Merge 21df001d53 into 7c81b2f27d

.gitignore

@@ -21,7 +21,7 @@
 *.html
 *.pdf
 *.whl
-cache
+*cache/
 __pycache__/
 storage/
 samples/
@@ -29,9 +29,11 @@ samples/
 !requirements.txt
 .DS_Store
 *DS_Store
+.vscode
 google/
 Wan2.1-T2V-14B/
 Wan2.1-T2V-1.3B/
 Wan2.1-I2V-14B-480P/
 Wan2.1-I2V-14B-720P/
 poetry.lock
+logs/

generate.py

@@ -25,8 +25,12 @@ import wan
 from wan.configs import MAX_AREA_CONFIGS, SIZE_CONFIGS, SUPPORTED_SIZES, WAN_CONFIGS
 from wan.utils.prompt_extend import DashScopePromptExpander, QwenPromptExpander
 from wan.utils.utils import cache_image, cache_video, str2bool
-from wan.utils.platform import get_torch_distributed_backend
-from wan.utils.chrono_inspector import ChronoInspector
+from wan.utils.platform import (
+    get_device_type,
+    get_torch_distributed_backend,
+    get_torch_profiler_activities,
+)
+
 
 
 EXAMPLE_PROMPT = {
@@ -252,6 +256,11 @@ def _parse_args():
         type=float,
         default=5.0,
         help="Classifier free guidance scale.")
+    parser.add_argument(
+        "--profile",
+        action="store_true",
+        default=False,
+        help="profile the generating procedure.")
 
     args = parser.parse_args()
 
@@ -272,6 +281,26 @@ def _init_logging(rank):
         logging.basicConfig(level=logging.ERROR)
 
 
+def _init_profiler():
+    return torch.profiler.profile(
+        activities=get_torch_profiler_activities(),
+        on_trace_ready=torch.profiler.tensorboard_trace_handler('./logs'),
+        record_shapes=True,
+        profile_memory=True,
+        with_stack=True,
+    )
+
+
+def _finalize_profiler(profiler):
+    profiler.stop()
+    table = profiler.key_averages().table(
+        sort_by=f"{get_device_type()}_time_total",
+        row_limit=20,
+    )
+    with open(f"logs/{datetime.now().strftime("%Y-%m-%d_%H-%M-%S")}-profiling.txt", "w") as f:
+        f.write(table)
+
+
 def generate(args):
     rank = int(os.getenv("RANK", 0))
     world_size = int(os.getenv("WORLD_SIZE", 1))
@@ -339,6 +368,10 @@ def generate(args):
         dist.broadcast_object_list(base_seed, src=0)
         args.base_seed = base_seed[0]
     
+    profiler = None
+    if args.profile and rank == 0:
+        profiler = _init_profiler()
+
     if "t2v" in args.task or "t2i" in args.task:
         if args.prompt is None:
             args.prompt = EXAMPLE_PROMPT[args.task]["prompt"]
@@ -366,7 +399,6 @@ def generate(args):
             logging.info(f"Extended prompt: {args.prompt}")
 
         logging.info("Creating WanT2V pipeline.")
-        with ChronoInspector("Creating WanT2V pipeline"):
         wan_t2v = wan.WanT2V(
             config=cfg,
             checkpoint_dir=args.ckpt_dir,
@@ -376,10 +408,11 @@ def generate(args):
             dit_fsdp=args.dit_fsdp,
             use_usp=(args.ulysses_size > 1 or args.ring_size > 1),
             t5_cpu=args.t5_cpu,
+            profiler=profiler,
         )
 
         logging.info("Warming up WanT2V pipeline ...")
-        with ChronoInspector("Warming up WanT2V pipeline"):
+        with torch.no_grad():
             _ = wan_t2v.generate(
                 args.prompt,
                 size=SIZE_CONFIGS[args.size],
@@ -392,7 +425,6 @@ def generate(args):
                 offload_model=args.offload_model)
 
         logging.info(f"Generating {'image' if 't2i' in args.task else 'video'} ...")
-        with ChronoInspector(f"Generating {'image' if 't2i' in args.task else 'video'}"):
         video = wan_t2v.generate(
             args.prompt,
             size=SIZE_CONFIGS[args.size],
@@ -437,7 +469,6 @@ def generate(args):
             logging.info(f"Extended prompt: {args.prompt}")
 
