Skip to content

Open-Sora: Democratizing Efficient Video Production for All

Python

fix #209

Zangwei Zheng6745d2069d50
diff --git a/configs/opensora/train/16x512x512.py b/configs/opensora/train/16x512x512.py
index 885aad1..22031dc 100644
--- a/configs/opensora/train/16x512x512.py
+++ b/configs/opensora/train/16x512x512.py
@@ -30,7 +30,7 @@ vae = dict(
)
text_encoder = dict(
type="t5",
- from_pretrained="./pretrained_models/t5_ckpts",
+ from_pretrained="DeepFloyd/t5-v1_1-xxl",
model_max_length=120,
shardformer=True,
)
diff --git a/configs/opensora/train/360x512x512.py b/configs/opensora/train/360x512x512.py
index 7a6f759..0f4e170 100644
--- a/configs/opensora/train/360x512x512.py
+++ b/configs/opensora/train/360x512x512.py
@@ -31,7 +31,7 @@ vae = dict(
)
text_encoder = dict(
type="t5",
- from_pretrained="./pretrained_models/t5_ckpts",
+ from_pretrained="DeepFloyd/t5-v1_1-xxl",
model_max_length=120,
shardformer=True,
)
diff --git a/configs/opensora/train/64x512x512-sp.py b/configs/opensora/train/64x512x512-sp.py
index b0b9062..f9b138e 100644
--- a/configs/opensora/train/64x512x512-sp.py
+++ b/configs/opensora/train/64x512x512-sp.py
@@ -30,7 +30,7 @@ vae = dict(
)
text_encoder = dict(
type="t5",
- from_pretrained="./pretrained_models/t5_ckpts",
+ from_pretrained="DeepFloyd/t5-v1_1-xxl",
model_max_length=120,
shardformer=True,
)
diff --git a/configs/opensora/train/64x512x512.py b/configs/opensora/train/64x512x512.py
index dfcdcc0..81162de 100644
--- a/configs/opensora/train/64x512x512.py
+++ b/configs/opensora/train/64x512x512.py
@@ -30,7 +30,7 @@ vae = dict(
)
text_encoder = dict(
type="t5",
- from_pretrained="./pretrained_models/t5_ckpts",
+ from_pretrained="DeepFloyd/t5-v1_1-xxl",
model_max_length=120,
shardformer=True,
)
diff --git a/configs/pixart/inference/16x256x256.py b/configs/pixart/inference/16x256x256.py
index 6fc8ee6..fed26f5 100644
--- a/configs/pixart/inference/16x256x256.py
+++ b/configs/pixart/inference/16x256x256.py
@@ -15,7 +15,7 @@ vae = dict(
)
text_encoder = dict(
type="t5",
- from_pretrained="./pretrained_models/t5_ckpts",
+ from_pretrained="DeepFloyd/t5-v1_1-xxl",
model_max_length=120,
)
scheduler = dict(
diff --git a/configs/pixart/inference/1x1024MS.py b/configs/pixart/inference/1x1024MS.py
index 41cc97a..3b34ebf 100644
--- a/configs/pixart/inference/1x1024MS.py
+++ b/configs/pixart/inference/1x1024MS.py
@@ -17,7 +17,7 @@ vae = dict(
)
text_encoder = dict(
type="t5",
- from_pretrained="./pretrained_models/t5_ckpts",
+ from_pretrained="DeepFloyd/t5-v1_1-xxl",
model_max_length=120,
)
scheduler = dict(
diff --git a/configs/pixart/inference/1x256x256.py b/configs/pixart/inference/1x256x256.py
index 11e06d7..44ce10b 100644
--- a/configs/pixart/inference/1x256x256.py
+++ b/configs/pixart/inference/1x256x256.py
@@ -16,7 +16,7 @@ vae = dict(
)
text_encoder = dict(
type="t5",
- from_pretrained="./pretrained_models/t5_ckpts",
+ from_pretrained="DeepFloyd/t5-v1_1-xxl",
model_max_length=120,
)
scheduler = dict(
