49 Releases published by 1 person

• b5874

  published Jul 12, 2025

• b5875

  published Jul 12, 2025

• b5876

  published Jul 12, 2025

• b5880

  published Jul 12, 2025

• b5882

  published Jul 12, 2025

• b5884

  published Jul 12, 2025

• b5886

  published Jul 13, 2025

• b5887

  published Jul 13, 2025

• b5888

  published Jul 13, 2025

• b5889

  published Jul 13, 2025

• b5890

  published Jul 13, 2025

• b5891

  published Jul 14, 2025

• b5892

  published Jul 14, 2025

• b5893

  published Jul 14, 2025

• b5894

  published Jul 14, 2025

• b5895

  published Jul 14, 2025

• b5896

  published Jul 14, 2025

• b5897

  published Jul 14, 2025

• b5898

  published Jul 15, 2025

• b5899

  published Jul 15, 2025

• b5900

  published Jul 15, 2025

• b5901

  published Jul 15, 2025

• b5902

  published Jul 15, 2025

• b5904

  published Jul 16, 2025

• b5908

  published Jul 16, 2025

• b5909

  published Jul 16, 2025

• b5910

  published Jul 16, 2025

• b5911

  published Jul 16, 2025

• b5912

  published Jul 16, 2025

• b5913

  published Jul 16, 2025

• b5914

  published Jul 16, 2025

• b5915

  published Jul 16, 2025

• b5916

  published Jul 16, 2025

• b5919

  published Jul 17, 2025

• b5920

  published Jul 17, 2025

• b5921

  published Jul 17, 2025

• b5922

  published Jul 17, 2025

• b5923

  published Jul 17, 2025

• b5924

  published Jul 17, 2025

• b5927

  published Jul 18, 2025

• b5928

  published Jul 18, 2025

• b5929

  published Jul 18, 2025

• b5930

  published Jul 18, 2025

• b5932

  published Jul 18, 2025

• b5933

  published Jul 18, 2025

• b5934

  published Jul 18, 2025

• b5935

  published Jul 18, 2025

• b5936

  published Jul 18, 2025

• b5937

  published Jul 18, 2025

61 Pull requests merged by 26 people

• metal : fuse add, mul

  #14596 merged Jul 18, 2025

• graph : fix graph reuse reset of params

  #14760 merged Jul 18, 2025

• parallel : add option for different RNG seeds

  #14757 merged Jul 18, 2025

• Fix Gemma3n not executed as CUDA_GRAPH on NVGPUs

  #14741 merged Jul 18, 2025

• graph : avoid huge warm-up graphs for MoE models

  #14753 merged Jul 18, 2025

• model : fix build after merge conflict

  #14754 merged Jul 18, 2025

• Add EXAONE 4.0 model architecture

  #14630 merged Jul 18, 2025

• CUDA: set_rows + cpy.cu refactor

  #14712 merged Jul 18, 2025

• graph : refactor context to not pass gf explicitly

  #14629 merged Jul 18, 2025

• Move the graph placeholder message to debug mode

  #14748 merged Jul 18, 2025

• use max work group size for device to replace the magic number

  #14732 merged Jul 18, 2025

• Fix Ernie4.5 MoE without shared experts

  #14746 merged Jul 17, 2025

• nix: use optionalAttrs for env mkDerivation attrset argument

  #14726 merged Jul 17, 2025

• Model: Add support for Ernie 4.5 MoE

  #14658 merged Jul 17, 2025

• kv-cache : fix k-shift for multiple streams

  #14742 merged Jul 17, 2025

• llama : reuse compute graphs

  #14482 merged Jul 17, 2025

• model : fix parallel processing for lfm2

  #14705 merged Jul 17, 2025

• kv-cache : opt mask set input

  #14600 merged Jul 17, 2025

• batch : fix uninitialized has_cpl flag

  #14733 merged Jul 17, 2025

• ci : disable failing vulkan crossbuilds

  #14723 merged Jul 16, 2025

• convert : make hf token optional

  #14717 merged Jul 16, 2025

• Fix parameter order issue for hybrid memory initialization

  #14725 merged Jul 16, 2025

• ggml: Add initial WebGPU backend

  #14521 merged Jul 16, 2025

• Support Cosyvoice2-0.5B By allowing Qwen2 architecture to have a optional bias tensor

