AI-ML
Ollama v0.17.8
RESUMEN
What's Changed parsers: repair unclosed arg_value tags in GLM tool calls by @BruceMacD in Reapply "don't require pulling stubs for cloud models" again by @jmorganca in * docs: format compat docs by @mxy
Descripción Detallada
What's Changed parsers: repair unclosed argvalue tags in GLM tool calls by @BruceMacD in Reapply "don't require pulling stubs for cloud models" again by @jmorganca in docs: format compat docs by @mxyng in create: fix localhost handling by @dhiltgen in build: smarter docker parallelism by @dhiltgen in mlx: int4 groupsize 64 by @pdevine in cloudproxy: handle stream disconnects gracefully by @drifkin in x/mlxrunner: replace sampler interface chain with single stateful Sampler by @pdevine in rocm: update linux to v7.2 by @dhiltgen in app: fix reset to defaults disabling auto-update by @hoyyeva in mlx: get parameters from modelfile during model creation by @pdevine in MLX: add header vendoring and remove go build tag by @dhiltgen in * ci: Fix windows build by @dhiltgen in Full Changelog:
Actualización de Ollama v0.17.8 con mejoras en parsers, manejo de localhost y correcciones en la documentación.
- Reparados tags arg_value no cerrados en llamadas de herramientas GLM.
- No es necesario descargar stubs para modelos en la nube.
- Mejoras en el manejo de localhost y desconexiones de stream.
- Correcciones en la documentación y en la construcción para Windows.
A quién le importa
Todos los que usan Ollama y necesitan las últimas correcciones y mejoras.
Generado por IA · puede contener errores
Releases Relacionados
AI-ML
Ollama v0.32.5
## What's Changed * Fixed an MLX Metal bug that could reduce output quality for NVFP4 models, particularly Laguna. **Full Changelog**: https://github.com/ollama/ollama/compare/v0.32.4...v0.32.5
AI-ML
Ollama v0.32.4
## What's Changed - Support Laguna on Apple GPUs via the MLX engine - Quantize draft-model output heads at the requested type when creating speculative-decoding drafts. - Fixed Qwen3 MoE decoding for differently-quantized experts, plus faster packed gate/up projection (~4–9% on M5 Max). **Full
AI-ML