44 Releases published by 1 person

• b5701

  published Jun 19, 2025

• b5702

  published Jun 19, 2025

• b5703

  published Jun 19, 2025

• b5704

  published Jun 19, 2025

• b5706

  published Jun 19, 2025

• b5707

  published Jun 19, 2025

• b5708

  published Jun 19, 2025

• b5709

  published Jun 19, 2025

• b5711

  published Jun 19, 2025

• b5712

  published Jun 20, 2025

• b5713

  published Jun 20, 2025

• b5714

  published Jun 20, 2025

• b5715

  published Jun 20, 2025

• b5716

  published Jun 20, 2025

• b5717

  published Jun 20, 2025

• b5718

  published Jun 20, 2025

• b5720

  published Jun 20, 2025

• b5719

  published Jun 20, 2025

• b5721

  published Jun 20, 2025

• b5723

  published Jun 20, 2025

• b5722

  published Jun 20, 2025

• b5726

  published Jun 20, 2025

• b5728

  published Jun 21, 2025

• b5729

  published Jun 21, 2025

• b5731

  published Jun 21, 2025

• b5733

  published Jun 22, 2025

• b5734

  published Jun 22, 2025

• b5735

  published Jun 22, 2025

• b5736

  published Jun 22, 2025

• b5737

  published Jun 22, 2025

• b5738

  published Jun 22, 2025

• b5740

  published Jun 22, 2025

• b5742

  published Jun 23, 2025

• b5743

  published Jun 23, 2025

• b5744

  published Jun 23, 2025

• b5745

  published Jun 23, 2025

• b5747

  published Jun 24, 2025

• b5749

  published Jun 24, 2025

• b5751

  published Jun 24, 2025

• b5752

  published Jun 24, 2025

• b5753

  published Jun 24, 2025

• b5754

  published Jun 25, 2025

• b5755

  published Jun 25, 2025

• b5756

  published Jun 25, 2025

55 Pull requests merged by 31 people

• ggml-cpu: enable IBM NNPA Vector Intrinsics

  #14317 merged Jun 25, 2025

• ggml : do not output unprintable characters on GGUF load failure

  #14381 merged Jun 25, 2025

• sycl: GGML_SYCL_DISABLE_OPT on by default for all Intel Devices

  #13973 merged Jun 25, 2025

• opencl: ref count ggml_backend_opencl_context and refactor profiling

  #14254 merged Jun 24, 2025

• batch : fix check for empty sequences in memory

  #14364 merged Jun 24, 2025

• cmake : use LLAMA_BUILD_NUMBER when defining LLAMA_INSTALL_VERSION

  #14362 merged Jun 24, 2025

• docs: Fix server API key doc for /props (move it to /health)

  #14352 merged Jun 24, 2025

• main : honor --verbose-prompt on interactive prompts

  #14350 merged Jun 24, 2025

• Add Mistral-Small-3.2-24B-Instruct-2506.jinja

  #14349 merged Jun 24, 2025

• CUDA/HIP: optimize mmv paths taken for HIP/CDNA

  #14324 merged Jun 23, 2025

• ci: add workflow for relocatable cmake package

  #14346 merged Jun 23, 2025

• vulkan: update windows SDK in release.yml

  #14344 merged Jun 23, 2025

• Fixes for rwkv-world template and the missing inputs.use_jinja in llama-cli

  #14336 merged Jun 23, 2025

• CUDA: mul_mat_v support for batch sizes > 1

  #14262 merged Jun 23, 2025

• kv-cells : fix tracking of seq_pos during cache reuse

  #14339 merged Jun 23, 2025

• vulkan: update windows SDK in CI

  #14334 merged Jun 23, 2025

• quantize: Handle user-defined pruning of whole layers (blocks)

  #13037 merged Jun 22, 2025

• gguf-py : fix SpecialVocab parsing when post_processor is null

  #14330 merged Jun 22, 2025

• run : avoid double tokenization

  #14327 merged Jun 22, 2025

• examples : fix is_first logic for tokenization

  #14329 merged Jun 22, 2025

• HIP: enable vec fattn on RDNA4

  #14323 merged Jun 22, 2025

• mtmd: fix Pixtral OOM with large images by capping image_size to 1024

  #14326 merged Jun 22, 2025

• common : use std::string_view now that we target c++17

  #14319 merged Jun 22, 2025

• CUDA: add mean operation

  #14313 merged Jun 22, 2025

• gguf-py : fix Qwen3-Embedding eos token

  #14314 merged Jun 21, 2025

• Add support for VK_EXT_debug_utils to add labels to Vulkan objects.

