Deschide articolul original într-un tab nou Schimbarea acestei opțiuni se aplică imediat și reîncarcă pagina.
LECTURA
← Înapoi la Urmăritori

Emre's Blog

emre570.bearblog.dev/feed/
Vizitează site-ul

Cel mai recent articol

Ultima verificare

Urmăritori 0 urmăritori
Articole: 7

  1. Writing A Megakernel For LLM Decode - A Worklog

    (Emre's Blog)

    This is the story of a megakernel: the 28-layer text decoder of Qwen3-ASR-0.6B, plus its lm_head, running as a single persistent CUDA kernel on an RTX 5080.

  2. Unleashing Blackwell's 4-bit: a surgical look at MXFP4 and NVFP4

    (Emre's Blog)

    If you do kernel-level inference optimization, you eventually hit the cold truth: the enemy is not compute, it is VRAM bandwidth.

  3. From 429 GB/s to the DRAM wall: writing an FP8 quantizer on an RTX 5080

    (Emre's Blog)

    This is the story of one CUDA kernel pair in cublade, my personal kernel library.

  4. 8.5x Faster Speech-to-Text: From 429ms to 50ms on a Single GPU

    (Emre's Blog)

    My STT journey started with Whisper.

  5. W8A16 Quantization with LLM.int8-Style Outlier Handling

    (Emre's Blog)

    In Phase 1, we demonstrated that outliers in transformer models are not "statistical noise"—they are critical carriers of model intelligence and context.

  6. How Critical Are Outliers in Transformer Models? A Live Experiment - Phase 1

    (Emre's Blog)

    Introduction Many assume that large language models (LLMs, Transformers) simply “memorize data” due to their scale.

  7. Transformer Architecture: Building Blocks Explained

    (Emre's Blog)

    Hey there!