No.
PyTorch was inspired by Torch7 which was developed at NEC Labs and NYU (in my lab).
But PyTorch supported dynamic graphs from the start, which Torch7 couldn't do. You are thinking of Torch3, which was developed at IDIAP in Switzerland. Very different beast.
OPEN SOURCE
-
PyTorch Dynamic Graphs vs Torch7 Technical Heritage
By
–
-
Apollo AI Adds Open-Source LLMs: Mixtral, Hermes, LLaMA-2
By
–
Added a few more OSS LLMs to @apolloaiapp:
— Aaron Ng (@localghost) 17 décembre 2023
+ Mixtral 8x7B
+ OpenOrca (Mistral 7B)
+ Nous Hermes LLaMA-2 7/13/70B
+ Meta LLaMA-2 Chat 7/13/70B pic.twitter.com/vLuhU9GehyAdded a few more OSS LLMs to @apolloaiapp
: + Mixtral 8x7B
+ OpenOrca (Mistral 7B)
+ Nous Hermes LLaMA-2 7/13/70B
+ Meta LLaMA-2 Chat 7/13/70B -
Competitive OSS Model Market Analysis via Major Endpoints
By
–
Some thoughts on how the actual market for competitive OSS models is manifesting via endpoints from @togethercompute
, @anyscalecompute
, @perplexity_ai
, and so on for Mixtral 8x7B—which is about as performant as GPT-3.5 Turbo on some benchmarks. -
Open Source Models Match GPT-3.5 at Half the Price
By
–
It’s kind of crazy we went from “oh yeah open source might one day be cheap/competitive with OpenAI” to getting like a half dozen endpoints for a model that basically matches GPT 3.5-Turbo in a week—and some of them are half the price.
-
LangChain Release Notes: New LLM Integrations and Templates
By
–
LangChain Release Notes Week of 12/11 LangChain Core, LangChain Community, and LangChain v0.1
New LLM integrations like Mixtral and Gemini
5 New LangChain Templates
Lot's of community apps to highlight Read here & sign up to get it in your inbox every other week: -
Lisp-to-C Compiler: Homoiconicity and Language Integration
By
–
I asked Ronan once. He just decided to use odd numbers after the failure of Torch4. Lush was actually pretty awesome, Thanks to some of Léon's magic. It had a Lisp-to-C compiler that allowed to intersperse Lisp and C in a single function.
The homoiconicity of Lisp made it super -
Ronan’s Torch Framework Evolution From PhD to NEC Labs
By
–
Here is the full history in all its gory details. Ronan developed Torch3 during his PhD at IDIAP.
He then developed Torch4 in 2004, but quickly abandoned it. In 2006, he started Torch5, and completed it while at NEC Labs. This was written in C (Ronan positively hated C++ at the -
Open Pretrained Transformers: Stanford MLSys Deep Learning Advances
By
–
Open Pretrained Transformers – Susan Zhang | Stanford MLSys #77 https://
youtube.com/watch?v=p9IxoS
kvZ-M&t=1s
…
#AI #MachineLearning #DeepLearning #LLMs #DataScience -
Mistral co-founders shaped by FAIR-Paris open research culture
By
–
Two of the three co-founders of Mistral did their PhD in residence at FAIR-Paris and worked there as research scientists for a few years before founding Mistral.
They were bathed in the open research culture of FAIR. -
Open Source AI History Before Llama and PyTorch
By
–
Open source AI didn't start with Llama, nor with PyTorch.