         logging.info("Creating WanI2V pipeline.")
-        with ChronoInspector("Creating WanI2V pipeline"):
         wan_i2v = wan.WanI2V(
             config=cfg,
             checkpoint_dir=args.ckpt_dir,
@@ -447,10 +478,11 @@ def generate(args):
             dit_fsdp=args.dit_fsdp,
             use_usp=(args.ulysses_size > 1 or args.ring_size > 1),
             t5_cpu=args.t5_cpu,
+            profiler=profiler,
         )
 
         logging.info("Warming up WanI2V pipeline ...")
-        with ChronoInspector("Warming up WanI2V pipeline"):
+        with torch.no_grad():
             _ = wan_i2v.generate(
                 args.prompt,
                 img,
@@ -464,7 +496,6 @@ def generate(args):
                 offload_model=args.offload_model)
 
         logging.info("Generating video ...")
-        with ChronoInspector("Generating video"):
         video = wan_i2v.generate(
             args.prompt,
             img,
@@ -511,7 +542,6 @@ def generate(args):
             logging.info(f"Extended prompt: {args.prompt}")
 
         logging.info("Creating WanFLF2V pipeline.")
-        with ChronoInspector("Creating WanFLF2V pipeline"):
         wan_flf2v = wan.WanFLF2V(
             config=cfg,
             checkpoint_dir=args.ckpt_dir,
@@ -521,10 +551,11 @@ def generate(args):
             dit_fsdp=args.dit_fsdp,
             use_usp=(args.ulysses_size > 1 or args.ring_size > 1),
             t5_cpu=args.t5_cpu,
+            profiler=profiler
         )
 
         logging.info("Warming up WanFLF2V pipeline ...")
-        with ChronoInspector("Warming up WanFLF2V pipeline"):
+        with torch.no_grad():
             _ = wan_flf2v.generate(
                 args.prompt,
                 first_frame,
@@ -539,7 +570,6 @@ def generate(args):
                 offload_model=args.offload_model)
 
         logging.info("Generating video ...")
-        with ChronoInspector("Generating video"):
         video = wan_flf2v.generate(
             args.prompt,
             first_frame,
@@ -576,7 +606,6 @@ def generate(args):
             logging.info(f"Extended prompt: {args.prompt}")
 
         logging.info("Creating VACE pipeline.")
-        with ChronoInspector("Creating VACE pipeline"):
         wan_vace = wan.WanVace(
             config=cfg,
             checkpoint_dir=args.ckpt_dir,
@@ -586,6 +615,7 @@ def generate(args):
             dit_fsdp=args.dit_fsdp,
             use_usp=(args.ulysses_size > 1 or args.ring_size > 1),
             t5_cpu=args.t5_cpu,
+            profiler=profiler
         )
 
         src_video, src_mask, src_ref_images = wan_vace.prepare_source(
@@ -595,8 +625,8 @@ def generate(args):
             ], args.frame_num, SIZE_CONFIGS[args.size], device)
 
         logging.info("Warming up VACE pipeline ...")
-        with ChronoInspector("Warming up VACE pipeline"):
-            video = wan_vace.generate(
+        with torch.no_grad():
+            _ = wan_vace.generate(
                 args.prompt,
                 src_video,
                 src_mask,
@@ -611,7 +641,6 @@ def generate(args):
                 offload_model=args.offload_model)
 
         logging.info(f"Generating video...")
-        with ChronoInspector("Generating video"):
         video = wan_vace.generate(
             args.prompt,
             src_video,
@@ -628,6 +657,9 @@ def generate(args):
     else:
         raise ValueError(f"Unkown task type: {args.task}")
 
+    if args.profile and rank == 0:
+        _finalize_profiler(profiler)
+
     if rank == 0:
         if args.save_file is None:
             formatted_time = datetime.now().strftime("%Y%m%d_%H%M%S")

wan/first_last_frame2video.py

@@ -6,6 +6,7 @@ import os
 import random
 import sys
 import types
+from time import perf_counter
 from contextlib import contextmanager
 from functools import partial
 
