some tests here, keeping same data, recipe, LLM but swap out the vision encoder



they compared with ViT‑L/14 and SigLIP‑SO400, a fully convolutional ConvNeXT, and hybrid FastViT models

FastViT is like 8× smaller and 20× faster than ViT‑L/14 while staying just as smart
SWAP3.45%
VSN0.06%
post-image
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • 8
  • Repost
  • Share
Comment
0/400
degenonymousvip
· 6h ago
The effect is better than expected.
View OriginalReply0
CryptoPunstervip
· 6h ago
Thin as paper, fast as the wind
View OriginalReply0
ForkMongervip
· 09-02 14:43
The speed improvement is really amazing!
View OriginalReply0
PretendingToReadDocsvip
· 09-02 14:42
Small and fast are worth studying.
View OriginalReply0
MissedTheBoatvip
· 09-02 14:32
The cost performance is really high.
View OriginalReply0
SilentObservervip
· 09-02 14:31
The speed has increased a lot.
View OriginalReply0
OfflineValidatorvip
· 09-02 14:30
The performance improvement is really good.
View OriginalReply0
GasFeeLovervip
· 09-02 14:21
Performance doubled and efficiency improved
View OriginalReply0
Trade Crypto Anywhere Anytime
qrCode
Scan to download Gate App
Community
English
  • 简体中文
  • English
  • Tiếng Việt
  • 繁體中文
  • Español
  • Русский
  • Français (Afrique)
  • Português (Portugal)
  • Bahasa Indonesia
  • 日本語
  • بالعربية
  • Українська
  • Português (Brasil)