  #13792 merged Jun 21, 2025

• gguf-py : fix TemplateProcessing pair when bos/eos is missing

  #14312 merged Jun 21, 2025

• metal : fix thread-safety

  #14300 merged Jun 21, 2025

• memory : rename interface to llama_memory_context_i

  #14296 merged Jun 21, 2025

• Fix Llama 4 conversion

  #14311 merged Jun 21, 2025

• sync : ggml

  #14308 merged Jun 20, 2025

• docs : fix the link to llama.h

  #14293 merged Jun 20, 2025

• CUDA: add conv_2d_transpose

  #14287 merged Jun 20, 2025

• lint : remove trailing whitepace

  #14304 merged Jun 20, 2025

• vocab : prevent tokenizer overflow

  #14301 merged Jun 20, 2025

• sycl: add usage of enqueue_functions extension

  #14244 merged Jun 20, 2025

• Implement GGML_CPU_ALL_VARIANTS for PowerPC

  #14286 merged Jun 20, 2025

• llama : improve sep token handling

  #14272 merged Jun 20, 2025

• cuda : synchronize graph capture and cublas handle destruction

  #14288 merged Jun 20, 2025

• ggml : fix repack work size for mul_mat_id

  #14292 merged Jun 20, 2025

• ggml: Update KleidiAI to v1.9.0

  #14277 merged Jun 20, 2025

• model : more uniform output id handling

  #14275 merged Jun 20, 2025

• ubatch : new splitting logic

  #14217 merged Jun 20, 2025

• CUDA: add conv_2d_dw

  #14265 merged Jun 20, 2025

• ggml-cpu : remove unnecesary arm feature detection

  #14281 merged Jun 19, 2025

• gguf-py: Make sentencepiece optional

  #14200 merged Jun 19, 2025

• server: args for draft model cache types (#11200)

  #13782 merged Jun 19, 2025

• fix: resolve gcc compile warnings

  #14261 merged Jun 19, 2025

• sycl: Cleanup codepaths in Get Rows in sycl backend

  #14215 merged Jun 19, 2025

• llama-bench : add --no-warmup flag (#14224)

  #14270 merged Jun 19, 2025

• scripts: Fix remote option in Windows (#14102)

  #14100 merged Jun 19, 2025

• llamafile: support s390x SIMD instruction set

  #14273 merged Jun 19, 2025

• Vulkan: Fix host-pinned memory for large allocations

  #14249 merged Jun 19, 2025

• Hybrid recurrent cache

  #13979 merged Jun 19, 2025

• metal : add mean kernel

  #14267 merged Jun 19, 2025

26 Pull requests opened by 20 people

• ggml : add ggml_set_rows

  #14274 opened Jun 19, 2025

• kv-cache : use ggml_set_rows

  #14285 opened Jun 19, 2025

• Fix Windows Null Pointer Bug and Enhance Memory Operations in ggml-sycl

  #14290 opened Jun 20, 2025

• GitHub workflow: set RPATH to "@loader_path" / "$ORIGIN" to ensure executables and dynamic libraries search for dependencies in their origin directory.

