Brendan McKeag

Falcon-180B is the largest open-source LLM to date, requiring 400GB of VRAM to run unquantized. This post explores how to deploy it on Runpod with A100s, L40s, and quantized alternatives like GGUF for more accessible use.

Klangio uses AI to transcribe music from recordings into readable notation, empowering learners and professionals alike. With Runpod’s infrastructure, Klangio can scale without the burden of managing GPU infrastructure.

his week’s roundup covers Alibaba’s vision-language model Qwen-VL, Meta’s new code-focused LLM Code Llama, and FACET—a benchmark for detecting bias in computer vision datasets.

Bench, Neuralangelo, and Marqo highlight this week’s updates—open-source tools for evaluating LLMs, reconstructing 3D scenes, and enabling GPU-powered vector search.

Learn how rank, learning rate, and training epochs impact the output of textual LoRAs—and how to balance these settings for coherent, stylistically faithful results.

This week’s Runpod RoundUp highlights Meta’s Audiocraft for AI-generated music and sound effects, new Chinese LLMs from Alibaba, and Salesforce’s DialogStudio dataset hub for building conversational AI.

This week’s Runpod RoundUp covers major releases including Llama-2 with 32k context support, SDXL 1.0’s public release, and StabilityAI’s new Stable Beluga LLMs—all now available to run on Runpod.