Views
No views yet
instagram_usernamesnapchat_usernameyoutube_usernametwitch_usernametiktok_usernamediscord_usernamex_usernamecashapp_usernameonlyfans_usernametumblr_usernamegithub_usernamekofi_usernamepatreon_usernameroblox_usernamegeneric_usernamegeneric_username is a fallback for handle-shaped strings without a clear platform
indicator.1from gliner import GLiNER
2
3model = GLiNER.from_pretrained("LumeData/HandleAtlas-166m")
4
5labels = ['instagram_username', 'snapchat_username', 'youtube_username', 'twitch_username', 'tiktok_username', 'discord_username', 'x_username', 'cashapp_username', 'onlyfans_username', 'tumblr_username', 'github_username', 'kofi_username', 'patreon_username', 'roblox_username', 'generic_username']
6
7text = "Insta: foodgrammer | Snap: chefchef | DC: gamer420 | $cashtag"
8for ent in model.predict_entities(text, labels, threshold=0.5):
9 print(f"{ent['text']!r} -> {ent['label']} ({ent['score']:.2f})")urchade/gliner_small-v2.1Dc:/dc:/DC: etc.)| metric | value |
|---|---|
| precision | 0.849 |
| recall | 0.929 |
| F1 | 0.887 |
threshold=0.5generic_username, bump to 0.65 to reduce false positives; it's the
catch-all label and over-fires at the default threshold.discord_invite is not predicted — invite codes will be classified as
discord_username or skipped.patreon.com/x | github.com/x chains).