  #14309 opened Jun 20, 2025

• ggml: adds CONV_2D op and direct GEMM Vulkan implementation

  #14316 opened Jun 21, 2025

• Conv2D: Add CPU version

  #14320 opened Jun 21, 2025

• Fix appearance of the chats list context menu for the browser Safari

  #14322 opened Jun 22, 2025

• make "serve" library

  #14331 opened Jun 22, 2025

• vulkan: lock accesses of pinned_memory vector

  #14333 opened Jun 22, 2025

• Make the shell scripts cross-platform

  #14341 opened Jun 23, 2025

• vulkan: Increase workgroup size for GLU, for performance

  #14345 opened Jun 23, 2025

• [server] webui DB import and export

  #14347 opened Jun 23, 2025

• Add script to test op perf and compare

  #14354 opened Jun 24, 2025

• build: refine toplevel .gitignore

  #14355 opened Jun 24, 2025

• llama : expose C API to get layer device type

  #14358 opened Jun 24, 2025

• server : fix assistant prefilling when content is an array

  #14360 opened Jun 24, 2025

• CUDA: add bf16 and f32 support to cublas_mul_mat_batched

  #14361 opened Jun 24, 2025

• llama : add high-throughput mode

  #14363 opened Jun 24, 2025

• ggml : add pointer to attach user data

  #14365 opened Jun 24, 2025

• vulkan: Add fusion support for RMS_NORM+MUL

  #14366 opened Jun 24, 2025

• test-backend-ops: add support for specifying output format

  #14368 opened Jun 25, 2025

• docs: fix broken url in main readme

  #14371 opened Jun 25, 2025

• Q2k interleaving implementation - x86/x64 SIMD

  #14373 opened Jun 25, 2025

• webui: preserve partial content when streaming errors occur

  #14374 opened Jun 25, 2025

• vulkan: handle noncontig in the final case of ggml_vk_get_cpy_pipeline

  #14378 opened Jun 25, 2025

• ggml-cpu: Build variant targeting Neoverse-V2

  #14380 opened Jun 25, 2025

47 Issues closed by 10 people

• Feature Request: Qwen 2.5 VL

  #11483 closed Jun 26, 2025

• Misc. bug: Model not loaded on Android with NDK

  #13399 closed Jun 26, 2025

• Eval bug: I cannot run llama 405b on CPU

  #13475 closed Jun 26, 2025

• web UI either doesn't scroll or jumps to the wrong element

  #13479 closed Jun 26, 2025

• Partial offload support for training

  #13486 closed Jun 26, 2025

• Misc. bug: ggml_cuda_compute_forward: MUL failed ROCm error: invalid device function

  #14370 closed Jun 25, 2025

• Misc. bug: llama-server slower on 4bit quantized model with f470bc36bed

  #14235 closed Jun 25, 2025

• Misc. bug: Completions hang after CUDA error, but health endpoint reports all OK

  #13281 closed Jun 25, 2025

• Misc. bug: The web UI of llama-server is not displaying correctly.

  #13428 closed Jun 25, 2025

• Compile bug: ld returned 1 exit status (file bigger than 2gb)

  #13446 closed Jun 25, 2025

• Drop support for sentencepiece

  #13448 closed Jun 25, 2025

• Misc. bug: Illegal CUDA memory access in ggml_backend_cuda_cpy_tensor_async

  #13449 closed Jun 25, 2025

• Feature Request: add draft model in llama-bench and more.

  #13456 closed Jun 25, 2025

• Misc. bug: llama-server webui overriding command line parameters

  #13277 closed Jun 24, 2025

• Eval bug: Regex

  #13347 closed Jun 24, 2025

• Differential mode for llama-bench + plotting code

  #13408 closed Jun 24, 2025

• Eval bug: Qwen3-30B-A3B-Q4_K_M: Slows down when using the \no_think mode.

  #13427 closed Jun 24, 2025

• Eval bug: llama-speculative core dump with Qwen3, GGML_ASSERT(batch.n_tokens > 0) failed

  #13433 closed Jun 24, 2025

• Misc. bug: Completion fails with error 500

  #14298 closed Jun 23, 2025

• Feature Request: update Windows actions builder to use *LATEST* Vulkan SDK (1.4.313 SDK) for VK_KHR_shader_bfloat16 support (like Linux builds)....

  #14230 closed Jun 23, 2025

• Eval bug: exmaple llama-simple-chat run failed in Android

  #14253 closed Jun 23, 2025

• Compile bug: I tried compiling llama.cpp for HIP on my system (elementaryOS 8/ubuntu 24.04, rocm 6.4.0, gfx1100) using the installation guide

  #13340 closed Jun 23, 2025

• Eval bug: llama-cli, Qwen3 jinja template will break CLI multiturn conversation

  #13404 closed Jun 23, 2025

• Misc. bug: convert: AttributeError: 'NoneType' object has no attribute 'get'

  #14328 closed Jun 22, 2025

• mtmd: Eval bug: Mistral small 2506 needs 1024x1024 image size cap

  #14310 closed Jun 22, 2025

• Feature Request: Add Support for ModernBert

  #11282 closed Jun 22, 2025

• Feature Request: allow setting jinja chat template from server webui

  #11689 closed Jun 22, 2025

• Misc. bug: Qwen 3.0 "enable_thinking" parameter not working

  #13160 closed Jun 22, 2025

• Token Generation Speed Decline with GGUF Models on M3 Ultra

  #13373 closed Jun 22, 2025

• Compile bug: ninja: build stopped: subcommand failed.

