Merge pull request #189 from silentoplayz/patch-1

Update env-configuration.md to remove depreciated env vars
This commit is contained in:
Justin Hayes 2024-08-21 22:15:12 -04:00 committed by GitHub
commit 825e6a4fca
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194

View File

@ -784,65 +784,6 @@ Query: [query]
- Type: `str`
- Description: Specifies the URL to the ComfyUI image generation API.
#### `COMFYUI_CFG_SCALE`
- Type: `int`
- Default: `7`
- Description: Specifies a `cfg_scale` value to use with ComfyUI. If you are using Stable Diffusion 3, the recommended value is `5.5`.
#### `COMFYUI_SAMPLER`
- Type: `str`
- Default: `euler`
- Description: Specifies a sampler to use with ComfyUI.
#### `COMFYUI_SCHEDULER`
- Type: `str`
- Default: `normal`
- Description: Specifies a scheduler to use with ComfyUI. If you are using Stable Diffusion 3, the recommended value is `sgm_uniform`.
#### `COMFYUI_SD3`
- Type: `bool`
- Default: `False`
- Description: Sets ComfyUI to Stable Diffusion 3 mode. SD3 will work without this set, but the image quality will be significantly lower. Requires a recent version of ComfyUI.
#### `COMFYUI_FLUX`
- Type: `bool`
- Default: `False`
- Description: Sets ComfyUI to Flux mode. Only set if Flux is required, as the workflow is completely different and is not at all compatible with other models.
:::info
[Find out more here](https://docs.openwebui.com/tutorial/images)
:::
#### `COMFYUI_FLUX_WEIGHT_DTYPE`
- Type: `str` (enum: `default`, `fp8_e4m3fn`)
- Options:
- `default` - Uses the default weight precision for Flux.
- `fp8_e4m3fn` - Uses 8-bit floating point precision for Flux, recommended for systems with limited VRAM.
- Default: `default`
- Description: Ignored if Flux is not enabled. Sets the weight precision for Flux.
#### `COMFYUI_FLUX_FP8_CLIP`
- Type: `bool`
- Default: `False`
- Description: Ignored if Flux is not enabled. Sets the Flux text encoder to 8-bit precision, saving VRAM.
:::tip
It is recommended to enable this environment variable if your machine doesn't have a lot of spare VRAM.
If enabled, the `t5xxl_fp8_e4m3fn.safetensors` file (can be downloaded [here](https://huggingface.co/comfyanonymous/flux_text_encoders/tree/main))
must exist in the `models/clip` directory in your ComfyUI installation.
:::
#### `IMAGES_OPENAI_API_KEY`
- Type: `str`