Viewing enriched Twitter post
The primary issue is that despite being known for compute efficiency in terms of raw FLOPs-to-performance, linear models are not very hardware efficient during *decoding* due to the fixed-size state. Mamba-2 only hits ~2.5 arithmetic intensity, whereas matmul for H100s peaks at around 300. This leaves lots of compute idle! 5/10
{
"media": [
{
"type": "photo",
"url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2033948596568264784/media_0.jpg",
"filename": "media_0.jpg"
}
],
"processed_at": "2026-03-17T16:53:05.189821",
"pipeline_version": "2.0"
}{
"type": "tweet",
"id": "2033948596568264784",
"url": "https://x.com/tri_dao/status/2033948596568264784",
"twitterUrl": "https://twitter.com/tri_dao/status/2033948596568264784",
"text": "The primary issue is that despite being known for compute efficiency in terms of raw FLOPs-to-performance, linear models are not very hardware efficient during *decoding* due to the fixed-size state.\nMamba-2 only hits ~2.5 arithmetic intensity, whereas matmul for H100s peaks at around 300.\nThis leaves lots of compute idle!\n5/10",
"source": "Twitter for iPhone",
"retweetCount": 0,
"replyCount": 1,
"likeCount": 0,
"quoteCount": 0,
"viewCount": 22,
"createdAt": "Tue Mar 17 16:48:29 +0000 2026",
"lang": "en",
"bookmarkCount": 0,
"isReply": true,
"inReplyToId": "2033948587131064393",
"conversationId": "2033948569502413245",
"displayTextRange": [
0,
278
],
"inReplyToUserId": "568879807",
"inReplyToUsername": "tri_dao",
"author": {
"type": "user",
"userName": "tri_dao",
"url": "https://x.com/tri_dao",
"twitterUrl": "https://twitter.com/tri_dao",
"id": "568879807",
"name": "Tri Dao",
"isVerified": false,
"isBlueVerified": true,
"verifiedType": null,
"profilePicture": "https://pbs.twimg.com/profile_images/1389486324516671496/owa9Z4AC_normal.jpg",
"coverPicture": "",
"description": "",
"location": "Stanford, CA",
"followers": 38193,
"following": 643,
"status": "",
"canDm": true,
"canMediaTag": true,
"createdAt": "Wed May 02 07:13:50 +0000 2012",
"entities": {
"description": {
"urls": []
},
"url": {}
},
"fastFollowersCount": 0,
"favouritesCount": 1610,
"hasCustomTimelines": true,
"isTranslator": false,
"mediaCount": 59,
"statusesCount": 910,
"withheldInCountries": [],
"affiliatesHighlightedLabel": {},
"possiblySensitive": false,
"pinnedTweetIds": [
"1811453622070444071"
],
"profile_bio": {
"description": "Asst. Prof @PrincetonCS, Chief Scientist @togethercompute. Machine learning & systems.",
"entities": {
"description": {
"hashtags": [],
"symbols": [],
"urls": [],
"user_mentions": [
{
"id_str": "0",
"indices": [
11,
23
],
"name": "",
"screen_name": "PrincetonCS"
},
{
"id_str": "0",
"indices": [
41,
57
],
"name": "",
"screen_name": "togethercompute"
}
]
},
"url": {
"urls": [
{
"display_url": "tridao.me",
"expanded_url": "https://tridao.me",
"indices": [
0,
23
],
"url": "https://t.co/uFTGOmLPGP"
}
]
}
}
},
"isAutomated": false,
"automatedBy": null
},
"extendedEntities": {
"media": [
{
"display_url": "pic.twitter.com/K183AZ9OzL",
"expanded_url": "https://twitter.com/tri_dao/status/2033948596568264784/photo/1",
"ext_media_availability": {
"status": "Available"
},
"features": {
"large": {
"faces": []
},
"orig": {
"faces": []
}
},
"id_str": "2033948591086354433",
"indices": [
279,
302
],
"media_key": "3_2033948591086354433",
"media_results": {
"id": "QXBpTWVkaWFSZXN1bHRzOgwAAQoAARw6CXefG8ABCgACHDoJeOXbEFAAAA==",
"result": {
"__typename": "ApiMedia",
"id": "QXBpTWVkaWE6DAABCgABHDoJd58bwAEKAAIcOgl45dsQUAAA",
"media_key": "3_2033948591086354433"
}
},
"media_url_https": "https://pbs.twimg.com/media/HDoJd58bwAEw_lS.jpg",
"original_info": {
"focus_rects": [
{
"h": 390,
"w": 696,
"x": 828,
"y": 0
},
{
"h": 390,
"w": 390,
"x": 981,
"y": 0
},
{
"h": 390,
"w": 342,
"x": 1005,
"y": 0
},
{
"h": 390,
"w": 195,
"x": 1079,
"y": 0
},
{
"h": 390,
"w": 2048,
"x": 0,
"y": 0
}
],
"height": 390,
"width": 2048
},
"sizes": {
"large": {
"h": 390,
"w": 2048
}
},
"type": "photo",
"url": "https://t.co/K183AZ9OzL"
}
]
},
"card": null,
"place": {},
"entities": {
"hashtags": [],
"symbols": [],
"timestamps": [],
"urls": [],
"user_mentions": []
},
"quoted_tweet": null,
"retweeted_tweet": null,
"isLimitedReply": false,
"article": null
}