Zengtudor/Wan2.1
1
0
Fork 0
mirror of https://github.com/Wan-Video/Wan2.1.git synced 2025-11-04 22:26:36 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

set default model loading to JIT

Browse Source
This commit is contained in:
Chris Malone 2025-03-24 23:38:12 +11:00
parent 1b125fe233
commit e686952a84
1 changed files with 6 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
gradio_server.py
View File

@ -492,7 +492,7 @@ if not Path(server_config_filename).is_file():
"boost" : 1, "boost" : 1,
"vae_config": 0, "vae_config": 0,
"profile" : profile_type.LowRAM_LowVRAM, "profile" : profile_type.LowRAM_LowVRAM,
"reload_model": 1 } "reload_model": 2 }
with open(server_config_filename, "w", encoding="utf-8") as writer: with open(server_config_filename, "w", encoding="utf-8") as writer:
writer.write(json.dumps(server_config)) writer.write(json.dumps(server_config))
@ -854,8 +854,8 @@ def load_models(i2v):
kwargs = { "extraModelsToQuantize": None} kwargs = { "extraModelsToQuantize": None}
if profile == 2 or profile == 4: if profile == 2 or profile == 4:
kwargs["budgets"] = { "transformer" : 100 if preload == 0 else preload, "text_encoder" : 100, "*" : 1000 } kwargs["budgets"] = { "transformer" : 100 if preload == 0 else preload, "text_encoder" : 100, "*" : 1000 }
if profile == 4: # if profile == 4:
kwargs["partialPinning"] = True # kwargs["partialPinning"] = True
elif profile == 3: elif profile == 3:
kwargs["budgets"] = { "*" : "70%" } kwargs["budgets"] = { "*" : "70%" }
offloadobj = offload.profile(pipe, profile_no= profile, compile = compile, quantizeTransformer = quantizeTransformer, loras = "transformer", **kwargs) offloadobj = offload.profile(pipe, profile_no= profile, compile = compile, quantizeTransformer = quantizeTransformer, loras = "transformer", **kwargs)
@ -867,6 +867,7 @@ if check_loras:
setup_loras(use_image2video, transformer, get_lora_dir(use_image2video), "", None) setup_loras(use_image2video, transformer, get_lora_dir(use_image2video), "", None)
exit() exit()
del transformer del transformer
gen_in_progress = False gen_in_progress = False
def get_auto_attention(): def get_auto_attention():
@ -2247,7 +2248,7 @@ def generate_configuration_tab():
("When changing tabs", 1), ("When changing tabs", 1),
("When pressing generate", 2), ("When pressing generate", 2),
], ],
value=server_config.get("reload_model",1), value=server_config.get("reload_model",2),
label="Reload model" label="Reload model"
) )
msg = gr.Markdown() msg = gr.Markdown()
@ -2289,7 +2290,7 @@ def on_tab_select(t2v_state, i2v_state, evt: gr.SelectData):
new_i2v = evt.index == 1 new_i2v = evt.index == 1
use_image2video = new_i2v use_image2video = new_i2v
if(server_config.get("reload_model",1) == 1): if(server_config.get("reload_model",2) == 1):
global wan_model, offloadobj global wan_model, offloadobj
if wan_model is not None: if wan_model is not None:
if offloadobj is not None: if offloadobj is not None: