Update README.md
Browse files
README.md
CHANGED
|
@@ -230,10 +230,10 @@ def load_image(image_file):
|
|
| 230 |
return image
|
| 231 |
|
| 232 |
|
| 233 |
-
config = AutoConfig.from_pretrained("
|
| 234 |
-
tokenizer = AutoTokenizer.from_pretrained("
|
| 235 |
-
processor = AutoProcessor.from_pretrained("
|
| 236 |
-
model = AutoModelForCausalLM.from_pretrained("
|
| 237 |
model.eval()
|
| 238 |
|
| 239 |
# For single image and text
|
|
@@ -356,7 +356,7 @@ for idx, o in enumerate(outs):
|
|
| 356 |
| MLP | 6.3M | 2 | - | 2048 | - |
|
| 357 |
| LM | 1.2B | 16 | 16 | 2048 | - |
|
| 358 |
|
| 359 |
-
We initialize the vision encoder from [CLIP-ViT-L/14@336](https://huggingface.co/openai/clip-vit-large-patch14-336) and initialize LM from [AMD OLMo 1B SFT](https://huggingface.co/
|
| 360 |
|
| 361 |
## Training Stages
|
| 362 |
|
|
|
|
| 230 |
return image
|
| 231 |
|
| 232 |
|
| 233 |
+
config = AutoConfig.from_pretrained("amd/Instella-VL-1B", trust_remote_code=True)
|
| 234 |
+
tokenizer = AutoTokenizer.from_pretrained("amd/Instella-VL-1B", config=config, trust_remote_code=True)
|
| 235 |
+
processor = AutoProcessor.from_pretrained("amd/Instella-VL-1B", trust_remote_code=True)
|
| 236 |
+
model = AutoModelForCausalLM.from_pretrained("amd/Instella-VL-1B", trust_remote_code=True).to('cuda') # or 'cpu'
|
| 237 |
model.eval()
|
| 238 |
|
| 239 |
# For single image and text
|
|
|
|
| 356 |
| MLP | 6.3M | 2 | - | 2048 | - |
|
| 357 |
| LM | 1.2B | 16 | 16 | 2048 | - |
|
| 358 |
|
| 359 |
+
We initialize the vision encoder from [CLIP-ViT-L/14@336](https://huggingface.co/openai/clip-vit-large-patch14-336) and initialize LM from [AMD OLMo 1B SFT](https://huggingface.co/amd/AMD-OLMo-1B-SFT)
|
| 360 |
|
| 361 |
## Training Stages
|
| 362 |
|