12月1月2月3月4月5月6月7月8月9月10月11月
周一周三周五
一年内 265876 次贡献
更少更多
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/peft 分支比较 3 提交 »e315fb9e7b WIP - still need to figure out why the keys are wrong.827ac4b841 Remove RawModel - it was just creating a weird layer of indirection in the AnyModel type without adding any value.73e5b08c1f Add peft dependency. Bump diffusers.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/multi-diffusion-batched 分支比较 10 提交 »6bcf48aa37 WIP - Started working towards MultiDiffusion batching.b1bb1511fe Delete rough notes.99046a8145 Fix advanced scheduler behaviour in MultiDiffusionPipeline.72be7e71e3 Fix handling of stateful schedulers in MultiDiffusionPipeline.35adaf1c17 Connect TiledMultiDiffusionDenoiseLatents to the MultiDiffusionPipeline backend.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/modular-inpaint-suggestion 分支9db0e9d696 Add better documentation/errors around the possibility that inpainting models may be incorrectly labelled as non-inpainting models.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/model-offload-prototype-lora-bnb-autocast-tensor 分支比较 10 提交 »510ed6ed1f Make CachedModelWithPartialLoad work with models that have non-persistent buffers.6bb3c96fa5 Update CachedModelWithPartialLoad to use new autocast modules.52c5061b99 Add CustomInvokeLinearNF4 to enable CPU -> GPU streaming for InvokeLinearNF4 layers.e3b1078e5e Add CustomInvokeLinear8bitLt layer for device streaming with InvokeLinear8bitLt layers.ea552821cd Simplify the state management in InvokeLinear8bitLt and add unit tests. This is in preparation for wrapping it to support streaming of weights from cpu to gpu.
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/model-offload-prototype-lora-bnb 分支比较 3 提交 »f01e41ceaf First pass at dynamically calculating the working memory requirements for the VAE decoding operation. Still need to tune SD3 and FLUX.609ed06265 Add AutoencoderKL to the list of models that opt-out of partial loading.f9e899a6ba Make pinned pytorch version slightly more specific. We need at least 2.4 for access to torch.nn.functional.rms_norm(...).
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/model-offload-prototype-lora-alternative-autocast-impl 分支f109914eb3 WIP - messing around with some alternative autocast implementations
- frostbyte_neo 推送了仓库 github-featured/invoke-ai--invokeai 的 ryan/model-offload-4-model-cache-partial-load-rb3.5 分支比较 10 提交 »b0be545381 Change definition of VRAM in use for the ModelCache from sum of model weights to the total torch.cuda.memory_allocated().5cfd6ee3e9 Add get_effective_device(...) utility to aid in determining the effective device of models that are partially loaded.d8ab66c139 Handle device casting in ia2_layer.py.146e6611a3 Remove all cases where we check the 'model.device'. This is no longer trustworthy now that partial loading is permitted.43c3da16d2 Improve handling of cases when application code modifies the size of a model after registering it with the model cache.