Albert Tseng
@tsengalb99
Followers
621
Following
41
Media
6
Statuses
28
Explore trending content on Musk Viewer
田中さん
• 615518 Tweets
田中敦子さん
• 580282 Tweets
声優さん
• 259687 Tweets
攻殻機動隊
• 191925 Tweets
Zize
• 139663 Tweets
Arhan
• 135299 Tweets
ネルフェス
• 107761 Tweets
Shadow
• 100942 Tweets
#RenjunThe8rightestStar
• 95659 Tweets
#8YearsWithRenjun
• 95554 Tweets
草薙素子
• 91702 Tweets
OUR BRILLIANT RENJUN
• 91211 Tweets
ショック
• 86107 Tweets
Salim
• 78041 Tweets
Black Myth
• 66776 Tweets
フランメ
• 62649 Tweets
Sinner
• 46470 Tweets
FELIZ CUMPLE XCRY
• 41453 Tweets
초코우유
• 37933 Tweets
ベヨネッタ
• 30189 Tweets
#कल_भारत_बंद_नहीं_होगा
• 24045 Tweets
Laporta
• 23190 Tweets
#يواكيم_اندرسن_مطلب_الاتحاديين
• 20696 Tweets
D-7 TO CHANYEOL SOLO
• 18933 Tweets
スーパームーン
• 16756 Tweets
Bahia
• 16257 Tweets
バニーの日
• 15568 Tweets
リサリサ先生
• 14840 Tweets
Azizah
• 14585 Tweets
MV Out Now
• 13286 Tweets
@CheeJerry
@qingyao_sun
@volokuleshov
@chrismdesa
It looks like a lot of people can't see the second post:
(2/n TLDR) We’ve released quantized LLaMA, Mistral, and OpenHermes models and a full codebase that you can quantize and deploy your own models with at .
1
13
94
👉QuIP# is even better now! Updated QuIP# adds fine-tuning for SOTA quantized 2, 3, & 4 bit LLMs. With QuIP#, 3 bit scales better than 4 bit for the first time!
💻
📜
3
25
90
🧵 (3/3)
QuIP (📜) will appear as a spotlight 💡 at
#NeurIPS
2023 next week in New Orleans 🎷. See you there!
QuIP# was done with
@CheeJerry
@qingyao_sun
@volokuleshov
@chrismdesa
1
6
46
Deep models need data, and annotating lots of behavior data is costly. In our
@CVPR
paper, we introduce AutoSWAP, a framework that uses program synthesis and weak supervision to train models with a fraction of the labels!
📄&👊🃏:
(1/n)
2
6
27
Turns out QuIP# works pretty well at 1 bit too. In fact, it's close to original QuIP at 2 bits.
Example generation from 1 bit 2-70b: "Cornell University is located in Ithaca, NY, which is known for its beautiful scenery. Ithaca is a small city of about 16,000 people ..."
👉QuIP# is even better now! Updated QuIP# adds fine-tuning for SOTA quantized 2, 3, & 4 bit LLMs. With QuIP#, 3 bit scales better than 4 bit for the first time!
💻
📜
3
25
90
2
2
15
0
0
6
🧵 (2/n)
QuIP# crushes all publicly available 2 bit PTQ methods on language modeling & zero shot tasks while being conceptually clean and simple.
We’ve released quantized LLaMA, Mistral, and OpenHermes models, and a full codebase at
1
2
6
Special thanks to co-authors
@JenJSun
and
@yisongyue
! Our poster will also make an appearance at , make sure to check out MABe 2022!
Also, why is there no code emoji?
(5/n)
0
1
1
In AutoSWAP, domain experts provide domain knowledge in a reusable domain-specific language (from program synthesis) and domain-level labeling functions (LFs, noisy heuristic functions from weak supervision).
(2/n)
1
0
1