@@ -54,6 +55,7 @@ class WanFLF2V:
         use_usp=False,
         t5_cpu=False,
         init_on_cpu=True,
+        profiler=None,
     ):
         r"""
         Initializes the image-to-video generation model components.
@@ -143,6 +145,7 @@ class WanFLF2V:
                 self.model.to(self.device)
 
         self.sample_neg_prompt = config.sample_neg_prompt
+        self.profiler = profiler
 
     def generate(self,
                  input_prompt,
@@ -197,6 +200,14 @@ class WanFLF2V:
                 - H: Frame height (from max_area)
                 - W: Frame width from max_area)
         """
+        if self.profiler and self.rank == 0:
+            self.profiler.start()
+
+        start_time = 0.0
+        end_time = 0.0
+        if self.rank == 0:
+            start_time = perf_counter()
+
         first_frame_size = first_frame.size
         last_frame_size = last_frame.size
         first_frame = TF.to_tensor(first_frame).sub_(0.5).div_(0.5).to(
@@ -289,6 +300,10 @@ class WanFLF2V:
         ])[0]
         y = torch.concat([msk, y])
 
+        if self.rank == 0:
+            end_time = perf_counter()
+            logging.info(f"[preprocess and VAE encode] Elapsed time: {end_time - start_time:.2f} seconds")
+
         @contextmanager
         def noop_no_sync():
             yield
@@ -339,8 +354,14 @@ class WanFLF2V:
             if offload_model:
                 empty_cache()
 
+            if self.rank == 0:
+                start_time = perf_counter()
+
             self.model.to(self.device)
             for _, t in enumerate(tqdm(timesteps)):
+                if self.profiler and self.rank == 0:
+                    self.profiler.step()
+
                 latent_model_input = [latent.to(self.device)]
                 timestep = [t]
 
@@ -370,6 +391,10 @@ class WanFLF2V:
                     generator=seed_g)[0]
                 latent = temp_x0.squeeze(0)
 
+            if self.rank == 0:
+                end_time = perf_counter()
+                logging.info(f"[sampling time steps] Elapsed time: {end_time - start_time:.2f} seconds")
+
             x0 = [latent.to(self.device)]
             del latent_model_input, timestep
 
@@ -378,7 +403,10 @@ class WanFLF2V:
                 empty_cache()
 
             if self.rank == 0:
+                start_time = perf_counter()
                 videos = self.vae.decode(x0)
+                end_time = perf_counter()
+                logging.info(f"[VAE decoding] Elapsed time: {end_time - start_time:.2f} seconds")
 
         del noise, latent
         del sample_scheduler

wan/image2video.py

@@ -6,6 +6,7 @@ import os
 import random
 import sys
 import types
+from time import perf_counter
 from contextlib import contextmanager
 from functools import partial
 
@@ -54,6 +55,7 @@ class WanI2V:
         use_usp=False,
         t5_cpu=False,
         init_on_cpu=True,
+        profiler=None,
     ):
         r"""
         Initializes the image-to-video generation model components.
@@ -143,6 +145,7 @@ class WanI2V:
                 self.model.to(self.device)
 
         self.sample_neg_prompt = config.sample_neg_prompt
+        self.profiler = profiler
 
     def generate(self,
                  input_prompt,
@@ -192,6 +195,14 @@ class WanI2V:
                 - H: Frame height (from max_area)
                 - W: Frame width from max_area)
         """
+        if self.profiler and self.rank == 0:
+            self.profiler.start()
+
+        start_time = 0.0
+        end_time = 0.0
+        if self.rank == 0:
+            start_time = perf_counter()
+
         img = TF.to_tensor(img).sub_(0.5).div_(0.5).to(self.device)
 
         F = frame_num
@@ -262,6 +273,10 @@ class WanI2V:
         ])[0]
         y = torch.concat([msk, y])
 
+        if self.rank == 0:
+            end_time = perf_counter()
+            logging.info(f"[preprocess and VAE encode] Elapsed time: {end_time - start_time:.2f} seconds")
+
         @contextmanager
         def noop_no_sync():
             yield
@@ -312,8 +327,14 @@ class WanI2V:
             if offload_model:
                 empty_cache()
 
