-
Notifications
You must be signed in to change notification settings - Fork 9
All issues
Issue creation is restricted in this repository
Issues
is:issue state:open
is:issue state:open
Search results
Implementation of Agent-Driven Memory Management (Targeted RAG Extraction & Selective Context Pruning)
status: needs triageNew report awaiting maintainer classificationNew report awaiting maintainer classificationtype: featureA new capability or user-facing improvementA new capability or user-facing improvementStatus: Open.#82 In engeldlgado/toshllm;Ability to turn off reasoning/thinkinh
status: needs triageNew report awaiting maintainer classificationNew report awaiting maintainer classificationtype: featureA new capability or user-facing improvementA new capability or user-facing improvementStatus: Open.#81 In engeldlgado/toshllm;Does “Whisper” not work?
status: needs triageNew report awaiting maintainer classificationNew report awaiting maintainer classificationtype: featureA new capability or user-facing improvementA new capability or user-facing improvementStatus: Open.#73 In engeldlgado/toshllm;- Status: Open.#72 In engeldlgado/toshllm;
Tensor Parallelism Errors when using 2x Vega II Duo Cards and using q8_0 keys, turbo4 values
status: needs triageNew report awaiting maintainer classificationNew report awaiting maintainer classificationtype: bugA reproducible defect in ToshLLMA reproducible defect in ToshLLMStatus: Open.#69 In engeldlgado/toshllm;Display total VRAM from all GPUs and whether Infinity Fabric Link/Bridge is present in "Your machine" area
status: needs triageNew report awaiting maintainer classificationNew report awaiting maintainer classificationtype: featureA new capability or user-facing improvementA new capability or user-facing improvementStatus: Open.#67 In engeldlgado/toshllm;MTP head not recognized in gemma-4 models
status: needs triageNew report awaiting maintainer classificationNew report awaiting maintainer classificationtype: bugA reproducible defect in ToshLLMA reproducible defect in ToshLLMStatus: Open.#60 In engeldlgado/toshllm;Dynamic host-RAM tiering for large MoE models (Rust experiment)
status: needs triageNew report awaiting maintainer classificationNew report awaiting maintainer classificationtype: featureA new capability or user-facing improvementA new capability or user-facing improvementStatus: Open.#54 In engeldlgado/toshllm;- Status: Open.#53 In engeldlgado/toshllm;
MTP and split mode: tensor
status: needs triageNew report awaiting maintainer classificationNew report awaiting maintainer classificationtype: featureA new capability or user-facing improvementA new capability or user-facing improvementStatus: Open.#51 In engeldlgado/toshllm;Test: Infinity Fabric Link peer transfer (TOSH_MGPU_PEER) on W6800X Duo / Vega II Duo
area: engine/metalInference engine, Metal backend, kernels, and GPU executionInference engine, Metal backend, kernels, and GPU executionStatus: Open.#49 In engeldlgado/toshllm;Removing Henbury FB from DeviceProperties resulted in a 35% increase in token generation
hardware: hackintoshOpenCore or other non-Apple macOS hardwareOpenCore or other non-Apple macOS hardwarestatus: needs reproductionPlausible report that needs independent or repeated reproductionPlausible report that needs independent or repeated reproductiontype: performancePerformance regression, optimization, or benchmark findingPerformance regression, optimization, or benchmark findingStatus: Open.#44 In engeldlgado/toshllm;