Sabitlenmiş Tweet
I've been a developer in Bittensor for ~1 year and I wanted to share some personal reflections 🧵
English
Brian McCrindle
814 posts
Brian McCrindle
@mccrinbc
Founding Engineer @MacrocosmosAI | @bitstarterAI 👾 Previous: Machine Learning (@opentensor, @parallel_domain) 🇨🇦
Katılım Ocak 2020
432 Takip Edilen644 Takipçiler
@0xbuzzzzzz @cryptoresetlife @IOTA_SN9 @MacrocosmosAI because the runs are full right now -- Making new ones as we speak!
English
@mccrinbc @cryptoresetlife @IOTA_SN9 @MacrocosmosAI Why does it keep showing “Attempting to fetch run info for miner” today?
English
‼️🦞有mac的来挖矿！tao这个子网！我测试一天能挖8u左右。根据 #iota sn9子网代币价格变动
大家养龙虾买的mac有用处了
最近研究tao，发现个 @IOTA_SN9 激励式编排训练架构——在 Bittensor 子网 9 上进行去中心化模型训练。由……@MacrocosmosAI构建的项目，目前只发布了mac系统的挖矿。
1️⃣mac电脑下载tah.iota.macrocosmos.ai/download/token…
2️⃣注册tao钱包bittensor.com/wallet
3️⃣打开软件，选第二个‼️填入自己的钱包地址。（一定不要选第一个为项目方贡献）
4️⃣到账的是iota 子网sn9 alpha代币
中文
English
@cryptoresetlife @IOTA_SN9 @MacrocosmosAI 老张，开了一个多小时了，还是没启动啊，日志一直显示Attempting to fetch run info for miner 5GsmJVv3...，是要等下一轮吗，多久一轮啊
中文
"Mac users, come mine! This subnet tao! I tested it and can mine about 8usd per day."
Never underestimate the crypto gang. This is the craziest TAH setup I have ever seen in my life
老张1982@cryptoresetlife
‼️🦞有mac的来挖矿！tao这个子网！我测试一天能挖8u左右。根据 #iota sn9子网代币价格变动 大家养龙虾买的mac有用处了 最近研究tao，发现个 @IOTA_SN9 激励式编排训练架构——在 Bittensor 子网 9 上进行去中心化模型训练。由……@MacrocosmosAI构建的项目，目前只发布了mac系统的挖矿。 1️⃣mac电脑下载tah.iota.macrocosmos.ai/download/token… 2️⃣注册tao钱包bittensor.com/wallet 3️⃣打开软件，选第二个‼️填入自己的钱包地址。（一定不要选第一个为项目方贡献） 4️⃣到账的是iota 子网sn9 alpha代币
English
Brian McCrindle retweetledi
Brian McCrindle retweetledi
Bittensor peeps: check out 31:44 - Templar sn3 discussed.
@chamath -- they've achieved a *72* billion parameter model with decentralized training, not a 4 billion parameter model :)
English
high-fidelity simulation and economic human+agentic optimization will move the idea of capturing the world's compute substrate for distributed training from a dream, to a reality.
Apex・SN1@Apex_SN1
We’ve built a simulation of the @IOTA_SN9 communication network. This is a high-fidelity digital twin - an abstracted version of our distributed training architecture, designed as a testing ground to run experiments and develop novel algorithms to increase the speed and quality of model training. We’re using the simulator as an environment for open competitions on Apex, outsourcing algorithmic innovations to the Bittensor miners. It’s the first time our simulator can be interacted with by the public. Our opening simulator competition is live now.
English
stoked to see this!! If you have any questions, feel free to hit me up
Jesus Martinez@JesusMartinez
Starting to get deeper into the rabbit hole Mining on Bittensor subnets I’ve always been a big fan of nodes & mining, so it was very easy to setup!
English
Brian McCrindle retweetledi
Jensen name-dropped me in the keynote and posed with our belt.
He has a physical belt too but they just showed the pic
Intially I made fun of the 35X perf improvement being bogus, I thought it was an exaggeration of performance
Turns out he was sandbagging, and perf is 50x
English
Brian McCrindle retweetledi
If you've wanted to get started mining Bittensor try out IOTA Train at Home. I just crossed 50 alpha tokens ($387) earned.
English
Brian McCrindle retweetledi
Good morning
Have earned ~$200 in the past two weeks contributing my mac mini to IOTA's decentralized model training runs
come taste the future
English
considering cohosting subnet summer nyc w/ @UnsupervisedCap in late march/early april:
- we will get the top ~10 subnet operators in. they show off vision and progress
- top range of capital allocators in, they get info, pushback on hopium, ask sharp q's
dm if interested
English
Brian McCrindle retweetledi
We recently concluded our matrix compression competitions on Apex. The purpose of these was to increase the training speed of @IOTA_SN9 by speeding up data transfer.
We found that by outsourcing the development of matrix compression algorithms, Apex achieved a ~3x reduction in data transmission size. Our winning algorithm strictly improves latency for network speeds < 50 MB/s.
See the full results in our Substack below
English
Brian McCrindle retweetledi
There's a lot of really exciting developments happening in decentralized AI training this year. Here's my take on why decentralized training is moving from "impossible" to "investable". 🧵👇
English
@IOTA_SN9 continues to build on this thesis, bootstrapping our research on this substrate, in efforts to build towards the vision of a truly decentralized future
English
@tplr_ai just showed the world what was possible, and most do not realize the scale of the worlds distributed compute substrate. Bittensor has shown us that compute is available, you just need to tap into it.
English
Get your AI to help you pull and plot your data from the db while on the go, and give summaries when asked. We run dozens of ablations a week in @IOTA_SN9, with hundreds of gigabytes of data, so I need to call in the Bishop🙏🏼🦀
English