Update train_text_to_image_sdxl.py

jiagaoxiang · web-flow · commit 91812beab0c1 · 2024-07-10T13:04:19.000-07:00
Enable VAE hash to be able to change with args change. If not, train_dataset_with_embeddiings may have row number inconsistency with train_dataset_with_vae.
diff --git a/examples/text_to_image/train_text_to_image_sdxl.py b/examples/text_to_image/train_text_to_image_sdxl.py
@@ -917,7 +917,7 @@ def preprocess_train(examples):
         # fingerprint used by the cache for the other processes to load the result
         # details: https://github.com/huggingface/diffusers/pull/4038#discussion_r1266078401
         new_fingerprint = Hasher.hash(args)
-        new_fingerprint_for_vae = Hasher.hash(vae_path)
+        new_fingerprint_for_vae = Hasher.hash((vae_path, args))
         train_dataset_with_embeddings = train_dataset.map(
             compute_embeddings_fn, batched=True, new_fingerprint=new_fingerprint
         )

Original file line number	Diff line number	Diff line change
`@@ -917,7 +917,7 @@ def preprocess_train(examples):`
`917`	`917`	`# fingerprint used by the cache for the other processes to load the result`
`918`	`918`	`# details: https://github.com/huggingface/diffusers/pull/4038#discussion_r1266078401`
`919`	`919`	`new_fingerprint = Hasher.hash(args)`
`920`		`- new_fingerprint_for_vae = Hasher.hash(vae_path)`
	`920`	`+ new_fingerprint_for_vae = Hasher.hash((vae_path, args))`
`921`	`921`	`train_dataset_with_embeddings = train_dataset.map(`
`922`	`922`	`compute_embeddings_fn, batched=True, new_fingerprint=new_fingerprint`
`923`	`923`	`)`