  #14711 merged Jul 16, 2025

• llama : add high-throughput mode

  #14363 merged Jul 16, 2025

• Support diffusion models: Add Dream 7B

  #14644 merged Jul 16, 2025

• ggml : add asserts

  #14720 merged Jul 16, 2025

• server : pre-calculate EOG logit biases

  #14721 merged Jul 16, 2025

• Bug: fix inputs to conv1d in mamba layer of plamo2

  #14716 merged Jul 16, 2025

• server : fix handling of the ignore_eos flag

  #14710 merged Jul 16, 2025

• scripts: synthetic prompt mode for server-bench.py

  #14695 merged Jul 16, 2025

• convert : only check for tokenizer folder if we need it

  #14704 merged Jul 16, 2025

• convert : add pre-computed hashes first to prevent order mishaps

  #14701 merged Jul 16, 2025

• llama: add LLAMA_API to deprecated llama_kv_self_seq_div

  #14708 merged Jul 16, 2025

• scripts: add bpw per layer and model

  #14703 merged Jul 15, 2025

• Model : Add support for Kimi-K2

  #14654 merged Jul 15, 2025

• vulkan: fix noncontig check for mat_mul_id splitting

  #14683 merged Jul 15, 2025

• vulkan: add RTE variants for glu/add/sub/mul/div

  #14653 merged Jul 15, 2025

• model : add PLaMo-2 model

  #14560 merged Jul 15, 2025

• cuda: fix build warnings in set-rows.cu (unused variable)

  #14687 merged Jul 15, 2025

• sycl: Hotfix for non dnnl codepath

  #14677 merged Jul 14, 2025

• PPC:Refactor llamafile_sgemm code

  #14673 merged Jul 14, 2025

• llama-context: add ability to get logits

  #14672 merged Jul 14, 2025

• scripts: benchmark for HTTP server throughput

  #14668 merged Jul 14, 2025

• SYCL: use 1D kernel for set_rows

  #14618 merged Jul 14, 2025

• sycl: Batched mulmat rework for oneDNN dispatch

  #14617 merged Jul 14, 2025

• llama : add jinja template for rwkv-world

  #14665 merged Jul 13, 2025

• quantize: fix minor logic flaw in --tensor-type

  #14572 merged Jul 13, 2025

• cuda : add set rows for bf16

  #14664 merged Jul 13, 2025

• Add ELU CUDA support

  #14657 merged Jul 13, 2025

• ggml : set_rows type coverage

  #14661 merged Jul 13, 2025

• Add missing unary ops Metal support

  #14660 merged Jul 13, 2025

• Add CMake presets for Linux and GCC

  #14656 merged Jul 13, 2025

• test-backend-ops : cover lfm2 cases in test_ssm_conv

  #14651 merged Jul 12, 2025

• readme : add LFM2 to models section

  #14650 merged Jul 12, 2025

• CUDA: add set rows for f32 and f16

  #14551 merged Jul 12, 2025

• sync : ggml

  #14648 merged Jul 12, 2025

• sync : ggml

  #14647 merged Jul 12, 2025

• server : fix pooled embedding output

  #14645 merged Jul 12, 2025

• vulkan: support SET_ROWS

  #14587 merged Jul 12, 2025

• vulkan: optimizations for deepseek prompt processing

  #14555 merged Jul 12, 2025

22 Pull requests opened by 19 people

• webui : add a preset feature to the settings

  #14649 opened Jul 12, 2025

• Add Pad Reflect 1D CUDA support

  #14659 opened Jul 13, 2025

• bug fix: handle saving/loading null layers in recurrent memory

  #14675 opened Jul 14, 2025

• kleidiai: add support for get_rows

  #14676 opened Jul 14, 2025

• Adding a simple-function-call example - hopefully not doing anything wrong

  #14682 opened Jul 14, 2025

• Fix KleidiAI compilation errors with -DGGML_NATIVE=OFF (issue #14464)

  #14700 opened Jul 15, 2025

• vulkan: Add logging for bf16 features to ggml_vk_print_gpu_info (#13274)

  #14707 opened Jul 16, 2025

• server: add prompt processing progress streaming for /completion endpoint #14685

  #14728 opened Jul 16, 2025

• mtmd : Support jinja in libmtmd (Only for QwenVL and Qwen Omni)

  #14730 opened Jul 17, 2025

• feat: Add optional prompt processing progress streaming

  #14731 opened Jul 17, 2025

• CUDA: skip masked out KQ slices in mma FA kernel

  #14735 opened Jul 17, 2025

• Documentation: Update build.md's Vulkan section

  #14736 opened Jul 17, 2025

• Improve Mistral models integration with llama.cpp

  #14737 opened Jul 17, 2025

• examples : predicted output for text generation

  #14739 opened Jul 17, 2025

• metal: SSM_SCAN performance

  #14743 opened Jul 17, 2025

• [ROCm] Fix HIP version check for HIPBLAS V2 API compatibility

  #14744 opened Jul 17, 2025

• Fix MinicpmV model converter and clip to avoid using hardcode.

