Search results

mul_mat_id produces wrong output with K-quantized source weights

Status: Open.

#1506 In ggml-org/ggml;

· OriPekelman opened on May 24, 2026

Segfault when loading arbitrary file as mmproj or model

Status: Open.

#1505 In ggml-org/ggml;

· joeyhub opened on May 23, 2026

ggml-cpu sched aliases intermediate grad-chain tensors in long backward graphs (CUDA correct)

Status: Open.

#1501 In ggml-org/ggml;

· OriPekelman opened on May 21, 2026

ggml-cuda: missing __syncthreads() in ssm-scan.cu

Status: Open.

#1498 In ggml-org/ggml;

· FrankZhuQS opened on May 19, 2026

ggml-cuda: cpy ignores destination strides (writes contiguously into a strided view_2d slot)

Status: Open.

#1497 In ggml-org/ggml;

· OriPekelman opened on May 18, 2026

ggml-cuda: add ggml_backend_cuda_buffer_from_ptr for UVA-direct mmap inference

Status: Open.

#1495 In ggml-org/ggml;

· OriPekelman opened on May 18, 2026

ggml_rms_norm_back: backend-sched compute produces different (wrong) result than legacy compute_with_ctx

Status: Open.

#1491 In ggml-org/ggml;

· OriPekelman opened on May 14, 2026

Windows build fails due to MLX backend (undefined symbols, no Windows build tags)

Status: Open.

#1486 In ggml-org/ggml;

· SkybuckFlying opened on May 12, 2026

Refactor large source files into smaller, modular units

Status: Open.

#1482 In ggml-org/ggml;

· Zabis13 opened on May 10, 2026

cmake: include path for ggml.h not exported if GGML_BACKEND_DL is set

Status: Open.

#1467 In ggml-org/ggml;

· SchaichAlonso opened on Apr 28, 2026

ggml-cuda: flash-attn MMA picker has no Blackwell (sm_120) entry — silently uses Ampere config

Status: Open.

#1466 In ggml-org/ggml;

· Zbig9000 opened on Apr 27, 2026

ggml-hexagon: HAP_power_set_HMX uses &ctx instead of ctx in htp_iface_open(), causing large HMX GEMM slowdown

Status: Open.

#1452 In ggml-org/ggml;

· happyyzy opened on Mar 31, 2026