rckrd

7 ms·

31.
by rckrd 3y ago
Logit-bias guidance goes a long way -- LLM structure for regex, context-free grammars, categorization, and typed construction. I'm working on a hosted and model-agnostic version of this with thiggle [0] https://thiggle.com
32.
The Age-Old Resistance to Generated Code(matt-rickard.com)
1 points by rckrd 3y ago|0 comments
33.
Type Constraints for LLM Output(matt-rickard.com)
2 points by rckrd 3y ago|0 comments
34.
by rckrd 3y ago
We use a similar trick and expose it via an API. Much easier to parse when you can guarantee the shape of the output [0] https://thiggle.com/
35.
by rckrd 3y ago
We've found the same. A lot of usage through our LLM Categorization endpoint. The toughest problem was actually constraining the model to only output valid categories and not hallucinate new ones. And to only return one for single-clas
36.
The new economics of generating code(matt-rickard.com)
10 points by rckrd 3y ago|0 comments
37.
The Inevitability of Scope Creep(matt-rickard.com)
1 points by rckrd 3y ago|0 comments
38.
Beyond Prompt Engineering(matt-rickard.com)
3 points by rckrd 3y ago|0 comments
39.
The Hardware / Software Cycle(matt-rickard.com)
1 points by rckrd 3y ago|0 comments
40.
Why Fast?(matt-rickard.com)
3 points by rckrd 3y ago|0 comments
41.
Hard Launches vs. Soft Launches(matt-rickard.com)
1 points by rckrd 3y ago|0 comments
42.
Technical Metrics to Track in Engineering Orgs(matt-rickard.com)
8 points by rckrd 3y ago|3 comments
43.
Local Llama-2 Playground: Run LLMs in the Browser(thiggle.com)
2 points by rckrd 3y ago|0 comments
44.
The Contrarian Strategy of OpenAI(matt-rickard.com)
2 points by rckrd 3y ago|0 comments
45.
AI and Text-First Interfaces(matt-rickard.com)
7 points by rckrd 3y ago|0 comments
46.
by rckrd 3y ago
In more impressive news, "38% of code generated by GPT-4 does not contain API misuses"
47.
The Truth About GPU Utilization(matt-rickard.com)
2 points by rckrd 3y ago|0 comments
48.
ScapeNet: Real-time object detection in RuneScape (2021)(matt-rickard.com)
1 points by rckrd 3y ago|0 comments
49.
Modern Samizdat Libraries(matt-rickard.com)
4 points by rckrd 3y ago|0 comments
50.
What Is a Prompt Engineer?(matt-rickard.com)
2 points by rckrd 3y ago|0 comments
51.
Learning via Negativa(matt-rickard.com)
34 points by rckrd 3y ago|8 comments
52.
by rckrd 3y ago
I also released a hosted version of my open-source libraries ReLLM and ParserLLM that already supports APIs for * Regex completion for LLMs * Context-free Grammar completion for LLMs https://thiggle.com/ [0] https:/&#
53.
Diffusion of Innovation(matt-rickard.com)
1 points by rckrd 3y ago|0 comments
54.
Apple's Hidden Social Network(matt-rickard.com)
33 points by rckrd 3y ago|9 comments
55.
Context-Free Grammar Parsing with LLMs(matt-rickard.com)
2 points by rckrd 3y ago|0 comments
56.
Why did Meta open-source Llama 2?(matt-rickard.com)
73 points by rckrd 3y ago|66 comments
57.
Why Open-Source a Model?(matt-rickard.com)
2 points by rckrd 3y ago|0 comments
58.
Stop Overengineering(matt-rickard.com)
2 points by rckrd 3y ago|0 comments
59.
Hard Launches vs. Soft Launches(matt-rickard.com)
2 points by rckrd 3y ago|0 comments
60.
Why Frameworks Fail(matt-rickard.com)
2 points by rckrd 3y ago|0 comments

More ›