upstream-archive/stable-diffusion.cpp
1
0
Fork 0
mirror of https://github.com/leejet/stable-diffusion.cpp.git synced 2026-07-23 11:20:53 -05:00
Code Issues Packages Projects Releases Wiki Activity

feat: add more built-in highres upscalers (#1456)

Browse Source
This commit is contained in:
leejet
2026-04-23 22:17:58 +08:00
committed by GitHub
parent c97702e105
commit b8bdffc199
11 changed files with 624 additions and 209 deletions
Download Patch File Download Diff File Expand all files Collapse all files

136
examples/server/README.md
View File

@@ -123,11 +123,11 @@ In this case, the server will load and serve the specified `index.html` file ins
usage: ./bin/sd-server [options]
Svr Options:
-l, --listen-ip <string> server listen ip (default: 127.0.0.1)
-l, --listen-ip <string> server listen ip (default: 127.0.0.1)
--serve-html-path <string> path to HTML file to serve at root (optional)
--listen-port <int> server listen port (default: 1234)
-v, --verbose print extra info
--color colors the logging tags according to level
--color colors the logging tags according to level
-h, --help show this help message and exit
Context Options:
@@ -136,7 +136,8 @@ Context Options:
--clip_g <string> path to the clip-g text encoder
--clip_vision <string> path to the clip-vision encoder
--t5xxl <string> path to the t5xxl text encoder
--llm <string> path to the llm text encoder. For example: (qwenvl2.5 for qwen-image, mistral-small3.2 for flux2, ...)
--llm <string> path to the llm text encoder. For example: (qwenvl2.5 for qwen-image,
mistral-small3.2 for flux2, ...)
--llm_vision <string> path to the llm vit
--qwen2vl <string> alias of --llm. Deprecated.
--qwen2vl_vision <string> alias of --llm_vision. Deprecated.
@@ -148,16 +149,16 @@ Context Options:
--control-net <string> path to control net model
--embd-dir <string> embeddings directory
--lora-model-dir <string> lora model directory
--hires-upscalers-dir <string> highres fix upscaler model directory
--tensor-type-rules <string> weight type per tensor pattern (example: "^vae\.=f16,model\.=q8_0")
--photo-maker <string> path to PHOTOMAKER model
--upscale-model <string> path to esrgan model.
-t, --threads <int> number of threads to use during computation (default: -1). If threads <= 0, then threads will be set to the number of
CPU physical cores
-t, --threads <int> number of threads to use during computation (default: -1). If threads <= 0,
then threads will be set to the number of CPU physical cores
--chroma-t5-mask-pad <int> t5 mask pad size of chroma
--vae-tile-overlap <float> tile overlap for vae tiling, in fraction of tile size (default: 0.5)
--vae-tiling process vae in tiles to reduce memory usage
--force-sdxl-vae-conv-scale force use of conv scale on sdxl vae
--offload-to-cpu place the weights in RAM to save VRAM, and automatically load them into VRAM when needed
--offload-to-cpu place the weights in RAM to save VRAM, and automatically load them into VRAM
when needed
--mmap whether to memory-map model
--control-net-cpu keep controlnet in cpu (for low vram)
--clip-on-cpu keep clip in cpu (for low vram)
@@ -172,20 +173,19 @@ Context Options:
--chroma-disable-dit-mask disable dit mask for chroma
--qwen-image-zero-cond-t enable zero_cond_t for qwen image
--chroma-enable-t5-mask enable t5 mask for chroma
--type weight type (examples: f32, f16, q4_0, q4_1, q5_0, q5_1, q8_0, q2_K, q3_K, q4_K). If not specified, the default is the
type of the weight file
--type weight type (examples: f32, f16, q4_0, q4_1, q5_0, q5_1, q8_0, q2_K, q3_K,
q4_K). If not specified, the default is the type of the weight file
--rng RNG, one of [std_default, cuda, cpu], default: cuda(sd-webui), cpu(comfyui)
--sampler-rng sampler RNG, one of [std_default, cuda, cpu]. If not specified, use --rng
--prediction prediction type override, one of [eps, v, edm_v, sd3_flow, flux_flow, flux2_flow]
--lora-apply-mode the way to apply LoRA, one of [auto, immediately, at_runtime], default is auto. In auto mode, if the model weights
contain any quantized parameters, the at_runtime mode will be used; otherwise,
immediately will be used.The immediately mode may have precision and
compatibility issues with quantized parameters, but it usually offers faster inference
speed and, in some cases, lower memory usage. The at_runtime mode, on the
other hand, is exactly the opposite.