  #14750 opened Jul 18, 2025

• tests : add non-cont K,V FA tests

  #14756 opened Jul 18, 2025

• cuda : implement bf16 cpy ops and enable bf16 cont

  #14763 opened Jul 18, 2025

• webui: add missing messages in export (#13552)

  #14764 opened Jul 18, 2025

• feat: Add extended sampling API with candidate token lists #14612

  #14765 opened Jul 19, 2025

• docs : mention apt installation method

  #14766 opened Jul 19, 2025

44 Issues closed by 15 people

• Eval bug: SIGILL

  #13161 closed Jul 19, 2025

• Eval bug: Can't run Qwen3-32B Q4_K_XL

  #13298 closed Jul 19, 2025

• Eval bug: Uncaught exception [json.exception.parse_error.101] during tool use crashes llama-server

  #13825 closed Jul 19, 2025

• Compile bug: numerous deprecation warnings when compiling in Termux

  #14011 closed Jul 19, 2025

• Feature Request: Support Llama-Nemotron-Nano-VL-8B-V1

  #14015 closed Jul 19, 2025

• Misc. bug: "error: invalid argument: /bin/sh" when using Docker image

  #14019 closed Jul 19, 2025

• Eval bug: b5922 causes gibberish on context shift

  #14759 closed Jul 18, 2025

• Eval bug: SYCL backend "invalid work-group size" error when using MoE models with Intel iGPU

  #14689 closed Jul 18, 2025

• Misc. bug: sentencepiece not included in requirements.txt

  #13982 closed Jul 18, 2025

• Compile bug:

  #13992 closed Jul 18, 2025

• Feature Request: allow spacebar to confirm web UI prompts [like the deleting a chat confirmation]

  #13999 closed Jul 18, 2025

• Feature Request: Add Ernie4.5MoE support

  #14465 closed Jul 17, 2025

• Compile bug: cannot compile get_rows_iq1_m

  #14542 closed Jul 17, 2025

• Misc. bug: mtmd cannot decode an image provided through valid OpenAI API request

  #14615 closed Jul 17, 2025

• Eval bug: Assertion failure when using LFM2 with parallel request processing

  #14670 closed Jul 17, 2025

• Eval bug: Llama-server on Mac starting at b5478 only producing 2-3 streaming tokens on Qwen3 and Deepseek R1 0528

  #14706 closed Jul 17, 2025

• Eval bug: microsoft/bitnet-b1.58-2B-4T-gguf

  #12997 closed Jul 17, 2025

• Feature Request: WINA

  #13964 closed Jul 17, 2025

• Eval bug: Unable to load the model on GPU

  #13967 closed Jul 17, 2025

• make using shifting context easier.

  #13969 closed Jul 17, 2025

• context shifting should be default option?

  #13971 closed Jul 17, 2025

• Misc. bug: llama-bench improper tensor split

  #13972 closed Jul 17, 2025

• Misc. bug: Hybrid models failing to load with assert GGML_ASSERT(kv_size % n_pad == 0)

  #14724 closed Jul 16, 2025

• Misc. bug: [SERVER] Multiple slots, generation speed is degraded after each generation/slot used

  #10860 closed Jul 16, 2025

• Data offline

  #14722 closed Jul 16, 2025

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

  #14379 closed Jul 16, 2025

• Eval bug: llama-tts abort

  #13955 closed Jul 16, 2025

• Eval bug: llama-mtmd-cli : option --image failed to load image

  #13959 closed Jul 16, 2025

• Eval bug: make_cpu_buft_list: no CPU backend found .... failed to load model

  #14691 closed Jul 15, 2025

• llama.cpp unable to compile for HIP

  #14694 closed Jul 15, 2025

• Feature Request: --swa-extra parameter needed to restore speculative decode function with SWA

  #13747 closed Jul 15, 2025

• Misc. bug: Decreased success rate for tool calling

  #13769 closed Jul 15, 2025

• Feature Request: Regarding Hardcoded GGML Tensor Name Length Limit (GGML_MAX_NAME)

