TwiScan
Hot
Communities
Account collections
Login
Register
English
日本語
한국의
简体中文
繁体中文
Register and share your invite link to earn from video plays and referrals.
Register now
不游
@fengbuyou
世界和平 | 做人类观察
Joined November 2023
1K
Following
4.3K
Followers
不游
@fengbuyou
2026.09.01 08:50
我最近训练了一些迷你模型,发现了什么呢? 1. 如果使用场景只是需要语义输入,然后主要逻辑是某种超多参数的复杂函数,梯度下降就是你最好的朋友,得到的结果比手写逻辑要好的多。 2. 训不明白的话,if/else也不错的。 3. 如果语义就是主要功能,那不如直接用qwen0.5B,在logits层上做文章。
Show more
0
0
0
2
0
Forward to community
Most Popular Users
Tibo
@thsottiaux
1M Followers
FinancialJuice
@financialjuice
959.2K Followers
Binance Wallet
@BinanceWallet
2.8M Followers
First Squawk
@FirstSquawk
579.8K Followers
Binance
@binance
16.1M Followers
LiveSquawk
@LiveSquawk
504.2K Followers
Serenity
@aleabitoreddit
1.1M Followers
ClaudeDevs
@ClaudeDevs
834.7K Followers
CNBC
@CNBC
6.3M Followers
Seeking Alpha
@SeekingAlpha
348.7K Followers
TheStreet
@TheStreet
1.2M Followers
The Wall Street Journal
@WSJ
22.3M Followers
Boris Cherny
@bcherny
641.8K Followers
Thariq
@trq212
405.5K Followers
New York Post
@nypost
4.3M Followers