--vae-tile-size tile size for vae tiling, format [X]x[Y] (default: 32x32)
--vae-relative-tile-size relative tile size for vae tiling, format [X]x[Y], in fraction of image size if < 1, in number of tiles per dim if >=1
(overrides --vae-tile-size)
--prediction prediction type override, one of [eps, v, edm_v, sd3_flow, flux_flow,
flux2_flow]
--lora-apply-mode the way to apply LoRA, one of [auto, immediately, at_runtime], default is
auto. In auto mode, if the model weights contain any quantized parameters,
the at_runtime mode will be used; otherwise, immediately will be used.The
immediately mode may have precision and compatibility issues with quantized
parameters, but it usually offers faster inference speed and, in some cases,
lower memory usage. The at_runtime mode, on the other hand, is exactly the
opposite.
Default Generation Options:
-p, --prompt <string> the prompt to render
@@ -194,65 +194,97 @@ Default Generation Options:
--end-img <string> path to the end image, required by flf2v
--mask <string> path to the mask image
--control-image <string> path to control image, control net
--control-video <string> path to control video frames, It must be a directory path. The video frames inside should be stored as images in
lexicographical (character) order. For example, if the control video path is
`frames`, the directory contain images such as 00.png, 01.png, ... etc.
--control-video <string> path to control video frames, It must be a directory path. The video frames
inside should be stored as images in lexicographical (character) order. For
example, if the control video path is `frames`, the directory contain images
such as 00.png, 01.png, ... etc.
--pm-id-images-dir <string> path to PHOTOMAKER input id images dir
--pm-id-embed-path <string> path to PHOTOMAKER v2 id embed
--hires-upscaler <string> highres fix upscaler, Lanczos, Nearest, Latent, Latent (nearest), Latent
(nearest-exact), Latent (antialiased), Latent (bicubic), Latent (bicubic
antialiased), or a model name under --hires-upscalers-dir (default: Latent)
-H, --height <int> image height, in pixel space (default: 512)
-W, --width <int> image width, in pixel space (default: 512)
--steps <int> number of sample steps (default: 20)
--high-noise-steps <int> (high noise) number of sample steps (default: -1 = auto)
--clip-skip <int> ignore last layers of CLIP network; 1 ignores none, 2 ignores one layer (default: -1). <= 0 represents unspecified,
will be 1 for SD1.x, 2 for SD2.x
--clip-skip <int> ignore last layers of CLIP network; 1 ignores none, 2 ignores one layer
(default: -1). <= 0 represents unspecified, will be 1 for SD1.x, 2 for SD2.x
-b, --batch-count <int> batch count
--video-frames <int> video frames (default: 1)
--fps <int> fps (default: 24)
--timestep-shift <int> shift timestep for NitroFusion models (default: 0). recommended N for NitroSD-Realism around 250 and 500 for
NitroSD-Vibrant
--timestep-shift <int> shift timestep for NitroFusion models (default: 0). recommended N for
NitroSD-Realism around 250 and 500 for NitroSD-Vibrant
--upscale-repeats <int> Run the ESRGAN upscaler this many times (default: 1)
--upscale-tile-size <int> tile size for ESRGAN upscaling (default: 128)
--hires-width <int> highres fix target width, 0 to use --hires-scale (default: 0)
--hires-height <int> highres fix target height, 0 to use --hires-scale (default: 0)
--hires-steps <int> highres fix second pass sample steps, 0 to reuse --steps (default: 0)
--hires-upscale-tile-size <int> highres fix upscaler tile size, reserved for model-backed upscalers (default:
128)
--cfg-scale <float> unconditional guidance scale: (default: 7.0)
--img-cfg-scale <float> image guidance scale for inpaint or instruct-pix2pix models: (default: same as --cfg-scale)
--img-cfg-scale <float> image guidance scale for inpaint or instruct-pix2pix models: (default: same
as --cfg-scale)
--guidance <float> distilled guidance scale for models with guidance input (default: 3.5)
--slg-scale <float> skip layer guidance (SLG) scale, only for DiT models: (default: 0). 0 means disabled, a value of 2.5 is nice for sd3.5
medium