  #13947 closed Jul 15, 2025

• Feature Request: Granite 4 Support

  #13275 closed Jul 14, 2025

• Compile bug: nvcc fatal : Unsupported gpu architecture 'compute_'

  #13893 closed Jul 14, 2025

• Feature Request: Generate Image Embeddings with llama.cpp

  #13913 closed Jul 14, 2025

• Compile bug: allocator.h:165:24 Call to implicitly-deleted copy constructor of 'std::unique_ptr<llama_adapter_lora, llama_adapter_lora_deleter>'

  #13925 closed Jul 14, 2025

• Eval bug: convert_hf_to_gguf.py: error: argument --outtype: invalid choice: 'q4_k_m'

  #14667 closed Jul 13, 2025

• Eval bug: Gemma 3n incoherent with HIP when prompt length > ubatch

  #14604 closed Jul 13, 2025

• Automatic optimization of runtime parameters such as -ngl given memory constraints

  #13860 closed Jul 13, 2025

• Feature Request: Make the `/completion` endpoint in `llama-server` work with multimodal models

  #13872 closed Jul 13, 2025

• Feature Request: Multimodal: llama-server support for Qwen2.5-VL chat template type: list of image paths (type: "video")

  #13905 closed Jul 13, 2025

• Eval bug:[5808-] qwen3 30B vulkan run with GGG

  #14583 closed Jul 12, 2025

• Misc. bug: Embedding/pooling: I receive 10xvector not 1xvector

  #14543 closed Jul 12, 2025

28 Issues opened by 27 people

• Feature Request: Direct FP8 conversion from convert_hf_to_gguf.py

  #14762 opened Jul 18, 2025

• Exaone-4 gibberish when using jinja template

  #14761 opened Jul 18, 2025

• Eval bug: Nemotron 49b doesnt load correctly

  #14752 opened Jul 18, 2025

• Error with unsloth DeepSeek-V3 BF16 and imatrix

  #14749 opened Jul 18, 2025

• Misc. bug: RPC flash attention bug on deepseek models (deepseek/kimi k2)

  #14747 opened Jul 17, 2025

• Compile bug: GGML Vulkan :: "format not a string literal and no format arguments [-Werror=format-security]"

  #14745 opened Jul 17, 2025

• Misc. bug: out of memory error after PR #13746

  #14740 opened Jul 17, 2025

• Misc. bug: b5921 release zip on github misses llama-embedding binary

  #14738 opened Jul 17, 2025

• Regarding the build for 8060S (gfx1151):

  #14734 opened Jul 17, 2025

• Eval bug: Nondeterministic output with ROCm backend despite zero temperature

  #14727 opened Jul 16, 2025

• Misc. bug: Llama server use some of the vram on another GPU, even I set -mg 1 and -sm 'none'

  #14719 opened Jul 16, 2025

• Misc. bug: After fine-tuning LLM-Research/Meta-Llama-3-8B-Instruct model with LLaMA Factory, an error occurs while converting it to the GGUF format.

  #14715 opened Jul 16, 2025

• Feature Request: Optimization of work for MOE architecture

  #14714 opened Jul 16, 2025

• Misc. bug: llamacpp crashes my PC whenever I close the console for it.

  #14713 opened Jul 16, 2025

• Compile bug: [SYCL][ARC A770] Regression: 双 A770 支持在 b5422 及以后版本失效

  #14709 opened Jul 16, 2025

• Misc. bug: OpenAI API v1/responses llama-server

  #14702 opened Jul 15, 2025

• Feature Request: ARMv7 / Termux Support on Mobile Devices

  #14699 opened Jul 15, 2025

• Eval bug: Gemma 3n on Vulkan fails to load

  #14698 opened Jul 15, 2025

• Eval bug: Regression: Tool calls still returned in content field as JSON string instead of tool_calls array

  #14697 opened Jul 15, 2025

• Eval bug: Unable to run with Qwen3 model

  #14696 opened Jul 15, 2025

• Compile bug: llama-llava-clip-quantize-cli not found

  #14693 opened Jul 15, 2025

• Eval bug: CUDA error: operation not supported

  #14692 opened Jul 15, 2025

• Misc. bug: Meta-Llama-3-8B-Instruct could not convert to .guuf. error:FileNotFoundError: File not found: /mnt/workspace/LLaMA-Factory/output/llama3_lora_sft/tokenizer.model

