media Hugging Face Forums · 2d ago · open_models

Cohesix 1.2.0 connects LoRA adapters to held-out comparisons and serving checks

from English

Cohesix 1.2.0 introduces a workflow that links trained adapters to base models for verification against held-out comparisons and real serving checks, including rollback capabilities. The release guide details this path, enabling users to submit jobs and inspect results on selected Mac MLX or Linux NVIDIA hosts.

Connects the adapter and base model to a held-out comparison.
Includes a real serving check to verify performance.
Provides a mechanism to revert to the previous version.
Supports execution on Mac MLX or Linux NVIDIA hardware.

Importance 1/3 Trust 1/3 Hugging Face Forums Inference efficiency Local inference

Read original

github llama.cpp · just now Live

llama.cpp b11302 fixes ThreadSanitizer data race in sparse indexer

The llama.cpp project released build b11302, which addresses a critical concurrency issue identified by ThreadSanitizer in the CI pipeline. The update resolves a data race within the sparse attention mechanism where multiple CPU threads were incorrectly writing to the same memory elements.

github llama.cpp · 1h ago · 1 view Live

llama.cpp b11302 fixes ThreadSanitizer data race in sparse indexer

llama.cpp b11298 adds dflash support to convert and feature extraction

llama.cpp b11284 fixes OpenVINO GET_ROWS on quantized weight views

llama.cpp adds GLM-5.3-Flash support with k-pool and MTP fixes

llama.cpp b11272 optimizes Hexagon concat op