Thomas 0451f10a42 Add GitHub data: filename sanitization (#640)		2025-07-23 13:31:53 +02:00
..
26 - Feature Request_ Improve CPU processing speed for large contexts.md
29 - Bug_ some ifdefs missing in ggml_src_iqk_iqk_quantize.cpp.md
30 - Bug_ Appcrash on Windows 7 with GGML_USE_IQK_MULMAT.md
34 - Bug_ FA fails when processing prompt lengths that are not a multiple of .md
59 - Bug_ GGML Compilation Error_ undefined references to _iqk_mul_mat_.md
60 - Bug_ Illegal instruction on NEON and Q4_0_4_4.md
67 - Feature Request_ Elliminate_reduce unnecessary copies.md
88 - Bug_ Won_t compile on MSVC.md
92 - Bug_ Quantized KV cache produces garbage in situation where llama.cpp do.md
103 - Bug_ K cache without FA.md
133 - Refactor_ update ggml library_.md
159 - Feature Request_ steps how to compile as cmake i struction on the origi.md
160 - Bug_ Can_t compile on MSVC 2022.md
167 - Bug_ Unable to quantize Falcon 10B 1.58 bitnet model.md
183 - Refactor_ iqk_mul_mat.md
196 - Refactor_ remove usage of Q8_1 for activation quantization.md
199 - Bug_ Changing system_prompt on llama-server at runtime breaks parallel .md
203 - Bug_ Compliation Error for Intel_R_ Xeon_R_ Gold 6326 CPU.md
209 - Does the iqk_mul_mat.cpp support 1.58-bit quantization model_.md
214 - AVX512 build error.md
217 - Bug_ CPU FA with fp16 K-cache is broken.md
224 - Bug_ IQK_FA_ALL_QUANTS causes failure to compile.md
227 - Prevent FA usage on CUDA when K and V head sizes are different.md
228 - Feature Request_ create tool to offline repack models.md
230 - Weird assert when using online repacking.md
245 - Bug_ Perplexity returns NaN with IQ4_KSS quantisation.md
249 - CUDA_ results for MoE models are not reproducible.md
254 - Split-mode row.md
255 - Feature Request_ dynamic layer by layer offloading during prompt proces.md
257 - Bug_ mla_2 in llama-server will crash when request done.md
263 - Benchmarking DeepSeek R1 - 16x3090.md
267 - Feature Request_ HugePage mmap alloc for DeepSeek V3_R1.md
271 - Possible regression computing _wk_b_ tensors on the fly after PR _265.md
281 - Bug_ Strange dips in TG performance.md
285 - llama-perplexity giving all NaNs on unsloth Q8_0 quant.md
293 - Feature Request_ IQ6_K row interleaved quant.md
296 - Possible numerical stability issue with experimental quant of DeepSeek-.md
297 - Update gguf-py scripts to support new quant types..md
300 - Bug_ IQK_FA_ALL_QUANTS causes failure to compile.md
305 - Gibberish output when using DeepSeek-V3-0324-IQ2_K_R4 on mixed CPU _ 4 .md
306 - Confused by the -mla flag. What_s supported_.md
308 - Bug_ Compiling for arm64_ error_ cannot convert _const uint32x4_t_ to _.md
314 - Llama 4 Support_.md
322 - Speculative decoding support.md
335 - Bug_ Llama 4 generates garbage with longer context _64K_ the issue is n.md
339 - Bug_ bitnet2b_2501 template issues.md
340 - Bug_ _unknown model architecture_ _cohere2_ when trying to load Command.md
345 - build question newbie.md
353 - Binaries releases for Windows _.md
358 - Bug_ IQK_FA_ALL_QUANTS causes failure to compile.md
361 - Bug_ Build not detecting some supported ARM CPUs.md
362 - README language is vague wrt. _quantization improvements_.md
363 - Bug_ Gibberish output when using flash attention using Mistral-Small-I.md
365 - Bug_ Updated BitNet arch bitnet-b1.58.md
367 - Bug_ IQ1_S_R4_ IQ1_M_R4 failed on Qwen3-235B-A22B.md
373 - DeepSeekV3 0324 can_t load newest UD quants _with MLA_. Older quant wor.md
376 - Bug_ unknown model architecture_ _deci_ _when loading Llama-3_1-Nemotro.md
378 - Feature Request_ Use ik_llama.cpp with llama-cpp-python.md
379 - Bug_ Cannot build on WoA.md
380 - Drop at the start of generation.md
