Felix
@felix_red_panda
Followers
5K
Following
22K
Media
184
Statuses
6K
speech synthesis and LLM nerd, DMs open, working on LLM stuff at @PrimeIntellect | prev @Aleph__Alpha
Berlin, Germany
Joined June 2020
RT @vtabbott_: Adding multi-level performance models to diagrams. This will allow performance models of FlashAttention / matmul / distribut….
0
3
0
RT @tugot17: I solved every single problem in the CUDA mode book. A quick thread summarizing this experience and what I learned 1/x https:/….
0
241
0
open source speech synthesis model trained on two 4090 GPUs!.
Open source notebooklm . Today we're open sourcing our 100M voice models that can render conversations. This includes a 40kh base finetune that is capable of voice cloning. Our models can do a variety of non speech sounds! Try them out yourself!.
1
8
123
our LLM inference blog post in 🇨🇳 And @tugot17 is traveling around China for the next 2 weeks. DM him to meetup :)
deep dive on LLM inference (read it if you haven't already!) link in the post post below
3
3
30
deep dive on LLM inference (read it if you haven't already!) link in the post post below
6
23
275
Qwen3 0.6b is a shockingly good draft model a lot of the time (96.6% acceptance rate on the 4b model for this particular task!)
1
0
24
RT @johannes_hage: if no one else is showing that RL isn't just eliciting latent behavior already learned in pretraining, but is actually a….
0
14
0
though the memory bandwidth look looks pretty meh. feels like Intel is trying to make cards with @tenstorrent performance characteristics, though much cheaper(?) xD.
Intel a 24GB GPU for ~500 USD, and there will be a single card version with GPU dies and 48GB combined memory
4
0
22
Intel a 24GB GPU for ~500 USD, and there will be a single card version with GPU dies and 48GB combined memory
7
1
112
if you want people to see your post on 🦋 site then you gotta post about it on X 😂
I’m surprised how dead the 🦋 site is now. I have a few hundred followers there but ~nobody cared about the LLM inference blog
0
0
12
I’m surprised how dead the 🦋 site is now. I have a few hundred followers there but ~nobody cared about the LLM inference blog
12
8
260
the actually relevant error message is quite a bit further up top in the stack trace :(
0
0
1
the huggingface-cli gives a super generic error message when the token is not valid anymore. iirc that used to be better(?)
4
0
9