Posts tagged “llama-cpp”

• llama.cpp b11413: How to Test Vulkan Sparse FlashAttention with Quantized KV

  Check eligibility for llama.cpp’s Vulkan sparse-attention change, understand the 16× cache threshold, and compare performance with the follow-up fix included.

• llama.cpp b11414: Vulkan Scratch-Buffer Fix and Long-Prompt Checks

  llama.cpp b11414 fixes stale Vulkan scratch-buffer reuse. Learn which workloads merit testing, what upstream results show and how to validate long prompts.

• llama.cpp b11415: Media Truncation Fix and Upgrade Checks

  llama.cpp b11415 fixes media-chunk boundary checks. What multimodal server operators should validate in prompt caches, client errors and staged upgrades.

• llama.cpp b11412: k-Pool Graph Fix and Regression Checks

  Test llama.cpp b11412’s k-pool graph fix for Qwen4Exp and GLM5-Next, with shared-sequence checks, diagnostic limits and cache-migration caveats.

• llama.cpp b11411: Saved KV-Cache Migration and Restore Checks

  llama.cpp b11411 rejects mismatched KV-cache restores and changes saved-state formats. How to rebuild caches, test restores and preserve rollback.

• llama.cpp b11405: CUDA Lightning-Indexer Upgrade Guide

  Test CUDA prefill gains on Qwen4Exp-style four-head lightning-indexer workloads, with depth-aware benchmarks and a check for missing baseline cases.

• llama.cpp b11404: How to Test Metal Speculative-Decoding Gains

  Test llama.cpp b11404 on Apple Silicon: Metal kernel eligibility, reported DFlash2 gains, and a four-run comparison of serial and speculative decoding.

• llama.cpp b11402: How to Evaluate Long-Context Prefill on DGX Spark

  A DGX Spark test plan for llama.cpp b11402: check FlashAttention eligibility, compare microbatches and validate real-server latency before upgrading.

• llama.cpp b11401: Router-Mode Fix and Upgrade Checks

  llama.cpp b11401 separates router child commands from logs. See who should upgrade and how to check downloads, model readiness, shutdown and logging.

• llama.cpp b11400: How to Test Mixed Text and Embedding Batches

  Evaluate llama.cpp b11400 mixed inputs: architecture gates, non-causal batch sizing, API steps and the limits of current multimodal server integration.

• llama.cpp b11398: When x86 CPU Vision Workloads Benefit

  Check whether llama.cpp b11398’s tinyBLAS K-tail support benefits your CPU vision workload, interpret reported timings and plan a controlled upgrade.

• llama.cpp b11393 Fixes Tool-Call Parser Use-After-Free

  llama.cpp b11393 clears a stale tool-call pointer. See which parser configurations warrant an upgrade and why the earlier b11390 CUDA fix is separate.

← Back to blog