Commits

Branch selector

 main

User selector

jessegross

Datepicker

All time

Commit History

Commits on Sep 2, 2025

• llm: Avoid underflow in free memory logging

  Show description for 8149a3c
  jessegross

  committed
  8149a3c

  Copy full SHA for 8149a3c

Commits on Aug 27, 2025

• ggml: Avoid allocating CUDA primary context on unused GPUs

  Show description for 9d97e6a
  jessegross

  committed
  9d97e6a

  Copy full SHA for 9d97e6a

Commits on Aug 20, 2025

• llm: Don't always evict models in CPU-only mode

  Show description for 073fa31
  jessegross

  committed
  073fa31

  Copy full SHA for 073fa31

Commits on Aug 19, 2025

• kvcache: Use Cast instead of Copy for flash attention masks

  Show description for 05ccb17
  jessegross

  committed
  05ccb17

  Copy full SHA for 05ccb17

Commits on Aug 18, 2025

• llm: Check for nil memory data before printing

  Show description for e3ade45
  jessegross

  committed
  e3ade45

  Copy full SHA for e3ade45

Commits on Aug 14, 2025

• llm: New memory management

  Show description for d5a0d8d
  jessegross

  committed
  d5a0d8d

  Copy full SHA for d5a0d8d

Commits on Aug 12, 2025

• ggml: Use ordinal IDs for AMD GPUs on Linux when UUID is unavailable

  Show description for a343ae5
  jessegross

  committed
  a343ae5

  Copy full SHA for a343ae5

Commits on Aug 11, 2025

• discover: CPU supports flash attention

  Show description for 8f4ec9a
  jessegross

  committed
  8f4ec9a

  Copy full SHA for 8f4ec9a

Commits on Aug 8, 2025

• ggml: No-alloc mode

  Show description for 79f6376
  jessegross

  committed
  79f6376

  Copy full SHA for 79f6376

• ggml: Support closing backends

  Show description for 756c78c
  jessegross

  committed
  756c78c

  Copy full SHA for 756c78c

• ggml: Use GGML's typedef'ed pointer types

  Show description for d7f4f78
  jessegross

  committed
  d7f4f78

  Copy full SHA for d7f4f78

Commits on Aug 7, 2025

• server: Reduce gpt-oss context length for small VRAM GPUs

  Show description for f2e9c9a
  jessegross

  committed
  f2e9c9a

  Copy full SHA for f2e9c9a

Commits on Aug 5, 2025

• ggml: Prevent kv cache quanitization on gpt-oss

  Show description for 8253ad4
  jessegross

  committed
  8253ad4

  Copy full SHA for 8253ad4

Commits on Aug 4, 2025

• kvcache: Log contents of cache when unable to find a slot

  Show description for 0d38b66
  jessegross

  committed
  0d38b66

  Copy full SHA for 0d38b66

Commits on Jul 31, 2025

• kvcache: Enable SWA to retain additional entries

  Show description for 4183bb0
  jessegross

  committed
  4183bb0

  Copy full SHA for 4183bb0

Commits on Jul 29, 2025

• kvcache: Don't shift empty batches

  Show description for c116a75
  jessegross

  committed
  c116a75

  Copy full SHA for c116a75

Commits on Jul 25, 2025

• kvcache: Group shift operations into batches

  Show description for 764be74
  jessegross

  committed
  764be74

  Copy full SHA for 764be74

Commits on Jul 11, 2025

• ggml: Use assigned layers when reporting loading stats

  Show description for acef9b4
  jessegross

  committed
  acef9b4

  Copy full SHA for acef9b4

• ggml: Disable unused pipeline parallelism

  Show description for 9a43994
  jessegross

  committed
  9a43994

  Copy full SHA for 9a43994

Commits on Jul 9, 2025

• ggml: Report ordinal IDs for AMD GPUs on Windows

  Show description for 35fda7b
  jessegross

  committed
  35fda7b

  Copy full SHA for 35fda7b

Commits on Jul 7, 2025

• Revert "ggml: Temporarily disable reporting UUIDs"

  Show description for 592d21e
  jessegross

  committed
  592d21e

  Copy full SHA for 592d21e

Commits on Jun 27, 2025

• ggml: Temporarily disable reporting UUIDs

  Show description for 45f216a
  jessegross

  committed
  45f216a

  Copy full SHA for 45f216a

Commits on Jun 20, 2025

• ggml: Check return status for computation.

  Show description for 87b7af6
  jessegross

  committed
  87b7af6

  Copy full SHA for 87b7af6

Commits on May 29, 2025

• ggml: Export GPU UUIDs

  Show description for aaa7818
  jessegross

  committed
  aaa7818

  Copy full SHA for aaa7818

• llm: Make "POST predict" error message more informative

  Show description for f15ffc4
  jessegross

  committed
  f15ffc4

  Copy full SHA for f15ffc4

Commits on May 27, 2025

• kvcache: Skip computing causal mask for worst case graph reservation

  Show description for ea79003
  jessegross

  committed
  ea79003

  Copy full SHA for ea79003

Commits on May 24, 2025

• ml: Improve slog formatting for BackendMemory

  jessegross

  committed
  f18e0cb

  Copy full SHA for f18e0cb

Commits on May 22, 2025

• ml: Panic rather than return error on tensor allocation failure

  Show description for 1f371ea
  jessegross

  committed
  1f371ea

  Copy full SHA for 1f371ea

• ollamarunner: Memory usage reporting

  Show description for 73d6a82
  jessegross

  committed
  73d6a82

  Copy full SHA for 73d6a82

• ggml: Report graph memory for failed allocations

  Show description for 6db8a37
  jessegross

  committed
  6db8a37

  Copy full SHA for 6db8a37

Commits on May 19, 2025

• llm: Use first layer as memory buffer in estimation

  Show description for 3fe74fb
  jessegross

  committed
  3fe74fb

  Copy full SHA for 3fe74fb

• ggml: Seperate tensor load from backend creation

  Show description for 94ab428
  jessegross

  committed
  94ab428

  Copy full SHA for 94ab428

• llm: Estimate projector memory correctly for Ollama engine

  Show description for d755577
  jessegross

  committed
  d755577

  Copy full SHA for d755577

• llm: Consistently track unassigned model data

  Show description for a2cc857
  jessegross

  committed
  a2cc857

  Copy full SHA for a2cc857

Commits on May 15, 2025

• ollamarunner: Multi-modal worst case graph

  Show description for fe623c2
  jessegross

  committed
  fe623c2

  Copy full SHA for fe623c2

Pagination

Previous

Next