@Teknium1
Somehow FSDP seems to have been the cause for hermes2 never converging. I dont understand it, but it seems to be the case. Deepspeed Zero 2 seems to work fine - testing zero 3 now https://t.co/PzDgFvaUhl
Viewing enriched Twitter post
Somehow FSDP seems to have been the cause for hermes2 never converging. I dont understand it, but it seems to be the case. Deepspeed Zero 2 seems to work fine - testing zero 3 now https://t.co/PzDgFvaUhl
{
"data": [
{
"id": "",
"type": "photo",
"url": null,
"media_url": "https://pbs.twimg.com/media/F8XcqbFbAAA_LLE.jpg",
"media_url_https": null,
"display_url": null,
"expanded_url": null
}
],
"score": 0.929,
"scored_at": "2025-08-09T13:46:07.550354",
"import_source": "manual_curation_2023",
"links_checked": true,
"checked_at": "2025-08-10T10:32:44.220523",
"media": [
{
"type": "photo",
"url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/1713017886099554641/media_0.jpg?",
"filename": "media_0.jpg"
}
],
"reprocessed_at": "2025-08-12T15:25:29.622927",
"reprocessed_reason": "missing_media_array"
} {
"user": {
"created_at": "2021-02-25T19:25:11.000Z",
"default_profile_image": false,
"description": "Github: https://t.co/LZwHTUFwPq\nHuggingFace: https://t.co/sN2FFU8PVE\nSupport me on Github Sponsors",
"fast_followers_count": 0,
"favourites_count": 27673,
"followers_count": 10711,
"friends_count": 1781,
"has_custom_timelines": true,
"is_translator": false,
"listed_count": 251,
"location": "USA",
"media_count": 1497,
"name": "Teknium (e/λ)",
"normal_followers_count": 10711,
"possibly_sensitive": false,
"profile_banner_url": "https://pbs.twimg.com/profile_banners/1365020011123773442/1679544051",
"profile_image_url_https": "https://pbs.twimg.com/profile_images/1642401912648777728/2KFikPsE_normal.jpg",
"screen_name": "Teknium1",
"statuses_count": 20319,
"translator_type": "none",
"url": "https://t.co/7DnNfZEOPh",
"verified": false,
"withheld_in_countries": [],
"id_str": "1365020011123773442"
},
"id": "1713017886099554641",
"conversation_id": "1713017886099554641",
"full_text": "Somehow FSDP seems to have been the cause for hermes2 never converging. I dont understand it, but it seems to be the case. Deepspeed Zero 2 seems to work fine - testing zero 3 now https://t.co/PzDgFvaUhl",
"reply_count": 6,
"retweet_count": 5,
"favorite_count": 67,
"hashtags": [],
"symbols": [],
"user_mentions": [],
"urls": [],
"media": [
{
"media_url": "https://pbs.twimg.com/media/F8XcqbFbAAA_LLE.jpg",
"type": "photo"
}
],
"url": "https://twitter.com/Teknium1/status/1713017886099554641",
"created_at": "2023-10-14T02:24:25.000Z",
"#sort_index": "1713017886099554641",
"view_count": 14692,
"quote_count": 0,
"is_quote_tweet": false,
"is_retweet": false,
"is_pinned": false,
"is_truncated": false,
"startUrl": "https://twitter.com/teknium1/status/1713017886099554641"
}