC)
Languages: English (2) · Chinese (Taiwan) (1) · Russian (1)
Avg boosts / post: 0
Top of the latest posts
🌘 偏差累積與方差抵銷：隨機捨入在模型訓練中的關鍵作用 ➤ 透過隨機捨入突破低精度計算的效能瓶頸 ✤ https://convergentthinking.sh/posts/bias-compounds-variance-washes-out/ 在神經網絡訓練中，數值精度往往受限於浮點數格式。作者指出，傳統的「四捨五入」（Round-to-Nearest, RNE）在進行微小數值累加時，會因為固定偏差而導致誤差累積，使得模型訓練停滯；相較之下，「隨機捨入」（Stochast
Clément Pillette (@ClementPillette) kim-dev 72B를 BF16으로 2 GPU 병렬화하는 시도는 다소 무리였고, 대신 AWQ 4-bit 양자화를 시도한다고 보고합니다. MLX 팀(특히 @ivanfioravanti) 덕분에 Mac Studio에서 모델 구동이 훨씬 수월해졌고, Minimax 2.5는 8비트에서 초당 30tps로 잘 동작하고 있다는 실무적 성과를 공유한 트윗입니다. https://x
New DFloat11 Technique Offers 30% Lossless Compression for LLMs, Easing Hardware Demands #AI #AIResearch #DFloat11 #LLMs #LLMcompression #MachineLearning #DeepLearning #BF16 #Inference #RiceUniversity #xMADai https://winbuzzer.com/2025/04/2
Every number above is measured from a named public API at the shown fetch time. Nothing is estimated or extrapolated. Platforms that lock their data behind paid APIs are not shown. Agents: the same numbers, as JSON, at /api/hashtags/bf16