felix_red_panda Profile Banner
Felix Profile
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
Don't wanna be here? Send us removal request.
@felix_red_panda
Felix
2 years
All evals of ML models suck - but some are useful πŸ™ƒ.
2
3
73
@felix_red_panda
Felix
9 days
RT @vtabbott_: Adding multi-level performance models to diagrams. This will allow performance models of FlashAttention / matmul / distribut….
0
3
0
@felix_red_panda
Felix
12 days
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
@felix_red_panda
Felix
23 days
hacker news doing hacker news things πŸ˜„
Tweet media one
2
0
21
@felix_red_panda
Felix
1 month
RT @SzymonOzog_: This matmul visualization is so cool it got me banned last time I posted it
0
2
0
@felix_red_panda
Felix
2 months
open source speech synthesis model trained on two 4090 GPUs!.
@harrycblum
Harry Coultas Blum
2 months
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
@felix_red_panda
Felix
2 months
@tugot17 is currently in Beijing and will also visit Chongqing, Wuhan, Hangzhou and Shanghai.
0
1
5
@felix_red_panda
Felix
2 months
@tugot17 thank you to @SiliconFlowAI for the translated version
1
0
4
@felix_red_panda
Felix
2 months
our LLM inference blog post in πŸ‡¨πŸ‡³ And @tugot17 is traveling around China for the next 2 weeks. DM him to meetup :)
Tweet media one
@felix_red_panda
Felix
2 months
deep dive on LLM inference (read it if you haven't already!) link in the post post below
Tweet media one
3
3
30
@felix_red_panda
Felix
2 months
deep dive on LLM inference (read it if you haven't already!) link in the post post below
Tweet media one
6
23
275
@felix_red_panda
Felix
2 months
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!)
Tweet media one
1
0
24
@felix_red_panda
Felix
2 months
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
@felix_red_panda
Felix
2 months
though the memory bandwidth look looks pretty meh. feels like Intel is trying to make cards with @tenstorrent performance characteristics, though much cheaper(?) xD.
@felix_red_panda
Felix
2 months
Intel a 24GB GPU for ~500 USD, and there will be a single card version with GPU dies and 48GB combined memory
Tweet media one
4
0
22
@felix_red_panda
Felix
2 months
Intel a 24GB GPU for ~500 USD, and there will be a single card version with GPU dies and 48GB combined memory
Tweet media one
7
1
112
@felix_red_panda
Felix
2 months
if you want people to see your post on πŸ¦‹ site then you gotta post about it on X πŸ˜‚
Tweet media one
@felix_red_panda
Felix
2 months
I’m surprised how dead the πŸ¦‹ site is now. I have a few hundred followers there but ~nobody cared about the LLM inference blog
Tweet media one
0
0
12
@felix_red_panda
Felix
2 months
I’m surprised how dead the πŸ¦‹ site is now. I have a few hundred followers there but ~nobody cared about the LLM inference blog
Tweet media one
12
8
260
@felix_red_panda
Felix
2 months
the actually relevant error message is quite a bit further up top in the stack trace :(
Tweet media one
0
0
1
@felix_red_panda
Felix
2 months
yes, i still use a light terminal theme πŸ˜….
1
0
6
@felix_red_panda
Felix
2 months
the huggingface-cli gives a super generic error message when the token is not valid anymore. iirc that used to be better(?)
Tweet media one
4
0
9