  #14690 opened Jul 15, 2025

• Feature Request: Server stream response for "prompt processing progress"

  #14685 opened Jul 15, 2025

• Misc. bug: DeepSeek-R1 0528 671b:Q4_K_XL think tags do not close sometimes

  #14679 opened Jul 14, 2025

• Eval bug: Qwen 2.5 VL gets stuck in a loop

  #14663 opened Jul 13, 2025

• Bunch of blank lines in prompt lead to segmentation fault in tokenizer with Qwen3

  #14655 opened Jul 12, 2025

• Feature Request: Add Explicit Context Reset for llama-cli or llama-server

  #14652 opened Jul 12, 2025

64 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.

• Add CUDA non-contiguous Unary Ops support

  #14639 commented on Jul 15, 2025 • 11 new comments

• ggml: adds CONV_2D op and direct GEMM Vulkan implementation

  #14316 commented on Jul 18, 2025 • 7 new comments

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

  #12718 commented on Jul 13, 2025 • 2 new comments

• OpenCL: add `mul_mat_f16_f32_image` kernel

  #14635 commented on Jul 15, 2025 • 2 new comments

• tool: add convertation of text/parquet to custom format

  #14622 commented on Jul 18, 2025 • 2 new comments

• Introduce New Lookup-Table(LUT)-Based Matrix Multiplication Method (TMAC)

  #13206 commented on Jul 17, 2025 • 1 new comment

• finetune.cpp command-line arg

  #13873 commented on Jul 18, 2025 • 1 new comment

• ggml: introduce GGML_NUMA_MIGRATE to optimize cross NUMA op computation

  #14232 commented on Jul 15, 2025 • 1 new comment

• OpenCL: add conv2d kernel

  #14403 commented on Jul 18, 2025 • 1 new comment

• Allow truncation when embedding

  #14493 commented on Jul 18, 2025 • 1 new comment

• docker : add cann build pipline

  #14591 commented on Jul 17, 2025 • 1 new comment

• common: add config presets for falcon

  #14638 commented on Jul 12, 2025 • 1 new comment

• Misc. bug: weird cursor placement in the web UI

  #14233 commented on Jul 18, 2025 • 0 new comments

• Misc. bug: prompt as pasted content in the server

  #14251 commented on Jul 18, 2025 • 0 new comments

• android built on GPU cannot comparable with CPU?

  #13910 commented on Jul 18, 2025 • 0 new comments

• Misc. bug: missing messages in JSON export via llama-server web UI

  #13552 commented on Jul 18, 2025 • 0 new comments

• Feature Request: s390x CI

  #13243 commented on Jul 18, 2025 • 0 new comments

• Feature Request: Gemma3n multimodal support

  #14429 commented on Jul 18, 2025 • 0 new comments

• main: failed to quantize model from 'gemma-3n-E2B-it.f16.gguf'

  #14405 commented on Jul 18, 2025 • 0 new comments

• Misc. bug: [CANN] memory leaky using CANN as backend

  #14257 commented on Jul 19, 2025 • 0 new comments

• Feature Request: Improve Sampling API: Expose Top‑K/Top‑P Candidate Token Lists in C API

  #14612 commented on Jul 19, 2025 • 0 new comments

• Revert "ggml : remove OpenCL (#7735) + (#8235)"

  #8986 commented on Jul 17, 2025 • 0 new comments

• imatrix : use GGUF to store importance matrices

  #9400 commented on Jul 19, 2025 • 0 new comments

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

  #12063 commented on Jul 18, 2025 • 0 new comments

• llama-server : implement universal assisted decoding

  #12635 commented on Jul 15, 2025 • 0 new comments

• Update llama-quant.cpp llama_tensor_get_type with DeepSeek friendly modifications

  #12727 commented on Jul 18, 2025 • 0 new comments

• HIP: Enable Matrix cores for MMQ Kernels, Enable stream-K for CDNA 3

  #14624 commented on Jul 17, 2025 • 0 new comments

• llama : support qwen3 rerank and embeddings

  #14029 commented on Jul 13, 2025 • 0 new comments

• metal : reuse graphs

  #14570 commented on Jul 17, 2025 • 0 new comments

• musa: upgrade musa sdk to 4.2.0

  #14498 commented on Jul 18, 2025 • 0 new comments

• Q2k interleaving implementation - x86/x64 SIMD

  #14373 commented on Jul 17, 2025 • 0 new comments

• [CANN] weight format to nz for Ascend310P3

  #14407 commented on Jul 16, 2025 • 0 new comments

• Misc. bug: linux/arm64 does not exist for the server docker image

  #13891 commented on Jul 12, 2025 • 0 new comments

• Metrics should not include : in Prometheus metric names

  #14150 commented on Jul 13, 2025 • 0 new comments

• Eval bug: (MAC) fail in `GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_FLASH_ATTN_EXT_Q8_0_H96, flash_attn_ext_q8_0_h96, has_simdgroup_mm);`

