update unipc

automatic.sh

@@ -1,12 +1,14 @@
 #!/bin/env bash
 
 export TF_CPP_MIN_LOG_LEVEL=2
+export ACCELERATE="True"
 export FORCE_CUDA="1"
 export ATTN_PRECISION=fp16
 export PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.9,max_split_size_mb:512
 export CUDA_LAUNCH_BLOCKING=0
 export CUDA_CACHE_DISABLE=0
 export CUDA_AUTO_BOOST=1
+export CUDA_MODULE_LOADING="LAZY"
 export CUDA_DEVICE_DEFAULT_PERSISTING_L2_CACHE_PERCENTAGE_LIMIT=0
 
 if [ "$PYTHON" == "" ]; then
@@ -151,6 +153,7 @@ if [ $MODE == optimized ]; then
 fi
 
 exec accelerate launch --no_python --quiet --num_cpu_threads_per_process=6 "$PYTHON" $CMD
+# exec "$PYTHON" $CMD
 
 # export LD_PRELOAD=libtcmalloc.so
 # TORCH_CUDA_ARCH_LIST="8.6"

cli/random/dynamotest.py

@@ -94,10 +94,4 @@ Inference-only backends:
     dynamo.optimize("ofi") - Uses Torchscript optimize_for_inference
     dynamo.optimize("fx2trt") - Uses Nvidia TensorRT for inference optimizations
     dynamo.optimize("onnxrt") - Uses ONNXRT for inference on CPU/GPU
-Results:
-    "default": 4.247040033340454,
-    "ofi": 3.820032000541687,
-    "aot_cudagraphs": 6.460927963256836,
-    "inductor": RuntimeError: CUDA: Error- no device
-    "fx2trt": ImportError: libtorch_cuda_cu.so: cannot open shared object file: No such file or directory
 """

config.json

@@ -132,7 +132,7 @@
     "localization": "None",
     "lora_apply_to_outputs": false,
     "memmon_poll_rate": 1,
-    "multiple_tqdm": true,
+    "multiple_tqdm": false,
     "n_rows": -1,
     "no_dpmpp_sde_batch_determinism": false,
     "outdir_extras_samples": "outputs/extras",

modules/models/diffusion/uni_pc/uni_pc.py

@@ -1,6 +1,7 @@
 import torch
 import torch.nn.functional as F
 import math
+from tqdm.auto import trange
 
 
 class NoiseScheduleVP:
@@ -766,7 +767,7 @@ class UniPC:
                         self.after_update(x, model_x)
                     model_prev_list.append(model_x)
                     t_prev_list.append(vec_t)
-                for step in range(order, steps + 1):
+                for step in trange(order, steps + 1):
                     vec_t = timesteps[step].expand(x.shape[0])
                     if lower_order_final:
                         step_order = min(order, steps + 1 - step)

ui-config.json

@@ -1032,7 +1032,7 @@
     "customscript/save_steps_animation.py/img2img/Skip steps/maximum": 100,
     "customscript/save_steps_animation.py/img2img/Skip steps/minimum": 0,
     "customscript/save_steps_animation.py/img2img/Skip steps/step": 1,
-    "customscript/save_steps_animation.py/img2img/Skip steps/value": 5,
+    "customscript/save_steps_animation.py/img2img/Skip steps/value": 0,
     "customscript/save_steps_animation.py/img2img/Skip steps/visible": true,
     "customscript/save_steps_animation.py/txt2img/Codec/value": "x264",
     "customscript/save_steps_animation.py/txt2img/Codec/visible": true,

webui.py

@@ -289,7 +289,10 @@ def webui():
 
         print(f"Startup time: {startup_timer.summary()}.")
 
-        wait_on_server(shared.demo)
+        try:
+            wait_on_server(shared.demo)
+        except KeyboardInterrupt as e:
+            pass
         print('Restarting UI...')
 
         startup_timer.reset()