Congrats to my former intern and longtime colleague Russ Cox @_rsc for 12 years of awesome impact as the TL for Go! https://
groups.google.com/g/golang-dev/c
/0OqBkS2RzWw
…
CODE
-
Russ Cox Celebrates 12 Years Leading Go Programming Language
By
–
-
Amazon Migrates 30000 Java Applications Using Q Code Transformation
By
–
'With Q's code transformation capabilities, Amazon has migrated over 30,000 Java JDK applications in a few months, saving the company $260 million and 4,500 developer years compared to what it would have otherwise cost.' -Andy Jassy
-
Web2MD: Convert HTML to Markdown Online Tool
By
–
Also I made something to turn HTML to MD: https://
web2md.answer.ai -
Creating Markdown Documentation Versions Optimized for LLMs
By
–
Yes I created md versions of our docs specifically designed for LLMs.
Examples medium size: https://
gist.github.com/jph00/26200368
915f6aabb450eaf33a03c3e8
…
Same but cut down: -
Building a Bignum Library for Mathematical Computing
By
–
Making a bignum library for fun https://
bit.ly/3xRELub
#AI #MachineLearning #DeepLearning #LLMs #DataScience -
Llama-3-EvoVLM-JP-v2: Open-Source Japanese Vision-Language Model Released
By
–
We are excited to release Llama-3-EvoVLM-JP-v2, the first open-source Japanese vision-language model capable of reasoning across multiple images! This model was developed quite inexpensively, in a resource-efficient manner, using evolutionary model merge. https://
sakana.ai/evovlm-jp/ -
Adam-atan2: Removing Epsilon Hyperparameter from Adam Optimizer
By
–
Adam-atan2 is a neat one line code change to remove the epsilon hyperparamter from Adam by changing the divide to an atan2(). This feels like a generally useful trick for many places you might be worried about divide by zero and numeric precision issues.
-
Groq Celebrates Developer Community SDK Integration Contributions
By
–
Thank you to @MSuryavansh and all of the other developers in our community who help integrate Groq SDK support into various frameworks for the benefit of the Open Source Community.
-
Post-Training Data Strategy: Key Competition Arena for LLMs
By
–
6/ Post-training data strategy is becoming a key arena for competition. Llama3.1 paper had 15 pg about post-training data (vs. 12 pg on pre-training), driving these capabilities: – Code
– Multilinguality
– Math/Reasoning
– Long-Context
– Tool Use
– Factuality
– Steerability -
Post-Training Data Strategies: SFT, RLHF, and DPO Approaches
By
–
4/We are also seeing remarkably similar data strategies for post-training from most labs at this point (at least what was published from Meta+Apple): – Hybrid data SFT, RLHF, & DPO setups
– Synthetic data on code and math
– Post-training data for most important capabilities