diff --git a/configs/pixart/inference/1x512x512.py b/configs/pixart/inference/1x512x512.py
index 5674259..243af4a 100644
--- a/configs/pixart/inference/1x512x512.py
+++ b/configs/pixart/inference/1x512x512.py
@@ -16,7 +16,7 @@ vae = dict(
)
text_encoder = dict(
type="t5",
- from_pretrained="./pretrained_models/t5_ckpts",
+ from_pretrained="DeepFloyd/t5-v1_1-xxl",
model_max_length=120,
)
scheduler = dict(
diff --git a/configs/pixart/train/16x256x256.py b/configs/pixart/train/16x256x256.py
index b47731e..dc91937 100644
--- a/configs/pixart/train/16x256x256.py
+++ b/configs/pixart/train/16x256x256.py
@@ -29,7 +29,7 @@ vae = dict(
)
text_encoder = dict(
type="t5",
- from_pretrained="./pretrained_models/t5_ckpts",
+ from_pretrained="DeepFloyd/t5-v1_1-xxl",
model_max_length=120,
shardformer=True,
)
diff --git a/configs/pixart/train/1x512x512.py b/configs/pixart/train/1x512x512.py
index 619c9aa..ec73b8c 100644
--- a/configs/pixart/train/1x512x512.py
+++ b/configs/pixart/train/1x512x512.py
@@ -30,7 +30,7 @@ vae = dict(
)
text_encoder = dict(
type="t5",
- from_pretrained="./pretrained_models/t5_ckpts",
+ from_pretrained="DeepFloyd/t5-v1_1-xxl",
model_max_length=120,
shardformer=True,
)
diff --git a/configs/pixart/train/64x512x512.py b/configs/pixart/train/64x512x512.py
index 628cf25..deebdf6 100644
--- a/configs/pixart/train/64x512x512.py
+++ b/configs/pixart/train/64x512x512.py
@@ -30,7 +30,7 @@ vae = dict(
)
text_encoder = dict(
type="t5",
- from_pretrained="./pretrained_models/t5_ckpts",
+ from_pretrained="DeepFloyd/t5-v1_1-xxl",
model_max_length=120,
shardformer=True,
)
diff --git a/docs/structure.md b/docs/structure.md
index 2d83620..3eb7c69 100644
--- a/docs/structure.md
+++ b/docs/structure.md
@@ -100,7 +100,7 @@ vae = dict(
)
text_encoder = dict(
type="t5", # Select text encoder type (t5, clip)
- from_pretrained="./pretrained_models/t5_ckpts", # Load from pretrained text encoder
+ from_pretrained="DeepFloyd/t5-v1_1-xxl", # Load from pretrained text encoder
model_max_length=120, # Maximum length of input text
)
scheduler = dict(
@@ -153,7 +153,7 @@ vae = dict(
)
text_encoder = dict(
type="t5",
- from_pretrained="./pretrained_models/t5_ckpts",
+ from_pretrained="DeepFloyd/t5-v1_1-xxl",
model_max_length=120,
shardformer=True, # Enable shardformer for T5 acceleration
)
diff --git a/docs/zh_CN/structure.md b/docs/zh_CN/structure.md
index 2d83620..3eb7c69 100644
--- a/docs/zh_CN/structure.md
+++ b/docs/zh_CN/structure.md
@@ -100,7 +100,7 @@ vae = dict(
)
text_encoder = dict(
type="t5", # Select text encoder type (t5, clip)
- from_pretrained="./pretrained_models/t5_ckpts", # Load from pretrained text encoder
+ from_pretrained="DeepFloyd/t5-v1_1-xxl", # Load from pretrained text encoder
model_max_length=120, # Maximum length of input text
)
scheduler = dict(
@@ -153,7 +153,7 @@ vae = dict(
)
text_encoder = dict(
type="t5",
- from_pretrained="./pretrained_models/t5_ckpts",
+ from_pretrained="DeepFloyd/t5-v1_1-xxl",
model_max_length=120,
shardformer=True, # Enable shardformer for T5 acceleration
)