Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-18 16:54:33 +02:00
Code Issues Packages Projects Releases Wiki Activity

new server info panel

Browse Source 
Signed-off-by: vladmandic <mandic00@live.com>
This commit is contained in:
vladmandic
2026-03-18 15:37:57 +01:00
parent c187aa706a
commit fb2f9ea650
11 changed files with 139 additions and 43 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/api/api.py
+4 -1
View File
@@ -39,12 +39,14 @@ class Api:
def register(self):
# fetch js/css
self.add_api_route("/js", server.get_js, methods=["GET"], auth=False)
# server api
self.add_api_route("/sdapi/v1/motd", server.get_motd, methods=["GET"], response_model=str)
self.add_api_route("/sdapi/v1/log", server.get_log, methods=["GET"], response_model=list[str])
self.add_api_route("/sdapi/v1/log", server.post_log, methods=["POST"])
self.add_api_route("/sdapi/v1/start", self.get_session_start, methods=["GET"])
self.add_api_route("/sdapi/v1/version", server.get_version, methods=["GET"])
self.add_api_route("/sdapi/v1/torch", server.get_torch, methods=["GET"])
self.add_api_route("/sdapi/v1/status", server.get_status, methods=["GET"], response_model=models.ResStatus)
self.add_api_route("/sdapi/v1/platform", server.get_platform, methods=["GET"])
self.add_api_route("/sdapi/v1/progress", server.get_progress, methods=["GET"], response_model=models.ResProgress)
@@ -54,7 +56,8 @@ class Api:
self.add_api_route("/sdapi/v1/shutdown", server.post_shutdown, methods=["POST"])
self.add_api_route("/sdapi/v1/memory", server.get_memory, methods=["GET"], response_model=models.ResMemory)
self.add_api_route("/sdapi/v1/cmd-flags", server.get_cmd_flags, methods=["GET"], response_model=models.FlagsModel)
self.add_api_route("/sdapi/v1/gpu", gpu.get_gpu_status, methods=["GET"], response_model=list[models.ResGPU])
self.add_api_route("/sdapi/v1/gpu", gpu.get_gpu, methods=["GET"])
self.add_api_route("/sdapi/v1/gpu-smi", gpu.get_gpu_smi, methods=["GET"], response_model=list[models.ResGPU])
# core api using locking
self.add_api_route("/sdapi/v1/txt2img", self.generate.post_text2img, methods=["POST"], response_model=models.ResTxt2Img, tags=["Generation"])
modules/api/gpu.py
+12 -2
View File
@@ -5,7 +5,17 @@ from modules.logger import log
device = None
def get_gpu_status():
def get_gpu():
import installer
res = {}
if len(installer.gpu_info) == 1:
return installer.gpu_info[0]
for i, item in enumerate(installer.gpu_info):
res[i] = item
return res
def get_gpu_smi():
"""Return real-time GPU metrics (utilization, temperature, memory, clock speeds) via vendor-specific APIs (NVML, ROCm SMI, XPU SMI)."""
global device # pylint: disable=global-statement
if device is None:
@@ -37,5 +47,5 @@ class ResGPU(BaseModel):
if __name__ == '__main__':
from rich import print as rprint
for gpu in get_gpu_status():
for gpu in get_gpu_smi():
rprint(gpu)
modules/api/middleware.py
+1 -1
View File
@@ -19,7 +19,7 @@ ignore_endpoints = [
'/sdapi/v1/version',
'/sdapi/v1/log',
'/sdapi/v1/browser',
'/sdapi/v1/gpu',
'/sdapi/v1/gpu-smi',
'/sdapi/v1/network/thumb',
'/sdapi/v1/progress',
]
modules/api/server.py
+21 -22
View File
@@ -8,14 +8,6 @@ from modules import shared
from modules.logger import log
from modules.api import models, helpers
def _get_version():
return installer.get_version()
def post_shutdown():
log.info('Shutdown request received')
import sys
sys.exit(0)
def get_js(request: Request):
file = request.query_params.get("file", None)
@@ -43,32 +35,34 @@ def get_js(request: Request):
media_type = 'application/octet-stream'
return FileResponse(file, media_type=media_type)
def get_version():
return installer.get_version()
def get_motd():
import requests
motd = ''
ver = _get_version()
if ver.get('updated', None) is not None:
motd = f"version <b>{ver['commit']} {ver['updated']}</b> <span style='color: var(--primary-500)'>{ver['url'].split('/')[-1]}</span><br>" # pylint: disable=use-maxsplit-arg
motd = ""
ver = get_version()
if ver.get("updated", None) is not None:
motd = f"version <b>{ver['commit']} {ver['updated']}</b> <span style='color: var(--primary-500)'>{ver['url'].split('/')[-1]}</span><br>" # pylint: disable=use-maxsplit-arg
if shared.opts.motd:
try:
res = requests.get('https://vladmandic.github.io/sdnext/motd', timeout=3)
res = requests.get("https://vladmandic.github.io/sdnext/motd", timeout=3)
if res.status_code == 200:
msg = (res.text or '').strip()
log.info(f'MOTD: {msg if len(msg) > 0 else "N/A"}')
msg = (res.text or "").strip()
log.info(f"MOTD: {msg if len(msg) > 0 else 'N/A'}")
motd += res.text
else:
log.error(f'MOTD: {res.status_code}')
log.error(f"MOTD: {res.status_code}")
except Exception as err:
log.error(f'MOTD: {err}')
log.error(f"MOTD: {err}")
return motd
def get_version():
return _get_version()
def get_platform():
from installer import get_platform as installer_get_platform
from modules.loader import get_packages as loader_get_packages
return { **installer_get_platform(), **loader_get_packages() }
return { **installer.get_platform(), **loader_get_packages() }
def get_torch():
return dict(installer.torch_info)
def get_log(req: models.ReqGetLog = Depends()):
lines = log.buffer[:req.lines] if req.lines > 0 else log.buffer.copy()
@@ -85,6 +79,11 @@ def post_log(req: models.ReqPostLog):
log.error(f'UI: {req.error}')
return {}
def post_shutdown():
log.info("Shutdown request received")
import sys
sys.exit(0)
def get_cmd_flags():
return vars(shared.cmd_opts)
modules/attention.py
+14 -4
View File
@@ -2,7 +2,7 @@ from functools import wraps
import torch
from modules import rocm
from modules.errors import log
from installer import install, installed
from installer import install, installed, torch_info
def set_dynamic_attention():
@@ -10,6 +10,7 @@ def set_dynamic_attention():
sdpa_pre_dyanmic_atten = torch.nn.functional.scaled_dot_product_attention
from modules.sd_hijack_dynamic_atten import dynamic_scaled_dot_product_attention
torch.nn.functional.scaled_dot_product_attention = dynamic_scaled_dot_product_attention
torch_info.set(attention='dynamic')
return sdpa_pre_dyanmic_atten
except Exception as err:
log.error(f'Torch attention: type="dynamic attention" {err}')
@@ -20,6 +21,7 @@ def set_triton_flash_attention(backend: str):
try:
if backend in {"rocm", "zluda"}: # flash_attn_triton_amd only works with AMD
from modules.flash_attn_triton_amd import interface_fa
sdpa_pre_triton_flash_atten = torch.nn.functional.scaled_dot_product_attention
@wraps(sdpa_pre_triton_flash_atten)
def sdpa_triton_flash_atten(query: torch.FloatTensor, key: torch.FloatTensor, value: torch.FloatTensor, attn_mask: torch.Tensor | None = None, dropout_p: float = 0.0, is_causal: bool = False, scale: float | None = None, enable_gqa: bool = False, **kwargs) -> torch.FloatTensor:
@@ -42,6 +44,7 @@ def set_triton_flash_attention(backend: str):
kwargs["enable_gqa"] = enable_gqa
return sdpa_pre_triton_flash_atten(query=query, key=key, value=value, attn_mask=attn_mask, dropout_p=dropout_p, is_causal=is_causal, scale=scale, **kwargs)
torch.nn.functional.scaled_dot_product_attention = sdpa_triton_flash_atten
torch_info.set(attention='triton')
log.debug('Torch attention: type="Triton Flash attention"')
except Exception as err:
log.error(f'Torch attention: type="Triton Flash attention" {err}')
@@ -78,6 +81,7 @@ def set_flex_attention():
return flex_attention(query, key, value, score_mod=score_mod, block_mask=block_mask, scale=scale, enable_gqa=enable_gqa)
torch.nn.functional.scaled_dot_product_attention = sdpa_flex_atten
torch_info.set(attention="flex")
log.debug('Torch attention: type="Flex attention"')
except Exception as err:
log.error(f'Torch attention: type="Flex attention" {err}')
@@ -93,6 +97,7 @@ def set_ck_flash_attention(backend: str, device: torch.device):
else:
install('flash-attn')
from flash_attn import flash_attn_func
sdpa_pre_flash_atten = torch.nn.functional.scaled_dot_product_attention
@wraps(sdpa_pre_flash_atten)
def sdpa_flash_atten(query: torch.FloatTensor, key: torch.FloatTensor, value: torch.FloatTensor, attn_mask: torch.Tensor | None = None, dropout_p: float = 0.0, is_causal: bool = False, scale: float | None = None, enable_gqa: bool = False, **kwargs) -> torch.FloatTensor:
@@ -120,6 +125,7 @@ def set_ck_flash_attention(backend: str, device: torch.device):
kwargs["enable_gqa"] = enable_gqa
return sdpa_pre_flash_atten(query=query, key=key, value=value, attn_mask=attn_mask, dropout_p=dropout_p, is_causal=is_causal, scale=scale, **kwargs)
torch.nn.functional.scaled_dot_product_attention = sdpa_flash_atten
torch_info.set(attention="flash")
log.debug('Torch attention: type="Flash attention"')
except Exception as err:
log.error(f'Torch attention: type="Flash attention" {err}')
