Image-to-Text
Transformers
Safetensors
Tibetan
vision-encoder-decoder
image-text-to-text
ocr
tibetan
document-ai
trocr
Instructions to use TibetanCodexAITeam/PechaBridgeOCR with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use TibetanCodexAITeam/PechaBridgeOCR with Transformers:
# Use a pipeline as a high-level helper # Warning: Pipeline type "image-to-text" is no longer supported in transformers v5. # You must load the model directly (see below) or downgrade to v4.x with: # 'pip install "transformers<5.0.0' from transformers import pipeline pipe = pipeline("image-to-text", model="TibetanCodexAITeam/PechaBridgeOCR")# Load model directly from transformers import AutoTokenizer, AutoModelForMultimodalLM tokenizer = AutoTokenizer.from_pretrained("TibetanCodexAITeam/PechaBridgeOCR") model = AutoModelForMultimodalLM.from_pretrained("TibetanCodexAITeam/PechaBridgeOCR", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Upload DONUT checkpoint checkpoint-154000 via PechaBridge
Browse files- repro/metrics.json +6 -0
repro/metrics.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"metric_name": "cer",
|
| 3 |
+
"definition": "sum(edit_distance(pred_norm, ref_norm)) / sum(max(1, len(ref_norm)))",
|
| 4 |
+
"aggregation": "global sums (DDP all_reduce)",
|
| 5 |
+
"empty_reference_policy": "references with empty normalized text are excluded from valid_n and counted in empty_ref_count"
|
| 6 |
+
}
|