--slg-scale <float> skip layer guidance (SLG) scale, only for DiT models: (default: 0). 0 means
disabled, a value of 2.5 is nice for sd3.5 medium
--skip-layer-start <float> SLG enabling point (default: 0.01)
--skip-layer-end <float> SLG disabling point (default: 0.2)
--eta <float> noise multiplier (default: 0 for ddim_trailing, tcd, res_multistep and res_2s; 1 for euler_a, er_sde and dpm++2s_a)
--eta <float> noise multiplier (default: 0 for ddim_trailing, tcd, res_multistep and
res_2s; 1 for euler_a, er_sde and dpm++2s_a)
--flow-shift <float> shift value for Flow models like SD3.x or WAN (default: auto)
--high-noise-cfg-scale <float> (high noise) unconditional guidance scale: (default: 7.0)
--high-noise-img-cfg-scale <float> (high noise) image guidance scale for inpaint or instruct-pix2pix models (default: same as --cfg-scale)
--high-noise-guidance <float> (high noise) distilled guidance scale for models with guidance input (default: 3.5)
--high-noise-slg-scale <float> (high noise) skip layer guidance (SLG) scale, only for DiT models: (default: 0)
--high-noise-img-cfg-scale <float> (high noise) image guidance scale for inpaint or instruct-pix2pix models
(default: same as --cfg-scale)
--high-noise-guidance <float> (high noise) distilled guidance scale for models with guidance input
(default: 3.5)
--high-noise-slg-scale <float> (high noise) skip layer guidance (SLG) scale, only for DiT models: (default:
0)
--high-noise-skip-layer-start <float> (high noise) SLG enabling point (default: 0.01)
--high-noise-skip-layer-end <float> (high noise) SLG disabling point (default: 0.2)
--high-noise-eta <float> (high noise) noise multiplier (default: 0 for ddim_trailing, tcd, res_multistep and res_2s; 1 for euler_a, er_sde and dpm++2s_a)
--high-noise-eta <float> (high noise) noise multiplier (default: 0 for ddim_trailing, tcd,
res_multistep and res_2s; 1 for euler_a, er_sde and dpm++2s_a)
--strength <float> strength for noising/unnoising (default: 0.75)
--pm-style-strength <float>
--control-strength <float> strength to apply Control Net (default: 0.9). 1.0 corresponds to full destruction of information in init image
--moe-boundary <float> timestep boundary for Wan2.2 MoE model. (default: 0.875). Only enabled if `--high-noise-steps` is set to -1
--pm-style-strength <float>
--control-strength <float> strength to apply Control Net (default: 0.9). 1.0 corresponds to full
destruction of information in init image
--moe-boundary <float> timestep boundary for Wan2.2 MoE model. (default: 0.875). Only enabled if
`--high-noise-steps` is set to -1
--vace-strength <float> wan vace strength
--increase-ref-index automatically increase the indices of references images based on the order they are listed (starting with 1).
--vae-tile-overlap <float> tile overlap for vae tiling, in fraction of tile size (default: 0.5)
--hires-scale <float> highres fix scale when target size is not set (default: 2.0)
--hires-denoising-strength <float> highres fix second pass denoising strength (default: 0.7)
--increase-ref-index automatically increase the indices of references images based on the order
they are listed (starting with 1).
--disable-auto-resize-ref-image disable auto resize of ref images
--disable-image-metadata do not embed generation metadata on image files
--vae-tiling process vae in tiles to reduce memory usage
--hires enable highres fix
-s, --seed RNG seed (default: 42, use random seed for < 0)
--sampling-method sampling method, one of [euler, euler_a, heun, dpm2, dpm++2s_a, dpm++2m, dpm++2mv2, ipndm, ipndm_v, lcm, ddim_trailing,
tcd, res_multistep, res_2s, er_sde] (default: euler for Flux/SD3/Wan, euler_a
otherwise)
--high-noise-sampling-method (high noise) sampling method, one of [euler, euler_a, heun, dpm2, dpm++2s_a, dpm++2m, dpm++2mv2, ipndm, ipndm_v, lcm,
ddim_trailing, tcd, res_multistep, res_2s, er_sde] default: euler for Flux/SD3/Wan,
euler_a otherwise
--scheduler denoiser sigma scheduler, one of [discrete, karras, exponential, ays, gits, smoothstep, sgm_uniform, simple,
kl_optimal, lcm, bong_tangent], default: discrete
--sigmas custom sigma values for the sampler, comma-separated (e.g., "14.61,7.8,3.5,0.0").
