All issues

搜索 results

Design check: Intel Arc (SYCL) backend

Status: Open.

#3921 In huggingface/candle;

· Hahihula opened on Aug 28, 2026

Unable to build whisper example

Status: Open.

#3915 In huggingface/candle;

· inferrna opened on Aug 24, 2026

qwen3_vl: no way to reset the KV cache, so a model instance is single-use

Status: Open.

#3901 In huggingface/candle;

· sidd-27 opened on Aug 17, 2026

CUDA is bound at link time, so a cuda-enabled binary cannot start without CUDA installed

Status: Open.

#3900 In huggingface/candle;

· nabbisen opened on Aug 17, 2026

F8E4M3 CUDA kernels are unreachable: DType::as_str() yields f8e4m3, the kernels are named f8_e4m3

Status: Open.

#3886 In huggingface/candle;

· astorise opened on Aug 14, 2026

get_or_load_func drops the kernel name from CUDA symbol-lookup errors

Status: Open.

#3884 In huggingface/candle;

· astorise opened on Aug 14, 2026

potential deadlock risk from multiple read locks over storage in tensor operations

Status: Open.

#3878 In huggingface/candle;

· scaspin opened on Aug 12, 2026

qwen3_5::Config deserializes without checking its fields agree, so an incoherent one fails deep instead of at load

Status: Open.

#3868 In huggingface/candle;

· astorise opened on Aug 9, 2026

ModelOpt checkpoints: the weight source stops at the projection; detection, shards and config are still on every consumer

Status: Open.

#3866 In huggingface/candle;

· astorise opened on Aug 8, 2026

Qwen3.5 quantized (GGUF) MoE checkpoints are not supported (qwen35moe)

Status: Open.

#3865 In huggingface/candle;

· astorise opened on Aug 8, 2026

QTensor::quantize on a narrowed view quantizes the parent (silently in release, debug_assert in debug)

Status: Open.

#3864 In huggingface/candle;

· oetiker opened on Aug 8, 2026

Metal: K-quant mul_mv kernels store past ne01; Q2K over-dispatches 2x on top of it

Status: Open.

#3863 In huggingface/candle;

· oetiker opened on Aug 8, 2026