BTW, Since Jamba supports a 256K context with high throughput, we also stumbled upon an issue where the fused_moe kernel didn’t work well in long contexts. Others seems to have had this too, according to some other open issues
RESEARCH
-
Jamba Modeling and Mamba Cache Management Updates
By
–
Things in this PR:
– Jamba modeling file, supporting Jamba variations
– Mamba cache management (also benefit other Mamba-based models)
– Send more request related properties to the forward pass, for model-specific implementations (including things needed for speculative decoding) -
Jamba Model Now Supported in vLLM for Efficient Serving
By
–
Jamba support is now live on vLLM Due to its novel hybrid SSM-Transformer arch, Jamba didn’t work out-of-the-box in vLLM. our own @MorZusman worked together with @vllm_project to integrate Jamba for an efficient serving in vLLM
-
Generative AI Simplifies Phase Transition Classification
By
–
[#Article] Researchers use generative AI to simplify phase transition classification https://actuia.com/actualite/des-chercheurs-utilisent-lia-generative-pour-simplifier-la-classification-des-transitions-de-phase/
… #AI #ArtificialIntelligence -
OpenAI Hack Exposed, Chinese AI Advances, New Tools Jobs
By
–
Top stories in AI today: -OpenAI's secret AI details stolen in 2023 hack
-Chinese AI firms flex muscles at WAIC
-Control AI videos with Luma Labs’ Keyframes
-DeepMind’s JEST supercharges AI training
-6 new AI tools & 4 new AI jobs Read more: http://
therundown.ai/p/openais-secr
et-hack-exposed
… -
Robot Audition and Human-Robot Interaction Workshop in Delft
By
–
Speakers include an opera singer, experts in robot audition, HRI, speech synthesis and a multimodal psychologist. Next Monday, July 15 in Delft, Netherlands and online. Details: https://
rosielab.github.io/wev/
Sign-up: https://
docs.google.com/forms/d/e/1FAI
pQLSfS1DnvpEA5w-n09vwWhIo1K3NsfveQYSLuIZ-qU1IIM8PROA/viewform
…
Sponsored by @SFUFASS Breaking Barriers Grant -
AI as Shakespeare, Not Terminator: Cultural Impact
By
–
Not sure where I read it, but this line stuck with me:
“AI was supposed to be The Terminator, instead it has turned out to be Shakespeare!” -
Neural Networks Dominance Narrows AI Field’s Perspective
By
–
The downside of neural networks having been a very insular community for so long is that now that they've taken over AI has become extremely narrow.
-
ICL versus Fine-tuning: Cost-Latency Trade-offs in LLM Optimization
By
–
I agree actually, if a problem can be solved with ICL (in terms of cost, latency, etc. and not just accuracy), then it's perfect. But I want to point out that it also makes fine-tuning even more relevant in a lot of use cases.
-
Unconventional Science Communication Appeals to Experts and Non-Tech Audiences
By
–
I’m internal and biased but have noticed that people who expect a traditional popular science approach i.e. “this is how AI works” seem to not like it. But both people who are AI experts already *and* those coming from non-tech backgrounds really enjoy it.