Jim B (@jimypbr) header
Jim B (@jimypbr) profile picture

🦔🏴󠁧󠁢󠁳󠁣󠁴󠁿🇪🇺 Jumbled Instructions Multiple Bugs AI Engineer

Scotland, United Kingdom
Joined March 2017
Photo by Jim B (@jimypbr)Photo by Jim B (@jimypbr)Photo by Jim B (@jimypbr)Photo by Jim B (@jimypbr)Photo by Jim B (@jimypbr)Photo by Jim B (@jimypbr)Photo by Jim B (@jimypbr)Photo by Jim B (@jimypbr)Photo by Jim B (@jimypbr)
Still feels surreal to have been on stage with one of the greatest CEO's of our time Dr Lisa Su where she explicitly called out GPU MODE and the work we did to enable the world's first $100K competitive kernel competition I never in a thousand years would have imagined that a humble reading group would morph into a machine that can generate more kernel data than what exists on all of Github combined and that the best members of our community could beat the best baselines by the best human experts Thank you GPU MODE, my favorite corner of the internet
9
21
4
210
24,164
Who would have thought it???
New modeling study finds that if everyone in the United Kingdom had worn N95 or FFP2 masks from the start of the pandemic, the rate of COVID-19 transmission would have been 9 times lower.
39
babe wake up, new muP paper dropped arxiv.org/abs/2407.17465 And holy smokes does this look promising!
10
51
4
333
36,191
Jim B retweeted
Scalify scale propagation for efficient low-precision LLM training Low-precision formats such as float8 have been introduced in machine learning accelerated hardware to improve computational efficiency for large language models training and inference. Nevertheless, adoption
3
20
1
68
13,255
It’s impossible not to cry every day. This disease has taken everything. For a year and a half I’ve lived in a nightmare and made no progress getting out. Severe #LongCovid is something no one should have to endure.
245
633
33
6,452
362,992
Exciting news: Graphcore joins @SoftBank_Group to build next generation of AI compute. graphcore.ai/posts/graphcore…
4
31
19
79
50,098
I had so many dreams. I wanted my own science show on Netflix. I wanted to start a family with Kyle. I wanted to scuba in Australia. Now, I just want to walk again. I want to have a real conversation. I want to read a book again. I want to listen to music and see my family.
334
639
92
9,485
281,718
Yes! I read this book cover to cover and completed the problems at the start of my programming career. One of the best investments I ever made.
Regret: Should have read this book earlier
2
214
Super long 🧵 incoming. I really enjoy looking at the full stack of computing - maximizing intelligence per picojoule. Been wanting to do a far-too-in-depth overview of the ENTIRE thing - from manufacturing a GPU to training a GPT :D Here we go!
7
47
8
226
44,512
Hi everyone yes, I left OpenAI yesterday. First of all nothing "happened" and it’s not a result of any particular event, issue or drama (but please keep the conspiracy theories coming as they are highly entertaining :)). Actually, being at OpenAI over the last ~year has been really great - the team is really strong, the people are wonderful, and the roadmap is very exciting, and I think we all have a lot to look forward to. My immediate plan is to work on my personal projects and see what happens. Those of you who’ve followed me for a while may have a sense for what that might look like ;) Cheers
1,393
1,263
560
21,240
3,344,714
Anyone else experience a trickle new followers with fake looking profiles in the past couple of weeks?
4
51
8-bit floating point is the next tour de force to train and deploy LLMs. Today we present our new @graphcoreai preprint: 📢 Training and inference of large language models using 8-bit floating point 📢 arxiv.org/abs/2309.17224 Details on the 🧵[1/7]
3
34
5
149
36,677
We are excited to release FastChat-T5: our compact and commercial-friendly chatbot! - Fine-tuned from Flan-T5, ready for commercial usage! - Outperforms Dolly-V2 with 4x fewer parameters. Link: github.com/lm-sys/FastChat#F…
27
138
26
701
155,529
Our new paper Unit Scaling enables FP16/FP8 training *out-of-the-📦* No need for loss scaling / complex automatic scaling methods 🙌 Hoping to change the way we think about low-precision training. Paper: arxiv.org/abs/2303.11257 Notebook: ipu.dev/AzbWWv 🧵[1/n]
1
30
7
154
18,773
I just got off the Zoom with @iScienceLuvr and @johnowhitaker where we completed recording the final @fastdotai part 2 video -- there are 18 of them! I'm working on creating the website for the new course now and will let you all know as soon as it's ready! :D
23
48
6
622
62,034
Those who partake in "culture wars" believe it's about the future direction of society, perhaps even the survival of their idea of civilization. The more mundane reality: it's about siphoning away the savings of gullible marks & consolidating the political power of a few conmen.
7
16
1
180
33,814
😍😍 literally my first computer...! Just as slow as I remembered 😂 I even have to reduce the screen size in doom to get a better frame rate like the old days. Didn't have real speakers though... PC speaker FTW.
Hidden feature at @huggingface for the nostalgic ones Run windows 3.1 and play doom in it huggingface.co/spaces/coolsp…
1