#NewListing on #bblist"Clearcom HBP-2x HelixNet Beltpack" - ift.tt/8xw9RIt

THIS. HelixNet: a cutting-edge Deep Learning structure with 3 Mistral-7B LLMs. Modeled after actor-critic methods in Reinforcement Learning, its DNA-inspired design reflects three networks collaborating seamlessly. (its components can be seamlessly transferred to other LLMs) 🔗: huggingface.co/migtissera/He… ↓

Reposting this since there's a lot of talk about "Reflection" right now. HelixNet was built with a similar insight. The difference here is that it consists of 3 separate LLMs: an actor, a critic and a regenerator. At that time I didn't know how to evaluate this, but I have now figured out a way. Will evaluate in the next few days and share the results with you.

Yeah! It also reminded me that HelixNet is due for an update 😁

Super interesting work! Congratulations to all involved. The approach however is very similar to HelixNet, that I released about a year ago. In HelixNet, the “thinker”, “critic” and “regenerator” was 3 separate 7B networks. Either way, super cool work and we’re absolutely destroying closed model moats! 👏🏾👏🏾👏🏾

Well we’ve got to change that! Sorry that you guys didn’t get the recognition as well. If I knew I would’ve made some noise then too. AI is moving so fast, and some of us folks hate writing and publishing. (I hated it during my PhD!). If anyone copies HelixNet and publishes without giving any credit, I’d love for folks to speak up as well.

Introducing Sensei (先生): A simple, powerful, minimal codebase to generate synthetic data using OpenAI. github.com/migtissera/Sensei This has been my framework for generating synthetic data using GPT-4. It includes Orca system contexts, as well as 10 new system contexts that I've designed for creating my models: Synthia, Tess and HelixNet. You have full control over the topics. The best part is that the prompt itself is generated using GPT-4, so there's minimal refusals. I hope you benefit from this as much as I have. Feel free to open PRs for other APIs, such as Mistral's.

Our latest training video is now live! A quick overview of how to integrate HelixNet to the Arcadia Central Station. Watch now: youtu.be/vA0kNhW0l5Q #ClearComTraining

Tune in to #NBC tonight for the annual Rockefeller Center Christmas Tree Lighting. Clear-Com’s Eclipse, FS Edge, FS II, Arcadia & Helixnet are supporting the event's communications for the event. Shout-out to #JetwaveWireless for their incredible support of this event!

Not your weights, not your AI! Tess-v1.1 is out, deprecating Tess-v1.0 models that had a few instruction following issues. Tess, short for Tessoro, is available in two sizes: Tess-M is trained on the Yi-34B with 200K context length, and Tess-XS is trained on the Mistral-7B with 8K context length. Tess-XS-v1.1: huggingface.co/migtissera/Te… Tess-M-v1.1: huggingface.co/migtissera/Te… This is my flagship series. SynthIA and HelixNet were just teasers. Enjoy!

Wow! I did a quick version of HelixNet based on a single quantised instance of @MistralAI's amazing 7B model and an actor/critic/re-generator prompting strategy. It works surprisingly well. And it can run on a Raspberry Pi!

You can now run full HelixNet (all 3 models) on a single 4090! Great work!

GPT-3.5 vs HelixNet (original response and the critique is not printed, as it's optional).

Folks have asked for an Open Source Mixture Of Experts platform for private local use. IT IS HERE! HelixNet. Just like ChatGPT-4 Mixture Of Experts “secret” technology. HelixNet is a novel Deep Learning architecture consisting of 3 x Mistral-7B LLMs. It has an actor, a critic, and a regenerator. HelixNet is insprired from an actor-critic architecture most prominent in Reinforcement Learning algorithms. The name Helix is referring to the spiral structure of a DNA molecule and it symbolizes the intertwined nature of the three networks, working in tandem, much like the strands of a DNA molecule. HelixNet regenerates very pleasing and accurate responses, due to the entropy preservation of the regenerator. The regenerator was only trained on a dataset of 1000 samples, similar to Meta's LIMA. The actor network here was trained on about 250K very high-quality samples, and the critic network was trained on further 10K samples. Helix is astounding! I will write a detailed ReadMultiplex.con article soon. Link: huggingface.co/migtissera/He…

It's been a big week for Open Source AI, and here's one more to cap the week off! Introducing HelixNet. HelixNet is a novel Deep Learning architecture consisting of 3 x Mistral-7B LLMs. It has an actor, a critic, and a regenerator. HelixNet is insprired from an actor-critic architecture most prominent in Reinforcement Learning algorithms. The name derives from Helix, referring to the spiral structure of a DNA molecule. It symbolizes the intertwined nature of the three networks, working in tandem, much like the strands of a DNA molecule. HelixNet regenerates very pleasing and accurate responses, due to the entropy preservation of the regenerator. Further, in testing, the critic and the regenerator seems readily transferrable to other LLMs. Here's the link to the model: huggingface.co/migtissera/He… Information on how to run it is provided on the readme file. Have a great weekend everyone!

.@Solotech_Inc deploys an extensive system featuring Clear-Com’s Eclipse HX Digital Matrix, HelixNet Digital Partyline, and FreeSpeak II for Resorts World Theatre 📶 mondodr.com/resorts-world-th…

We are excited to be a finalist in the @AVNews Project of the Year Awards for our @NatGeo Awards Ceremony project with FreeSpeak II and HelixNet beltpacks last summer! Winners will be announced at #ISE2023. #NatGeo story: clearcom.me/natgeo | #AVAwards #AVtweeps

.@clearcomsystem FreeSpeak, Dante and HelixNet integration offers the ideal combination of features for River Oak Church ⛪️ mondodr.com/clear-coms-arcad…

“Honestly, I feel like every church needs to know about Arcadia." @RiverOakChurch finds the perfect intercom solution for #HousesofWorship. Read the full story: clearcom.me/3PbN5te #FSII #HelixNet #Arcadia #ClearComStories #AVtweeps

German speakers: tune in to @AudioTechnicaUK's training on the #Arcadia with HelixNet Integration this Thurs. at 10:00 AM CET! Register here: attendee.gotowebinar.com/reg… | #audiotechnica #techtraining #avtweeps