-
Notifications
You must be signed in to change notification settings - Fork 32
All issues
Issue creation is restricted in this repository
Issues
is:issue state:open
is:issue state:open
Search results
fix: Paged block budget test underflows when the paged-v2 workspace reserve exceeds the byte budget
area:coremlxcel-core: MLX FFI, primitives, KV cache, layersmlxcel-core: MLX FFI, primitives, KV cache, layerspriority:mediumMedium priorityMedium prioritystatus:readyReady to be worked onReady to be worked ontype:bugBug fixes, error corrections, or issue resolutionsBug fixes, error corrections, or issue resolutionsStatus: Open.#1091 In lablup/mlxcel;feat(core): Support mixed-bit-width q/k/v projections in the fused QKV loader
area:coremlxcel-core: MLX FFI, primitives, KV cache, layersmlxcel-core: MLX FFI, primitives, KV cache, layersarea:modelsModel architectures, weights, loading, metadataModel architectures, weights, loading, metadatapriority:mediumMedium priorityMedium prioritystatus:readyReady to be worked onReady to be worked ontype:enhancementNew features, capabilities, or significant additionsNew features, capabilities, or significant additionsStatus: Open.#1090 In lablup/mlxcel;fix(detect): docling-layout-heron boxes miss page content on tall pages
area:modelsModel architectures, weights, loading, metadataModel architectures, weights, loading, metadatapriority:mediumMedium priorityMedium prioritystatus:readyReady to be worked onReady to be worked ontype:bugBug fixes, error corrections, or issue resolutionsBug fixes, error corrections, or issue resolutionsStatus: Open.#1089 In lablup/mlxcel;test: verify-test-cuda is red on clean main because the gate leaves MLX TF32 enabled
area:coremlxcel-core: MLX FFI, primitives, KV cache, layersmlxcel-core: MLX FFI, primitives, KV cache, layerspriority:highHigh priorityHigh prioritystatus:readyReady to be worked onReady to be worked ontype:testTest related changesTest related changesStatus: Open.#1088 In lablup/mlxcel;fix: molmo-7b generates incoherent output on image prompts while molmo2-4b is correct
area:modelsModel architectures, weights, loading, metadataModel architectures, weights, loading, metadatapriority:mediumMedium priorityMedium prioritystatus:readyReady to be worked onReady to be worked ontype:bugBug fixes, error corrections, or issue resolutionsBug fixes, error corrections, or issue resolutionsStatus: Open.#1087 In lablup/mlxcel;fix(server): close shared gaps in the single-stream worker paths
area:inferenceGeneration, sampling, decoding (incl. speculative, DRY)Generation, sampling, decoding (incl. speculative, DRY)priority:mediumMedium priorityMedium prioritystatus:readyReady to be worked onReady to be worked ontype:bugBug fixes, error corrections, or issue resolutionsBug fixes, error corrections, or issue resolutionsStatus: Open.#1086 In lablup/mlxcel;fix(models): Florence-2 -large-ft releases generate degenerate output
area:modelsModel architectures, weights, loading, metadataModel architectures, weights, loading, metadatapriority:lowLow priorityLow prioritystatus:backlogIn the backlog, not yet readyIn the backlog, not yet readytype:bugBug fixes, error corrections, or issue resolutionsBug fixes, error corrections, or issue resolutionsStatus: Open.#1085 In lablup/mlxcel;[nightly-verify] the run did not finish
priority:highHigh priorityHigh prioritystatus:readyReady to be worked onReady to be worked ontype:bugBug fixes, error corrections, or issue resolutionsBug fixes, error corrections, or issue resolutionsStatus: Open.#1084 In lablup/mlxcel;fix(test): two autotune tests decide on microsecond thread::sleep accuracy and flake under host load
area:coremlxcel-core: MLX FFI, primitives, KV cache, layersmlxcel-core: MLX FFI, primitives, KV cache, layerspriority:lowLow priorityLow prioritystatus:readyReady to be worked onReady to be worked ontype:bugBug fixes, error corrections, or issue resolutionsBug fixes, error corrections, or issue resolutionsStatus: Open.#1079 In lablup/mlxcel;chore(packaging): decide the fate of debian/, which is maintained every release but has never shipped and cannot build
platform:linuxLinux (CUDA / packaging) specificLinux (CUDA / packaging) specificpriority:mediumMedium priorityMedium prioritystatus:readyReady to be worked onReady to be worked ontype:choreMaintenance tasks (build, CI, etc.)Maintenance tasks (build, CI, etc.)Status: Open.#1068 In lablup/mlxcel;fix(core): f32 matmul and single-query SDPA lose ~fp16 precision on M5, breaking four numeric tests
area:coremlxcel-core: MLX FFI, primitives, KV cache, layersmlxcel-core: MLX FFI, primitives, KV cache, layersarea:inferenceGeneration, sampling, decoding (incl. speculative, DRY)Generation, sampling, decoding (incl. speculative, DRY)platform:macosmacOS (Apple Silicon) specificmacOS (Apple Silicon) specificpriority:highHigh priorityHigh prioritystatus:readyReady to be worked onReady to be worked ontype:bugBug fixes, error corrections, or issue resolutionsBug fixes, error corrections, or issue resolutionsStatus: Open.#1065 In lablup/mlxcel;fix: Resolve flaky text_only_forward_produces_finite_logits in granite4_vision and hunyuan_vl parity tests
priority:mediumMedium priorityMedium prioritystatus:readyReady to be worked onReady to be worked ontype:bugBug fixes, error corrections, or issue resolutionsBug fixes, error corrections, or issue resolutionsStatus: Open.#997 In lablup/mlxcel;