--sampling-method sampling method, one of [euler, euler_a, heun, dpm2, dpm++2s_a, dpm++2m,
dpm++2mv2, ipndm, ipndm_v, lcm, ddim_trailing, tcd, res_multistep, res_2s,
er_sde] (default: euler for Flux/SD3/Wan, euler_a otherwise)
--high-noise-sampling-method (high noise) sampling method, one of [euler, euler_a, heun, dpm2, dpm++2s_a,
dpm++2m, dpm++2mv2, ipndm, ipndm_v, lcm, ddim_trailing, tcd, res_multistep,
res_2s, er_sde] default: euler for Flux/SD3/Wan, euler_a otherwise
--scheduler denoiser sigma scheduler, one of [discrete, karras, exponential, ays, gits,
smoothstep, sgm_uniform, simple, kl_optimal, lcm, bong_tangent], default:
discrete
--sigmas custom sigma values for the sampler, comma-separated (e.g.,
"14.61,7.8,3.5,0.0").
--skip-layers layers to skip for SLG steps (default: [7,8,9])
--high-noise-skip-layers (high noise) layers to skip for SLG steps (default: [7,8,9])
-r, --ref-image reference image for Flux Kontext models (can be used multiple times)
--cache-mode caching method: 'easycache' (DiT), 'ucache' (UNET), 'dbcache'/'taylorseer'/'cache-dit' (DiT block-level), 'spectrum' (UNET/DiT Chebyshev+Taylor forecasting)
--cache-mode caching method: 'easycache' (DiT), 'ucache' (UNET),
'dbcache'/'taylorseer'/'cache-dit' (DiT block-level), 'spectrum' (UNET/DiT
Chebyshev+Taylor forecasting)
--cache-option named cache params (key=value format, comma-separated). easycache/ucache:
threshold=,start=,end=,decay=,relative=,reset=; dbcache/taylorseer/cache-dit: Fn=,Bn=,threshold=,warmup=. Examples:
"threshold=0.25" or "threshold=1.5,reset=0"
--scm-mask SCM steps mask for cache-dit: comma-separated 0/1 (e.g., "1,1,1,0,0,1,0,0,1,0") - 1=compute, 0=can cache
threshold=,start=,end=,decay=,relative=,reset=; dbcache/taylorseer/cache-dit:
Fn=,Bn=,threshold=,warmup=; spectrum: w=,m=,lam=,window=,flex=,warmup=,stop=.
Examples: "threshold=0.25" or "threshold=1.5,reset=0"
--scm-mask SCM steps mask for cache-dit: comma-separated 0/1 (e.g.,
"1,1,1,0,0,1,0,0,1,0") - 1=compute, 0=can cache
--scm-policy SCM policy: 'dynamic' (default) or 'static'
--vae-tile-size tile size for vae tiling, format [X]x[Y] (default: 32x32)
--vae-relative-tile-size relative tile size for vae tiling, format [X]x[Y], in fraction of image size
if < 1, in number of tiles per dim if >=1 (overrides --vae-tile-size)
```

10
examples/server/api.md
View File

@@ -219,7 +219,7 @@ Currently supported request fields:
| `lora` | `array<object>` | Structured LoRA list |
| `extra_images` | `array<string>` | Base64 or data URL images |
| `enable_hr` | `boolean` | Enable highres fix for `txt2img` |
| `hr_upscaler` | `string` | `Latent (nearest)` or an upscaler model name from `/sdapi/v1/upscalers` |
| `hr_upscaler` | `string` | `Lanczos`, `Nearest`, a latent mode such as `Latent (nearest-exact)`, or an upscaler model name from `/sdapi/v1/upscalers` |
| `hr_scale` | `number` | Highres scale when resize target is not set |
| `hr_resize_x` | `integer` | Highres target width, `0` to use scale |
| `hr_resize_y` | `integer` | Highres target height, `0` to use scale |
@@ -303,6 +303,8 @@ Built-in entries include `None`, `Lanczos`, and `Nearest`. Model-backed entries
| --- | --- | --- |
| `[].name` | `string` | WebUI-compatible latent upscale mode name |
Built-in latent modes include `Latent`, `Latent (nearest)`, `Latent (nearest-exact)`, `Latent (antialiased)`, `Latent (bicubic)`, and `Latent (bicubic antialiased)`.
