Update README.md
Browse files
README.md
CHANGED
|
@@ -24,7 +24,8 @@ Training configuration:
|
|
| 24 |
* Hardware: 8xH100s
|
| 25 |
|
| 26 |
Training was conducted using FP16 mixed-precision and DeepSpeed Zero2 scheme. The vision tower of the model
|
| 27 |
-
was kept frozen during the training.
|
|
|
|
| 28 |
|
| 29 |
## Inference
|
| 30 |
|
|
@@ -35,7 +36,7 @@ from PIL import Image
|
|
| 35 |
import requests
|
| 36 |
|
| 37 |
|
| 38 |
-
folder_path = "diffusers
|
| 39 |
model = (
|
| 40 |
AutoModelForCausalLM.from_pretrained(folder_path, torch_dtype=torch.float16, trust_remote_code=True)
|
| 41 |
.to("cuda")
|
|
@@ -44,7 +45,7 @@ model = (
|
|
| 44 |
processor = AutoProcessor.from_pretrained(folder_path, trust_remote_code=True)
|
| 45 |
|
| 46 |
prompts = ["<COLOR>", "<LIGHTING>", "<LIGHTING_TYPE>", "<COMPOSITION>"]
|
| 47 |
-
|
| 48 |
image = Image.open(img_path).convert("RGB")
|
| 49 |
|
| 50 |
with torch.no_grad() and torch.inference_mode():
|
|
|
|
| 24 |
* Hardware: 8xH100s
|
| 25 |
|
| 26 |
Training was conducted using FP16 mixed-precision and DeepSpeed Zero2 scheme. The vision tower of the model
|
| 27 |
+
was kept frozen during the training. We used the [diffusers/ShotDEAD-v0](https://huggingface.co/datasets/diffusers/ShotDEAD-v0)
|
| 28 |
+
dataset for conducting training.
|
| 29 |
|
| 30 |
## Inference
|
| 31 |
|
|
|
|
| 36 |
import requests
|
| 37 |
|
| 38 |
|
| 39 |
+
folder_path = "diffusers/shot-categorizer-v0"
|
| 40 |
model = (
|
| 41 |
AutoModelForCausalLM.from_pretrained(folder_path, torch_dtype=torch.float16, trust_remote_code=True)
|
| 42 |
.to("cuda")
|
|
|
|
| 45 |
processor = AutoProcessor.from_pretrained(folder_path, trust_remote_code=True)
|
| 46 |
|
| 47 |
prompts = ["<COLOR>", "<LIGHTING>", "<LIGHTING_TYPE>", "<COMPOSITION>"]
|
| 48 |
+
img_path = "./assets/image_3.jpg"
|
| 49 |
image = Image.open(img_path).convert("RGB")
|
| 50 |
|
| 51 |
with torch.no_grad() and torch.inference_mode():
|