  #13375 closed Jun 22, 2025

• Misc. bug: invalid regex grammar causes segment violation

  #13390 closed Jun 22, 2025

• Feature Request: fix handling of Qwen3-Embedding-0.6B input to add EOS token

  #14252 closed Jun 21, 2025

• Not able to convert hf models to gguf anymore

  #14315 closed Jun 21, 2025

• Llama 4 mmproj fails `unable to find tensor mm.model.fc.weight`

  #14237 closed Jun 21, 2025

• Eval bug: IQ2_M broken for mradermacher / Llama-4-Maverick-17B-128E-Instruct-GGUF

  #12913 closed Jun 21, 2025

• Feature Request: tensor split needs control over where CPU layers go

  #13314 closed Jun 21, 2025

• Misc. bug: error in remote conversion for the new ServiceNow Nemotron 15B model

  #13354 closed Jun 21, 2025

• Compile bug: clang-18.1.3 compile fail (vsetivli)

  #13358 closed Jun 21, 2025

• Misc. bug: Extended swap/unswap times when loading large models on Apple Silicon

  #13361 closed Jun 21, 2025

• Misc. bug: `ggml_backend_reg_count()` returns 0 on linux from b5587 and onwards

  #14302 closed Jun 20, 2025

• Misc. bug: RPC immediate closing of the connection

  #14307 closed Jun 20, 2025

• Compile bug: [blas] choose blas backend to run llama2-7b model, but system info doesn't have the blas flag.

  #14259 closed Jun 20, 2025

• Eval bug: Abort is called in a thread from a custom thread pool during a llama_decode call

  #13990 closed Jun 20, 2025

• Eval bug: Qwen3 30B A3B Q4_0 failed to run

  #13168 closed Jun 20, 2025

• Misc. bug: batch in the mtmd-cli.cpp not freed

  #13620 closed Jun 19, 2025

• Feature Request: Add --no-warmup to llama-bench

  #14224 closed Jun 19, 2025

• Misc. bug: option --remote of convert_hf_to_gguf.py does not work in Windows

  #14102 closed Jun 19, 2025

15 Issues opened by 15 people

• Eval bug: Regression: Function calling using Mistral does not work anymore, refers to a character named Mistral in Tekken 7

  #14383 opened Jun 26, 2025

• Feature Request: allow running llama with an idle (lowest) priority as well

  #14382 opened Jun 25, 2025

• Feature Request: Exclude thinking tokens from server cache for reasoning models

  #14379 opened Jun 25, 2025

• Request for Official Support of AMD Ryzen AI Platform NPU

  #14377 opened Jun 25, 2025

• Compile error for ggml_gemv_q4_K_8x8_q8_K on Intel x86_64 MacOS (AVX2)

  #14372 opened Jun 25, 2025

• Misc. bug: Inconsistent Gemma3 implementation in rope factor

  #14367 opened Jun 24, 2025

• Misc. bug: llama-server assistant prefill only works when message content is a string (not a list of objects)

  #14353 opened Jun 24, 2025

• Feature Request: Suggest to provide armv7l version to run on Raspberry Pi devices.

  #14348 opened Jun 23, 2025

• OpenCL backend with Qualcomm Adreno GPUs load time is too long

  #14337 opened Jun 23, 2025

• Eval bug: Program crashes during long input inference when batch size is set to 16384

  #14325 opened Jun 22, 2025

• Feature Request: Add support for moonshotai/Kimi-VL-A3B-Instruct

  #14318 opened Jun 21, 2025

• Eval bug: When I store the model on my hard drive, llama.cpp attempts to load it and then says it's warming it up with a blank run after which it crashes the terminal session.

  #14297 opened Jun 20, 2025

• mtmd: Any plan for mtmd to support video input and audio output?

  #14295 opened Jun 20, 2025

• Eval bug: [CANN]AutoDL Ascend 910B instance running DeepSeek-r1 32B_Q8 error

  #14291 opened Jun 20, 2025

• Eval bug: Inconsistent Embedding Similarity between llama-server and LlamaCppEmbeddings for BGE-M3 Model

  #14280 opened Jun 19, 2025

69 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• imatrix: add option to display importance score statistics for a given imatrix file

  #12718 commented on Jun 25, 2025 • 11 new comments

• imatrix : use GGUF to store importance matrices

  #9400 commented on Jun 25, 2025 • 6 new comments

• llama : initial Mamba-2 support

  #9126 commented on Jun 25, 2025 • 0 new comments

• ggml: avoid rebuild of GGML graph for each token (#7456)

