add intermediate logging for dreambooth training script

examples/dreambooth/train_dreambooth.py

@@ -24,6 +24,7 @@
 from typing import Optional
 
 import accelerate
+import numpy as np
 import torch
 import torch.nn.functional as F
 import torch.utils.checkpoint
@@ -40,18 +41,71 @@
 from transformers import AutoTokenizer, PretrainedConfig
 
 import diffusers
-from diffusers import AutoencoderKL, DDPMScheduler, DiffusionPipeline, UNet2DConditionModel
+from diffusers import (
+    AutoencoderKL,
+    DDPMScheduler,
+    DiffusionPipeline,
+    DPMSolverMultistepScheduler,
+    UNet2DConditionModel,
+)
 from diffusers.optimization import get_scheduler
-from diffusers.utils import check_min_version
+from diffusers.utils import check_min_version, is_wandb_available
 from diffusers.utils.import_utils import is_xformers_available
 
 
+if is_wandb_available():
+    import wandb
+
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
 check_min_version("0.15.0.dev0")
 
 logger = get_logger(__name__)
 
 
+def log_validation(text_encoder, tokenizer, unet, vae, args, accelerator, weight_dtype, epoch):
+    logger.info(
+        f"Running validation... \n Generating {args.num_validation_images} images with prompt:"
+        f" {args.validation_prompt}."
+    )
+    # create pipeline (note: unet and vae are loaded again in float32)
+    pipeline = DiffusionPipeline.from_pretrained(
+        args.pretrained_model_name_or_path,
+        text_encoder=accelerator.unwrap_model(text_encoder),
+        tokenizer=tokenizer,
+        unet=accelerator.unwrap_model(unet),
+        vae=vae,
+        revision=args.revision,
+        torch_dtype=weight_dtype,
+    )
+    pipeline.scheduler = DPMSolverMultistepScheduler.from_config(pipeline.scheduler.config)
+    pipeline = pipeline.to(accelerator.device)
+    pipeline.set_progress_bar_config(disable=True)
+
+    # run inference
+    generator = None if args.seed is None else torch.Generator(device=accelerator.device).manual_seed(args.seed)
+    images = []
+    for _ in range(args.num_validation_images):
+        with torch.autocast("cuda"):
+            image = pipeline(args.validation_prompt, num_inference_steps=25, generator=generator).images[0]
+        images.append(image)
+
+    for tracker in accelerator.trackers:
+        if tracker.name == "tensorboard":
+            np_images = np.stack([np.asarray(img) for img in images])
+            tracker.writer.add_images("validation", np_images, epoch, dataformats="NHWC")
+        if tracker.name == "wandb":
+            tracker.log(
+                {
+                    "validation": [
+                        wandb.Image(image, caption=f"{i}: {args.validation_prompt}") for i, image in enumerate(images)
+                    ]
+                }
+            )
+
+    del pipeline
+    torch.cuda.empty_cache()
+
+
 def import_model_class_from_model_name_or_path(pretrained_model_name_or_path: str, revision: str):
     text_encoder_config = PretrainedConfig.from_pretrained(
         pretrained_model_name_or_path,
@@ -306,6 +360,28 @@ def parse_args(input_args=None):
             ' (default), `"wandb"` and `"comet_ml"`. Use `"all"` to report to all integrations.'
         ),
     )
+    parser.add_argument(
+        "--validation_prompt",
+        type=str,
+        default=None,
+        help="A prompt that is used during validation to verify that the model is learning.",
+    )
+    parser.add_argument(
+        "--num_validation_images",
+        type=int,
+        default=4,
+        help="Number of images that should be generated during validation with `validation_prompt`.",
+    )
+    parser.add_argument(
+        "--validation_steps",
+        type=int,
+        default=100,
+        help=(
+            "Run validation every X steps. Validation consists of running the prompt"
+            " `args.validation_prompt` multiple times: `args.num_validation_images`"
+            " and logging the images."
+        ),
+    )
     parser.add_argument(
         "--mixed_precision",
         type=str,
@@ -508,6 +584,10 @@ def main(args):
         project_config=accelerator_project_config,
     )
 
+    if args.report_to == "wandb":
+        if not is_wandb_available():
+            raise ImportError("Make sure to install wandb if you want to use it for logging during training.")
+
     # Currently, it's not possible to do gradient accumulation when training two models with accelerate.accumulate
     # This will be enabled soon in accelerate. For now, we don't allow gradient accumulation when training two models.
     # TODO (patil-suraj): Remove this check when gradient accumulation with two models is enabled in accelerate.
@@ -920,6 +1000,8 @@ def load_model_hook(models, input_dir):
                         save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
                         accelerator.save_state(save_path)
                         logger.info(f"Saved state to {save_path}")
+                if args.validation_prompt is not None and global_step % args.validation_steps == 0:
+                    log_validation(text_encoder, tokenizer, unet, vae, args, accelerator, weight_dtype, epoch)
 
             logs = {"loss": loss.detach().item(), "lr": lr_scheduler.get_last_lr()[0]}
             progress_bar.set_postfix(**logs)