I’d say that Triton kind of is that for CUDA. I.e summarizes code into more compact building blocks.
@rasbt
-
Training Data Determines Model Behavior and Outputs
By
–
More generally, it’s whatever is predominant in the training data.
-
Machine Code as Universal Programming Standard: LLM Capabilities
By
–
Hm, if we assume that all programming languages ultimately run machine code, then we already have that scenario where things are standardized. You have Python on the one end, Assembly on the other hand. You could ask LLMs to write machine code today, but it won’t be as good as
-
Creating Concise Languages That Map to Popular Targets
By
–
The best you can do for now is create a less verbose language that maps onto a widely used target language. You can then generate paired training examples programmatically so the LLM learns to operate with fewer tokens. (Congratulations, you also just reinvented Triton.)
-
LLMs Benefit from Human-Designed Programming Languages
By
–
LLMs are quite capable coders. Now, people say they are limited by programming languages designed for humans. I think it is the opposite: they are enabled by programming languages designed for humans. You cannot just design a new programming language for LLMs because where
-
LLM Comparison Issues Apply Equally Between Companies
By
–
Yeah, but if company x compares their LLM to that of company Y the same concept (issue) applies.
-
The Challenge of Evaluating LLMs: Balancing Marketing Claims and Independent Testing
By
–
It’s kind of a dilemma. You want to check independent evals because the original ones might be inflated for marketing purposes.
At the same time, independent evals may also undersell the LLM because of accidental bad prompting, bad batching, bad optimization etc -
Black Forest Labs FLUX model licensing uncertainty clarified
By
–
Black forest labs engineer says its their FLUX model. Maybe they licensed different ones. Not sure.
-
Heavy Lifting with Subword Splits and Max Length Settings
By
–
Literally heavy lifting with subword splits and max_length sets and all
-
Open Weight Models and On-Device AI Privacy Benefits
By
–
Releasing the open weight models to the research community and developers in general is not a bad thing though. On the topic of on-device. It can be interesting for a lot contexts though, which is why Apple does/tried it. One is privacy (I don’t think most people want third