@@ -174,6 +180,7 @@ def set_sage_attention(backend: str, device: torch.device):
kwargs["enable_gqa"] = enable_gqa
return sdpa_pre_sage_atten(query=query, key=key, value=value, attn_mask=attn_mask, dropout_p=dropout_p, is_causal=is_causal, scale=scale, **kwargs)
torch.nn.functional.scaled_dot_product_attention = sdpa_sage_atten
torch_info.set(attention="sage")
log.debug(f'Torch attention: type="Sage attention" backend={"cuda" if use_cuda_backend else "auto"}')
except Exception as err:
log.error(f'Torch attention: type="Sage attention" {err}')
@@ -208,19 +215,22 @@ def set_diffusers_attention(pipe, quiet:bool=False):
log.quiet(quiet, f'Setting model: attention="{shared.opts.cross_attention_optimization}"')
if shared.opts.cross_attention_optimization == "Disabled":
pass # do nothing
elif shared.opts.cross_attention_optimization == "Scaled-Dot-Product": # The default set by Diffusers
torch_info.set(attention="disabled")
elif shared.opts.cross_attention_optimization == "Scaled-Dot-Product": # The default set by Diffusers
torch_info.set(attention="sdpa")
# set_attn(pipe, p.AttnProcessor2_0(), name="Scaled-Dot-Product")
pass
elif shared.opts.cross_attention_optimization == "xFormers":
if hasattr(pipe, 'enable_xformers_memory_efficient_attention'):
torch_info.set(attention="xformers")
pipe.enable_xformers_memory_efficient_attention()
else:
log.warning(f"Attention: xFormers is not compatible with {pipe.__class__.__name__}")
elif shared.opts.cross_attention_optimization == "Batch matrix-matrix":
torch_info.set(attention="bmm")
set_attn(pipe, p.AttnProcessor(), name="Batch matrix-matrix")
elif shared.opts.cross_attention_optimization == "Dynamic Attention BMM":
from modules.sd_hijack_dynamic_atten import DynamicAttnProcessorBMM
torch_info.set(attention="dynamic_bmm")
set_attn(pipe, DynamicAttnProcessorBMM(), name="Dynamic Attention BMM")
if shared.opts.attention_slicing != "Default" and hasattr(pipe, "enable_attention_slicing") and hasattr(pipe, "disable_attention_slicing"):
modules/devices.py
+27 -2
View File
@@ -4,8 +4,10 @@ import time
import contextlib
import importlib.metadata
import torch
from installer import torch_info
from modules.logger import log
from modules import rocm, attention
from modules.errors import log, display, install as install_traceback
from modules.errors import display, install as install_traceback
debug = os.environ.get('SD_DEVICE_DEBUG', None) is not None
@@ -398,17 +400,35 @@ def test_triton(early: bool = False):
test_triton_func(torch.randn(16, device=device), torch.randn(16, device=device), torch.randn(16, device=device))
triton_ok = True
else:
torch_info.set(triton=False)
triton_ok = False
except Exception as e:
torch_info.set(triton=False)
triton_ok = False
line = str(e).splitlines()[0]
log.warning(f"Triton test fail: {line}")
if debug:
from modules import errors
errors.display(e, 'Triton')
triton_version = False
if triton_ok:
if triton_version is None:
try:
import torch._inductor.triton as torch_triton
triton_version = torch_triton.__version__
except Exception:
pass
if triton_version is None:
try:
import triton
triton_version = triton.__version__
except Exception:
pass
torch_info.set(triton=triton_version)
t1 = time.time()
fn = f'{sys._getframe(2).f_code.co_name}:{sys._getframe(1).f_code.co_name}' # pylint: disable=protected-access
log.debug(f'Triton: pass={triton_ok} fn={fn} time={t1-t0:.2f}')
log.debug(f'Triton: pass={triton_ok} version={triton_version} fn={fn} time={t1-t0:.2f}')
if not triton_ok and opts is not None:
opts.sdnq_dequantize_compile = False
return triton_ok
@@ -469,6 +489,7 @@ def set_sdpa_params():
torch.backends.cuda.enable_math_sdp('Math' in opts.sdp_options or 'Math attention' in opts.sdp_options)
if hasattr(torch.backends.cuda, "allow_fp16_bf16_reduction_math_sdp"): # only valid for torch >= 2.5
torch.backends.cuda.allow_fp16_bf16_reduction_math_sdp(True)
torch_info.set(attention="sdpa")
log.debug(f'Torch attention: type="sdpa" kernels={opts.sdp_options} overrides={opts.sdp_overrides}')
except Exception as err:
log.warning(f'Torch attention: type="sdpa" {err}')
@@ -559,6 +580,10 @@ def set_dtype():
inference_context = contextlib.nullcontext
else:
inference_context = torch.no_grad
if dtype == dtype_vae:
torch_info.set(dtype=str(dtype))
else:
torch_info.set(dtype=str(dtype), vae=str(dtype_vae))
def set_cuda_params():