+            if self.rank == 0:
+                start_time = perf_counter()
+
             self.model.to(self.device)
             for _, t in enumerate(tqdm(timesteps)):
+                if self.profiler and self.rank == 0:
+                    self.profiler.step()
+
                 latent_model_input = [latent.to(self.device)]
                 timestep = [t]
 
@@ -343,6 +364,10 @@ class WanI2V:
                     generator=seed_g)[0]
                 latent = temp_x0.squeeze(0)
 
+            if self.rank == 0:
+                end_time = perf_counter()
+                logging.info(f"[sampling time steps] Elapsed time: {end_time - start_time:.2f} seconds")
+
             x0 = [latent.to(self.device)]
             del latent_model_input, timestep
 
@@ -351,7 +376,10 @@ class WanI2V:
                 empty_cache()
 
             if self.rank == 0:
+                start_time = perf_counter()
                 videos = self.vae.decode(x0)
+                end_time = perf_counter()
+                logging.info(f"[VAE decoding] Elapsed time: {end_time - start_time:.2f} seconds")
 
         del noise, latent
         del sample_scheduler

wan/modules/model.py

@@ -16,7 +16,7 @@ try:
     from wan.modules.attention import attention as flash_attention
     torch.backends.mudnn.allow_tf32 = True
 except ModuleNotFoundError:
-    pass
+    torch_musa = None
 
 __all__ = ['WanModel']
 

wan/text2video.py

@@ -6,6 +6,7 @@ import os
 import random
 import sys
 import types
+from time import perf_counter
 from contextlib import contextmanager
 from functools import partial
 
@@ -50,6 +51,7 @@ class WanT2V:
         dit_fsdp=False,
         use_usp=False,
         t5_cpu=False,
+        profiler=None,
     ):
         r"""
         Initializes the Wan text-to-video generation model components.
@@ -124,6 +126,7 @@ class WanT2V:
             self.model.to(self.device)
 
         self.sample_neg_prompt = config.sample_neg_prompt
+        self.profiler = profiler
 
     def generate(self,
                  input_prompt,
@@ -169,6 +172,14 @@ class WanT2V:
                 - H: Frame height (from size)
                 - W: Frame width from size)
         """
+        if self.profiler and self.rank == 0:
+            self.profiler.start()
+
+        start_time = 0.0
+        end_time = 0.0
+        if self.rank == 0:
+            start_time = perf_counter()
+
         # preprocess
         F = frame_num
         target_shape = (self.vae.model.z_dim, (F - 1) // self.vae_stride[0] + 1,
@@ -208,6 +219,10 @@ class WanT2V:
                 generator=seed_g)
         ]
 
+        if self.rank == 0:
+            end_time = perf_counter()
+            logging.info(f"[preprocess] Elapsed time: {end_time - start_time:.2f} seconds")
+
         @contextmanager
         def noop_no_sync():
             yield
@@ -244,8 +259,14 @@ class WanT2V:
             arg_c = {'context': context, 'seq_len': seq_len}
             arg_null = {'context': context_null, 'seq_len': seq_len}
 
+            if self.rank == 0:
+                start_time = perf_counter()
+
             self.model.to(self.device)
             for _, t in enumerate(tqdm(timesteps)):
+                if self.profiler and self.rank == 0:
+                    self.profiler.step()
+
                 latent_model_input = latents
                 timestep = [t]
 
@@ -267,12 +288,19 @@ class WanT2V:
                     generator=seed_g)[0]
                 latents = [temp_x0.squeeze(0)]
 