  #8366 commented on Jun 19, 2025 • 0 new comments

• Error while converting peft finetuned merged model to gguf

  #12494 commented on Jun 26, 2025 • 0 new comments

• Feature Request: Add support of convert.py for model Qwen2.5-Omni-7B

  #12641 commented on Jun 26, 2025 • 0 new comments

• Compile bug: Vulkan shaders build fails due to missing vulkan-shaders directory during ExternalProject\_Add configure step

  #13753 commented on Jun 26, 2025 • 0 new comments

• Compile bug: Vulkan Build Fails in Termux/Proot Due to Missing Cooperative Matrix Shader Variables

  #13801 commented on Jun 26, 2025 • 0 new comments

• Eval bug: Output garbled on DeepSeek-R1-Distill-Qwen-7B-Q4_K_M.gguf from unsloth using musa backend with VMM off

  #13788 commented on Jun 26, 2025 • 0 new comments

• ERROR:hf-to-gguf:Model Qwen2_5_VLModel is not supported

  #13802 commented on Jun 26, 2025 • 0 new comments

• ERROR:hf-to-gguf:Model MllamaForConditionalGeneration is not supported

  #13805 commented on Jun 26, 2025 • 0 new comments

• Misc. bug: ROCm images cannot be found

  #11913 commented on Jun 26, 2025 • 0 new comments

• Feature Request: (webui) do not throw away message if there is error in stream

  #13709 commented on Jun 25, 2025 • 0 new comments

• something with llama_server? slow vs llama_cli

  #13560 commented on Jun 25, 2025 • 0 new comments

• Misc. bug: RUNPATH properties are not properly set

  #13740 commented on Jun 25, 2025 • 0 new comments

• Eval bug: terminate called after throwing an instance of 'std::runtime_error' what(): Unexpected empty grammar stack after accepting piece: [control_36]

  #13690 commented on Jun 25, 2025 • 0 new comments

• Eval bug: Command-A generates a single repeating token when using split mode row on P40

  #14228 commented on Jun 24, 2025 • 0 new comments

• open source dataset for low bit quantization?

  #13736 commented on Jun 24, 2025 • 0 new comments

• Add SmolLM3

  #14240 commented on Jun 19, 2025 • 0 new comments

• MODEL: Falcon-H1 support

  #14238 commented on Jun 25, 2025 • 0 new comments

• ggml: introduce GGML_NUMA_MIGRATE to optimize cross NUMA op computation

  #14232 commented on Jun 25, 2025 • 0 new comments

• ggml : implement op fusion, starting with REGLU/GEGLU/SWIGLU

  #14158 commented on Jun 24, 2025 • 0 new comments

• ggml: aarch64: Implement SVE Kernels for Int 8 Quantization

  #14117 commented on Jun 25, 2025 • 0 new comments

• llama : support qwen3 rerank and embeddings

  #14029 commented on Jun 26, 2025 • 0 new comments

• Add plamo2

  #13930 commented on Jun 25, 2025 • 0 new comments

• convert: add eagle2 draft arch

  #13908 commented on Jun 24, 2025 • 0 new comments

• finetune.cpp command-line arg

  #13873 commented on Jun 25, 2025 • 0 new comments

• musa: enable fp16 mma (all) and cublas on qy2

  #13842 commented on Jun 25, 2025 • 0 new comments

• cmake : set `RPATH` to `$ORIGIN` on Linux (#13740)

  #13741 commented on Jun 25, 2025 • 0 new comments

• Granite Four

  #13550 commented on Jun 20, 2025 • 0 new comments

• cuda: set cuda compiler path (#13527)

  #13528 commented on Jun 20, 2025 • 0 new comments

• Added dynamic context size. This is perfect for servers running llama models as a service.

  #13295 commented on Jun 22, 2025 • 0 new comments

• WIP: Add support for CogAgent

  #12679 commented on Jun 24, 2025 • 0 new comments

• PR: Refine ggml-hexagon backend(Qualcomm Hexagon NPU backend) for latest ggml,whisper.cpp,llama.cpp

  #12326 commented on Jun 23, 2025 • 0 new comments

• [WIP]backend: Integrating QNN (Qualcomm AI Engine Direct) as a dedicated backend for Qualcomm NPUs

  #12063 commented on Jun 23, 2025 • 0 new comments

• Eval bug: A100 GPU not working with CUDA 12.8 in llama.cpp

  #13609 commented on Jun 21, 2025 • 0 new comments

• Eval bug: Output garbled in dual-GPU environment

  #13673 commented on Jun 21, 2025 • 0 new comments

• Misc. bug: AMX is not ready to be used!

