Franz Kafka would have loved Gemini API I mean Google AI Studio I mean Vertex AI I mean Gemini Enterprise Agent Platform I mean Gemini Enterprise Agent Platform Inference I mean
gemini is a very good accountantView quoted post on X
Johannes Messner
@atomicflndr
🇪🇺 automating accounting. Prev. research @Aleph__Alpha
304 Following 320 Followers 413 Posts
Latest posts 5
The last thing I worked on at Aleph Alpha!
With @SohirMaskey Constantin Eichenberg @douglasahorr
tl;dr:
- Quantisation-aware training works really well
- If you have a fixed memory budget you should probably go many parameters - few bits
- k-means quant. is better than uniform
With @SohirMaskey Constantin Eichenberg @douglasahorr
tl;dr:
- Quantisation-aware training works really well
- If you have a fixed memory budget you should probably go many parameters - few bits
- k-means quant. is better than uniform
Would you rather use 1 million × 16-bit weights, 4 million × 4-bit weights, or even 16 million × 1-bit weights?
In joint work between Aleph Alpha Research and Graphcore, we asked this question of LLMs — the answer encouraged us to embrace the wonder ✨ of 1-bit weights, which can outperform 4-bit and 16-bit weights on a fixed weight memory budget.
In our work
- ⚖️ A scaling laws evaluation prompts us to consider very low-bit formats
- 📈 Scaled-up tests show the power of memory-matched models with 1-bit weights
- ⚡ Kernel benchmarking demonstrates their feasibility for autoregressive inference
Read all about it in our blog and paper (link below! ⬇️)
Massive thanks to our collaborators at Aleph Alpha Research!
Authors: @SohirMaskey, Constantin Eichenberg, @atomicflndr and @douglasahorrView quoted post on X
Tears in my eyes
🤯 MERZ AND MACRON JUST CONFIRMED A PAN-EUROPEAN LEGAL ENTITY IS COMING
It's now up to all of us to ensure the solution that gets passed into law is fit for purpose for European startups.
That means: EU–INC. 🚀
Support us and we all will get this done together. 🇪🇺🤝View quoted post on X
Seeing this pushback a lot - and it‘s fair!
However, these models don’t have a fixed vocabulary, i.e. there are infinitely many words the model can operate over instead of a finite set of tokens.
However, these models don’t have a fixed vocabulary, i.e. there are infinitely many words the model can operate over instead of a finite set of tokens.
I wouldn't really consider these to be tokenizer-free tbh.View quoted post on X
Unlike Hnets, these models are word level. The sequence is turned into words (this is literally called tokenization).
Then, the bytes of these words are turned into embeddings, which are then processed by a model.