+            if self.rank == 0:
+                end_time = perf_counter()
+                logging.info(f"[sampling time steps] Elapsed time: {end_time - start_time:.2f} seconds")
+
             x0 = latents
             if offload_model:
                 self.model.cpu()
                 empty_cache()
             if self.rank == 0:
+                start_time = perf_counter()
                 videos = self.vae.decode(x0)
+                end_time = perf_counter()
+                logging.info(f"[VAE decoding] Elapsed time: {end_time - start_time:.2f} seconds")
 
         del noise, latents
         del sample_scheduler

wan/utils/platform.py

@@ -1,4 +1,4 @@
-from typing import Optional
+from typing import Optional, List
 
 import torch
 
@@ -8,7 +8,7 @@ except ModuleNotFoundError:
     torch_musa = None
 
 
-def _is_musa():
+def _is_musa() -> bool:
     if torch_musa is None:
         return False
     else:
@@ -48,3 +48,14 @@ def get_torch_distributed_backend() -> str:
         return "mccl"
     else:
         raise NotImplementedError("No Accelerators(NV/MTT GPU) available")
+
+
+def get_torch_profiler_activities() -> List[torch.profiler.ProfilerActivity]:
+    activities: List[torch.profiler.ProfilerActivity] = [
+        torch.profiler.ProfilerActivity.CPU
+    ]
+    if torch.cuda.is_available():
+        activities.append(torch.profiler.ProfilerActivity.CUDA)
+    elif _is_musa():
+        activities.append(torch.profiler.ProfilerActivity.MUSA)
+    return activities

wan/vace.py

@@ -8,6 +8,7 @@ import sys
 import time
 import traceback
 import types
+from time import perf_counter
 from contextlib import contextmanager
 from functools import partial
 
@@ -58,6 +59,7 @@ class WanVace(WanT2V):
         dit_fsdp=False,
         use_usp=False,
         t5_cpu=False,
+        profiler=None,
     ):
         r"""
         Initializes the Wan text-to-video generation model components.
@@ -150,6 +152,8 @@ class WanVace(WanT2V):
             seq_len=75600,
             keep_last=True)
 
+        self.profiler = profiler
+
     def vace_encode_frames(self, frames, ref_images, masks=None, vae=None):
         vae = self.vae if vae is None else vae
         if ref_images is None:
@@ -354,6 +358,14 @@ class WanVace(WanT2V):
                 - H: Frame height (from size)
                 - W: Frame width from size)
         """
+        if self.profiler and self.rank == 0:
+            self.profiler.start()
+
+        start_time = 0.0
+        end_time = 0.0
+        if self.rank == 0:
+            start_time = perf_counter()
+
         # preprocess
         # F = frame_num
         # target_shape = (self.vae.model.z_dim, (F - 1) // self.vae_stride[0] + 1,
@@ -404,6 +416,10 @@ class WanVace(WanT2V):
                             (self.patch_size[1] * self.patch_size[2]) *
                             target_shape[1] / self.sp_size) * self.sp_size
 
+        if self.rank == 0:
+            end_time = perf_counter()
+            logging.info(f"[preprocess] Elapsed time: {end_time - start_time:.2f} seconds")
+
         @contextmanager
         def noop_no_sync():
             yield
@@ -440,13 +456,19 @@ class WanVace(WanT2V):
             arg_c = {'context': context, 'seq_len': seq_len}
             arg_null = {'context': context_null, 'seq_len': seq_len}
 
+            if self.rank == 0:
+                start_time = perf_counter()
+
+            self.model.to(self.device)
             for _, t in enumerate(tqdm(timesteps)):
+                if self.profiler and self.rank == 0:
+                    self.profiler.step()
+
                 latent_model_input = latents
                 timestep = [t]
 
                 timestep = torch.stack(timestep)
 
-                self.model.to(self.device)
                 noise_pred_cond = self.model(
                     latent_model_input,
                     t=timestep,
@@ -471,12 +493,19 @@ class WanVace(WanT2V):
                     generator=seed_g)[0]
                 latents = [temp_x0.squeeze(0)]
 
+            if self.rank == 0:
+                end_time = perf_counter()
+                logging.info(f"[sampling time steps] Elapsed time: {end_time - start_time:.2f} seconds")
+
             x0 = latents
             if offload_model:
                 self.model.cpu()
                 empty_cache()
             if self.rank == 0:
+                start_time = perf_counter()
                 videos = self.decode_latent(x0, input_ref_images)
+                end_time = perf_counter()
+                logging.info(f"[VAE decoding] Elapsed time: {end_time - start_time:.2f} seconds")
 
         del noise, latents
         del sample_scheduler