HN Explorer

gdiamos

↗ HN profile [ 63.6 ms ] full profile
Karma
0
Created
()
Submissions
0
  1. CUDA-like programming of Cerebras WSE (github.com)
    2 points 2
  2. I aint gonna work on Maggie's Datacenter no more [video] (youtube.com)
    3 points 0
  3. Demo: Fold your coding sessions into LLM weights (app.scalarlmforge.com)
    2 points 0
  4. LLM-Deflate: Extracting LLMs into Datasets (scalarlm.com)
    77 points 12
  5. AMD MI300X Memcpy Peer Deep Dive (scalarlm.com)
    1 points 0
  6. Craylm: Open-source unified LLM training and inference for R1 (blog.cray-lm.com)
    9 points 2
  7. Notes on Building LLMs in 2024

    Here's what I'm observing about developing LLMs. What are others seeing? 1. Prompting a base model - very fast for prototyping - eventually hits an accuracy limit, e.g. after 100 iterations of prompts and around 4 hours…

    4 points 0
  8. LLM Inference at the Memory Wall (lamini.ai)
    1 points 0
  9. LLMs hallucinate in critical enterprise scenarios (lamini.ai)
    6 points 1
  10. Training LLMs in the Wild Wild West – AMD multi-node (lamini.ai)
    2 points 0
  11. Open source ICD11 Mistral trained on AMD GPUs [video] (youtube.com)
    2 points 1
  12. Good old fashion ML is dying

    LLMs popularized zero-shot learning, or “prompt engineering” which is drastically easier to use and more effective than labeling data. You can also retrofit “prompt engineering” onto good old fashion ML like text…

    20 points 16
  13. AI startup Lamini bets future on AMD's Instinct GPUs (theregister.com)
    22 points 7
  14. How to use a LLM to classify text (github.com)
    2 points 1
  15. Laminify: Instantly classify data with Lamini and Llama 2 (github.com)
    2 points 2
  16. Cramming an LLM platform into one container (lamini-ai.github.io)
    3 points 1
  17. 1.109B times faster serving of finetuned LLMs (lamini.ai)
    2 points 1
  18. Specializing LLMs on Databricks (databricks.com)
    3 points 0
  19. MLPerf Releases Inference Results (mlperf.org)
    2 points 1