  #13678 commented on Jun 21, 2025 • 0 new comments

• Eval bug: SYCL branch produces mul_mat bug when trying to run.

  #13674 commented on Jun 21, 2025 • 0 new comments

• devops/nix: `flake.lock` is very obsolete

  #13679 commented on Jun 21, 2025 • 0 new comments

• Eval bug: std::regex to split the text

  #13691 commented on Jun 21, 2025 • 0 new comments

• Feature Request: Support Jina V3 arch

  #9585 commented on Jun 20, 2025 • 0 new comments

• llama_eval removed, no deprecation info, still referenced in comments

  #14271 commented on Jun 20, 2025 • 0 new comments

• Eval bug: Cannot load Qwen3 ranking models

  #13820 commented on Jun 20, 2025 • 0 new comments

• Compile bug: llama.cpp-master/ggml/src/ggml-cpu/ggml-cpu-aarch64.cpp:80:54:error '_mm256_set_m128i' was not declared in this scope

  #11385 commented on Jun 20, 2025 • 0 new comments

• Eval bug: repeated output for llama-server

  #12782 commented on Jun 20, 2025 • 0 new comments

• How to start gemma3 multimodal model service using llama_server

  #13465 commented on Jun 20, 2025 • 0 new comments

• Eval bug: Not splitting model across rows correctly

  #13661 commented on Jun 20, 2025 • 0 new comments

• Feature Request: Procedure for reproducing test models

  #13662 commented on Jun 20, 2025 • 0 new comments

• Feature Request: Llama-bench improvement

  #13671 commented on Jun 20, 2025 • 0 new comments

• Feature Request: Save Model Name in Conversation Chats (WebUI)

  #13570 commented on Jun 19, 2025 • 0 new comments

• Feature Request:

  #13989 commented on Jun 19, 2025 • 0 new comments

• Feature Request: Add keep_alive function for llama-server

  #13748 commented on Jun 24, 2025 • 0 new comments

• Misc. bug: segfault in test-gbnf-validator

  #13762 commented on Jun 24, 2025 • 0 new comments

• changelog : `libllama` API

  #9289 commented on Jun 23, 2025 • 0 new comments

• Eval bug: Server and mtmd both crashing when starting Ultravox

  #13727 commented on Jun 23, 2025 • 0 new comments

• Eval bug: [CUDA] MoE model (Qwen3-30B-A3B) loads to GPU but does not utilize CUDA for inference in build b5466

  #13729 commented on Jun 23, 2025 • 0 new comments

• [Tracker] Docker build fails on CI for arm64

  #11888 commented on Jun 22, 2025 • 0 new comments

• Compile bug: gcc-12: error: unrecognized command-line option ‘-compress-mode=size’

  #14260 commented on Jun 22, 2025 • 0 new comments

• Misc. bug: Inconsistent Vulkan segfault

  #10528 commented on Jun 22, 2025 • 0 new comments

• Eval bug: KV cache stopped working in b5554 version

  #14071 commented on Jun 22, 2025 • 0 new comments

• Feature Request: add jina embeddings model availible convert to gguf

  #12327 commented on Jun 22, 2025 • 0 new comments

• Feature Request: support for image input in llama-server (and web ui)

  #12792 commented on Jun 22, 2025 • 0 new comments

• Misc. bug: -sm row results in gibberish output on HIP (ROCm 6.3.3)

  #13545 commented on Jun 22, 2025 • 0 new comments

• Eval bug: swa_full = true is slower than false

  #13683 commented on Jun 22, 2025 • 0 new comments

• Eval bug: Server Returns Empty Responses Under High Load

  #13703 commented on Jun 22, 2025 • 0 new comments

• Misc. bug: ./llama-server API max_completion_tokens Parameter Not Working

  #13700 commented on Jun 22, 2025 • 0 new comments

• tts : add support for Orpheus

  #12476 commented on Jun 21, 2025 • 0 new comments

• Misc. bug: Gemma3 multimodal (or all VL models?): </think> tag in the image or PDF text breaks prompt processing (or token generation?)

  #14143 commented on Jun 21, 2025 • 0 new comments