Spaces:
Paused
Paused
Update train.py
Browse files
train.py
CHANGED
@@ -1,80 +1,55 @@
|
|
1 |
import os
|
2 |
import torch
|
3 |
-
from
|
4 |
-
|
5 |
-
|
6 |
-
|
7 |
-
|
8 |
-
UNet2DConditionModel
|
9 |
)
|
10 |
-
from transformers import CLIPTextModel, CLIPTokenizer
|
11 |
-
from peft import LoraConfig, get_peft_model
|
12 |
|
13 |
-
|
|
|
|
|
|
|
|
|
14 |
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
use_auth_token=True
|
20 |
)
|
21 |
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
# 1) grab the model locally
|
30 |
-
print("📥 Downloading Flux‑Dev model…")
|
31 |
-
model_path = snapshot_download(MODEL_ID, local_dir="./fluxdev-model")
|
32 |
|
33 |
-
# 2
|
34 |
-
|
35 |
-
|
36 |
-
|
|
|
|
|
37 |
)
|
38 |
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
)
|
43 |
|
44 |
-
|
45 |
-
|
46 |
-
model_path, subfolder="text_encoder", torch_dtype=torch.float16
|
47 |
-
)
|
48 |
-
tokenizer = CLIPTokenizer.from_pretrained(
|
49 |
-
model_path, subfolder="tokenizer"
|
50 |
-
)
|
51 |
|
52 |
-
|
53 |
-
|
54 |
-
|
|
|
|
|
55 |
)
|
56 |
|
57 |
-
|
58 |
-
|
59 |
-
pipe = StableDiffusionPipeline(
|
60 |
-
vae=vae,
|
61 |
-
text_encoder=text_encoder,
|
62 |
-
tokenizer=tokenizer,
|
63 |
-
unet=unet,
|
64 |
-
scheduler=scheduler
|
65 |
-
).to("cuda")
|
66 |
-
|
67 |
-
# 4) apply LoRA
|
68 |
-
print("🧠 Applying LoRA…")
|
69 |
-
lora_config = LoraConfig(r=16, lora_alpha=16, bias="none", task_type="CAUSAL_LM")
|
70 |
-
pipe.unet = get_peft_model(pipe.unet, lora_config)
|
71 |
-
|
72 |
-
# 5) your training loop (or dummy loop for illustration)
|
73 |
-
print("🚀 Starting fine‑tuning…")
|
74 |
-
for step in range(100):
|
75 |
-
print(f"Training step {step+1}/100")
|
76 |
-
# …insert your actual data‑loader and loss/backprop here…
|
77 |
|
78 |
-
|
79 |
-
pipe.save_pretrained(output_dir)
|
80 |
-
print("✅ Done. LoRA weights in", output_dir)
|
|
|
1 |
import os
|
2 |
import torch
|
3 |
+
from aitoolkit import (
|
4 |
+
LoRATrainer,
|
5 |
+
StableDiffusionModel,
|
6 |
+
LoRAConfig,
|
7 |
+
ImageTextDataset,
|
|
|
8 |
)
|
|
|
|
|
9 |
|
10 |
+
# 1. Configuration
|
11 |
+
MODEL_ID = "HiDream-ai/HiDream-I1-Dev" # or your gated FLUX model if you have access
|
12 |
+
DATA_DIR = "/workspace/data"
|
13 |
+
OUTPUT_DIR = "/workspace/lora-trained"
|
14 |
+
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
15 |
|
16 |
+
lora_cfg = LoRAConfig(
|
17 |
+
rank=16,
|
18 |
+
alpha=16,
|
19 |
+
bias="none",
|
|
|
20 |
)
|
21 |
|
22 |
+
training_args = {
|
23 |
+
"num_train_steps": 100,
|
24 |
+
"batch_size": 4,
|
25 |
+
"learning_rate": 1e-4,
|
26 |
+
"save_every_n_steps": 50,
|
27 |
+
"output_dir": OUTPUT_DIR,
|
28 |
+
}
|
|
|
|
|
|
|
29 |
|
30 |
+
# 2. Load base diffusion model
|
31 |
+
model = StableDiffusionModel.from_pretrained(
|
32 |
+
MODEL_ID,
|
33 |
+
torch_dtype=torch.float16,
|
34 |
+
device=DEVICE,
|
35 |
+
use_auth_token=True, # if it’s a gated repo
|
36 |
)
|
37 |
|
38 |
+
# 3. Prepare your dataset
|
39 |
+
# Expects pairs of image files + .txt captions in DATA_DIR
|
40 |
+
dataset = ImageTextDataset(data_root=DATA_DIR, image_size=512)
|
|
|
41 |
|
42 |
+
# 4. Hook up the LoRA adapter
|
43 |
+
model.apply_lora(lora_cfg)
|
|
|
|
|
|
|
|
|
|
|
44 |
|
45 |
+
# 5. Create the trainer and kickoff
|
46 |
+
trainer = LoRATrainer(
|
47 |
+
model=model,
|
48 |
+
dataset=dataset,
|
49 |
+
args=training_args,
|
50 |
)
|
51 |
|
52 |
+
print("🚀 Starting training with AI‑Toolkit…")
|
53 |
+
trainer.train()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
|
55 |
+
print(f"✅ Done! Fine-tuned weights saved to {OUTPUT_DIR}")
|
|
|
|