381 - ik_llama.cpp_ggml_src_ggml-cuda_fattn.cu_66_ fatal error after latest.md
383 - Bug_ Loading DeepSeek R1T Chimera causes _llama_model_load_ error loadi.md
387 - Bug_ bitnet 1.58 on termux segmentation fault.md
388 - Bug_ Clash with mainline llama.cpp .so files.md
389 - Bug_ llama-batched-bench crashed with batch size _2.md
398 - Bug_ -fmoe causing illegal memory access.md
407 - Feature Request_ Support for function calling in llama-server.md
412 - Bug_ Static asserts trip during compile..md
419 - qwen3 metrics in expert parallel_2x P100_.md
420 - Bug_ standard attention is broken.md
423 - Bug_ Compile failure undefined reference to _void mul_mat_q_case.md
425 - Bug_ CUDA error_ an illegal memory access was encountered.md
432 - Refactor_ GGUF v14 broke compatibility with IQx_KS quants.md
433 - Feature Request_ CORS support.md
436 - Bug_ Saving the prompt cache causes Segfault.md
437 - Feature Request_ support intel amx for further accelerate.md
440 - Feature Request_ Top n-sigma sampler.md
447 - Compilation Error_ Error C2676.md
450 - Bug_ Performance regression.md
452 - Falcon H1 Support.md
455 - Bug_ KV cache is never reused in OpenAI compatible Chat Completion api.md
456 - Bug_ no compilation without IQK_MULMAT.md
463 - Research_ V100 Flash Attention Implementation.md
464 - Bug_ The streaming every couple of rows blocks for 5-8s.md
467 - Bug_ Server does not send data_ _DONE_ for OpenAI-compatible streaming .md
472 - Bug_ Don_t build ggml-aarch64 regardless of CPU arch type.md
474 - Bug_ Perf Regression in PP throughput after Pull _461 _...R4 CUDA impl_.md
476 - Research_ performance divergence.md
479 - Bug_ _ggml_backend_cuda_graph_compute_ disabling CUDA graphs due to GPU.md
485 - Bug_ Illegal Memory Access loading model to CUDA1.md
490 - Bug_ Performance drop with 14292913 _461.md
498 - question_ about quantize method.md
499 - Bug_ cache quantization crash with IQK_FORCE_BF16.md
500 - Bug_ Insane cudaMalloc OOM Error on Dual 3090 GPUs.md
503 - Bug_ server_cli fails with segmentation fault.md
507 - Compatible gguf models _.md
514 - CUDA Kernel Error on RTX 5090 _Compute Capability 12.0_ _no kernel imag.md
521 - When offloading semi layers to some GPUs with -ot_ TG t_s performance t.md
522 - Bug_ disabling CUDA graphs due to mul_mat_id.md
523 - Bug_ tg speed drop after https_github.com_ikawrakow_ik_llama.cpp_pull_5.md
527 - Bug_ Webui improvement _481 core dump with a certain question..md
530 - Getting crash on second prompt..md
538 - Bug_ GGML_ASSERT failed at first prompt.md
539 - Bug_ garbage output.md
551 - Feature Request_ Support for Falcon Edge series.md
561 - Feature Request_ Tencent Hunyuan-A13B model support.md
568 - Feature Request_ ERNIE MoE Model Support.md
572 - Bug_ Oops_ggml_compute_forward_sum_rows_f32_ ffn_moe_weights_sum-60_ fo.md
575 - Bug_ llama-server crash with sampling order.md
576 - Bug_ llama-server crash with _Deepseek2 does not support K-shift_.md
596 - Bug_ Lastest commit broke llama-cli on Windows - mmq.cuh_107_ fatal err.md
597 - Feature Request_ Add THUDM_GLM-4-MoE-100B-A10B support.md
600 - Feature Request_ Port --reasoning-budget from main llamacpp _llamaserve.md
601 - Bug_ llama-imatrix crashing.md
605 - Bug_ IQ3_KS missing from GGMLQuantizationType - gguf_reader.py script c.md
614 - Feature Request_ port no-mmproj-offload.md
615 - Bug_ Gemma3 Vision not working.md
625 - Bug_ undefined symbol errors after successful compilation.md
626 - Feature Request_ Add IQK GEMM for IQ1_M.md
627 - Feature Request_ Tensor Parallelism.md
629 - Multi-GPU performance _Windows_ is significantly worse than single-GPU.md