`GET /sdapi/v1/samplers`
| Field | Type | Notes |
@@ -462,7 +464,7 @@ Shared nested fields:
| --- | --- | --- |
| `upscalers[].name` | `string` | Built-in name or model stem; use this value in `hires.upscaler` |
Built-in entries include `None` and `Latent (nearest)`. Model-backed entries are scanned from the top level of `--hires-upscalers-dir`; subdirectories are not scanned.
Built-in entries include `None`, `Lanczos`, `Nearest`, `Latent`, `Latent (nearest)`, `Latent (nearest-exact)`, `Latent (antialiased)`, `Latent (bicubic)`, and `Latent (bicubic antialiased)`. Model-backed entries are scanned from the top level of `--hires-upscalers-dir`; subdirectories are not scanned.
`limits`
@@ -677,7 +679,7 @@ Example:
"lora": [],
"hires": {
"enabled": false,
"upscaler": "Latent (nearest)",
"upscaler": "Latent",
"scale": 2.0,
"target_width": 0,
"target_height": 0,
@@ -804,7 +806,7 @@ Other native fields:
| `scm_mask` | `string` |
| `scm_policy_dynamic` | `boolean` |
For `hires.upscaler`, use `Latent (nearest)` for latent upscale or an `upscalers[].name` value from `GET /sdcpp/v1/capabilities`. Model-backed upscalers are resolved as `--hires-upscalers-dir / (name + ext)` and must live directly in that directory.
For `hires.upscaler`, use `Lanczos`, `Nearest`, `Latent`, `Latent (nearest)`, `Latent (nearest-exact)`, `Latent (antialiased)`, `Latent (bicubic)`, `Latent (bicubic antialiased)`, or an `upscalers[].name` value from `GET /sdcpp/v1/capabilities`. Model-backed upscalers are resolved as `--hires-upscalers-dir / (name + ext)` and must live directly in that directory.
HTTP-only output fields:

7
examples/server/routes_sdapi.cpp
View File

@@ -381,6 +381,8 @@ void register_sdapi_endpoints(httplib::Server& svr, ServerRuntime& rt) {
json result = json::array();
result.push_back(make_builtin("None"));
result.push_back(make_builtin("Lanczos"));
result.push_back(make_builtin("Nearest"));
{
std::lock_guard<std::mutex> lock(*runtime->upscaler_mutex);
@@ -400,7 +402,12 @@ void register_sdapi_endpoints(httplib::Server& svr, ServerRuntime& rt) {
svr.Get("/sdapi/v1/latent-upscale-modes", [](const httplib::Request&, httplib::Response& res) {
json result = json::array({
{{"name", "Latent"}},
{{"name", "Latent (nearest)"}},
{{"name", "Latent (nearest-exact)"}},
{{"name", "Latent (antialiased)"}},
{{"name", "Latent (bicubic)"}},
{{"name", "Latent (bicubic antialiased)"}},
});
res.set_content(result.dump(), "application/json");
});

21
examples/server/routes_sdcpp.cpp
View File

@@ -227,9 +227,30 @@ static json make_capabilities_json(ServerRuntime& runtime) {
available_upscalers.push_back({
{"name", "None"},
});
available_upscalers.push_back({
{"name", "Lanczos"},
});
available_upscalers.push_back({
{"name", "Nearest"},
});
available_upscalers.push_back({
{"name", "Latent"},
});
available_upscalers.push_back({
{"name", "Latent (nearest)"},
});
available_upscalers.push_back({
{"name", "Latent (nearest-exact)"},
});
available_upscalers.push_back({
{"name", "Latent (antialiased)"},
});
available_upscalers.push_back({
{"name", "Latent (bicubic)"},
});
available_upscalers.push_back({
{"name", "Latent (bicubic antialiased)"},
});
{
std::lock_guard<std::mutex> lock(*runtime.upscaler_mutex);
for (const auto& entry : *runtime.upscaler_cache) {