  #14110 commented on Jul 13, 2025 • 0 new comments

• Misc. bug: llama-server webui with --jinja flag does not show thinking when using reasoning models

  #14007 commented on Jul 13, 2025 • 0 new comments

• prismatic-vlms to gguf?

  #14159 commented on Jul 14, 2025 • 0 new comments

• Research: mmap eviction

  #14154 commented on Jul 14, 2025 • 0 new comments

• Feature request: Graphical GGUF viewer

  #6715 commented on Jul 14, 2025 • 0 new comments

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

  #14337 commented on Jul 14, 2025 • 0 new comments

• Feature Request: add tool calling for deepseek-r1-0528

  #14557 commented on Jul 15, 2025 • 0 new comments

• Feature Request: Support EXAONE 4.0

  #14474 commented on Jul 15, 2025 • 0 new comments

• Misc. bug: convert_hf_to_gguf.py not working on qwen3-embedding and qwen3-embedding lora tuned models

  #14459 commented on Jul 15, 2025 • 0 new comments

• Compile bug: zero-size array ‘gemm_gemv_kernels’ / invalid feature modifier ‘sme’

  #14464 commented on Jul 15, 2025 • 0 new comments

• ggml : add WebGPU backend

  #7773 commented on Jul 16, 2025 • 0 new comments

• Misc. bug: full-cuda docker build needs ldconfig before launching llama-*

  #14195 commented on Jul 16, 2025 • 0 new comments

• Misc. bug: LLAMA-SERVER is 40% slower than LLAMA-CLI when using identical parameters including -ot option for tensor offloading

  #14201 commented on Jul 16, 2025 • 0 new comments

• Misc. bug: evaluate_and_capture_cuda_graph NULL POINTER DEREFERENCE

  #14186 commented on Jul 16, 2025 • 0 new comments

• Misc. bug: Failure to allocate buffer with ROCm 6.4

  #14178 commented on Jul 16, 2025 • 0 new comments

• Misc. bug: Potential out of bound in rerank

  #13549 commented on Jul 16, 2025 • 0 new comments

• Misc. bug: Qwen3-Embedding-0.6B-GGUF doesn't work for 32768 context size (too much memory used)

  #14084 commented on Jul 16, 2025 • 0 new comments

• changelog : `libllama` API

  #9289 commented on Jul 16, 2025 • 0 new comments

• Misc. bug: crash on vulkan with new max mem alloc size calculations since b5703

  #14553 commented on Jul 16, 2025 • 0 new comments

• Feature Request: Generic CPU in ggml-cpu/arch

  #14402 commented on Jul 16, 2025 • 0 new comments

• Feature Request: Support Kimi K2

  #14642 commented on Jul 16, 2025 • 0 new comments

• Feature Request: llama-server: a flag for limiting input image size

  #14216 commented on Jul 17, 2025 • 0 new comments

• Misc. bug: OAI response_format json_schema and json_object not applied with Llama 3.x models

  #14218 commented on Jul 17, 2025 • 0 new comments

• Eval bug: RWKV inference with llama-parallel gets wrong output with lmhead offloaded to GPU

  #14211 commented on Jul 17, 2025 • 0 new comments

• Misc. bug: [Windows] GPU layers/tensors still consume system memory after load when mmap = true

  #14187 commented on Jul 17, 2025 • 0 new comments

• Misc. bug: Stuck while loading the model

  #14114 commented on Jul 17, 2025 • 0 new comments

• Feature Request: Support GLM-4.1V-9B-Thinking

  #14495 commented on Jul 17, 2025 • 0 new comments

• Feature Request: Add support for Kokoro TTS

  #11050 commented on Jul 17, 2025 • 0 new comments

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

  #14280 commented on Jul 17, 2025 • 0 new comments

• Misc. bug: Complex tool calling schema causes an "Unrecognized Schema" exception

  #14227 commented on Jul 17, 2025 • 0 new comments