🐦 Twitter Post Details

Viewing enriched Twitter post

@teortaxesTex

They are hinting at that, sure. But they're testing on OPT, as in most of those Hype-Aware Quantization papers Why? OPT's FF layers use ReLU. It sacrifices perplexity but makes activations sparse. I'm skeptical it'll work for SwiGLU in LLaMA… without retrain. (paper:MoEfication) https://t.co/LbkvP5LW5i

Media 1
Media 2

📊 Media Metadata

{
  "media": [
    {
      "url": "https://pbs.twimg.com/media/F3tkVcDXoAAcB8B.jpg",
      "type": "photo",
      "original_url": "https://pbs.twimg.com/media/F3tkVcDXoAAcB8B.jpg",
      "format_converted_from_list": true
    },
    {
      "url": "https://pbs.twimg.com/media/F3tkim1WcAExARf.jpg",
      "type": "photo",
      "original_url": "https://pbs.twimg.com/media/F3tkim1WcAExARf.jpg",
      "format_converted_from_list": true
    }
  ],
  "conversion_date": "2025-08-13T00:32:48.703667",
  "format_converted": true,
  "original_structure": "had_media_only"
}

🔧 Raw API Response

{
  "user": {
    "created_at": "2010-09-18T13:32:22.000Z",
    "default_profile_image": false,
    "description": "Ours is the age of unaligned utilitarians. Other problems are relatively unimportant, but sometimes I tweet about them anyway.\n(кто/кого)",
    "fast_followers_count": 0,
    "favourites_count": 14778,
    "followers_count": 1652,
    "friends_count": 686,
    "has_custom_timelines": true,
    "is_translator": false,
    "listed_count": 24,
    "location": "",
    "media_count": 1112,
    "name": "Teortaxes",
    "normal_followers_count": 1652,
    "possibly_sensitive": false,
    "profile_banner_url": "https://pbs.twimg.com/profile_banners/192201556/1682742940",
    "profile_image_url_https": "https://pbs.twimg.com/profile_images/1652169745037242368/KRPTShbG_normal.jpg",
    "screen_name": "teortaxesTex",
    "statuses_count": 12359,
    "translator_type": "none",
    "verified": false,
    "withheld_in_countries": [],
    "id_str": "192201556"
  },
  "id": "1692057168659271988",
  "conversation_id": "1692057168659271988",
  "full_text": "They are hinting at that, sure.\nBut they're testing on OPT, as in most of those Hype-Aware Quantization papers\nWhy?\nOPT's FF layers use ReLU. It sacrifices perplexity but makes activations sparse.\nI'm skeptical it'll work for SwiGLU in LLaMA… without retrain.\n(paper:MoEfication) https://t.co/LbkvP5LW5i",
  "reply_count": 2,
  "retweet_count": 1,
  "favorite_count": 24,
  "hashtags": [],
  "symbols": [],
  "user_mentions": [],
  "urls": [],
  "media": [
    {
      "media_url": "https://pbs.twimg.com/media/F3tkVcDXoAAcB8B.jpg",
      "type": "photo"
    },
    {
      "media_url": "https://pbs.twimg.com/media/F3tkim1WcAExARf.jpg",
      "type": "photo"
    }
  ],
  "url": "https://twitter.com/teortaxesTex/status/1692057168659271988",
  "created_at": "2023-08-17T06:14:00.000Z",
  "#sort_index": "1695781238999568600",
  "view_count": 32546,
  "quote_count": 1,
  "is_quote_tweet": true,
  "is_retweet": false,
  "is_pinned": false,
  "is_truncated": false,
  "startUrl": "https://twitter.com/teortaxestex/status/1695781238999568624"
}