Release OCR checkpoint-184000

._sentencepiece.model

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:9624dfbd84676e99c6fc8cb4b73d15f619d757d51bfb2cff84af484a0c3b1061
-size 4096

._spiece.model

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:9624dfbd84676e99c6fc8cb4b73d15f619d757d51bfb2cff84af484a0c3b1061
-size 4096

README.md

@@ -1,88 +1,110 @@
+---
+language:
+- bo
+license: mit
+library_name: transformers
+pipeline_tag: image-to-text
+base_model: microsoft/trocr-base-stage1
+tags:
+- ocr
+- tibetan
+- document-ai
+- trocr
+- vision-encoder-decoder
+---
+
 # PechaBridgeOCR
 
-Tibetan OCR model (~200M parameters, DONUT / VisionEncoderDecoder architecture) fine-tuned on Tibetan pecha line images with PechaBridge.
+Tibetan line OCR model (323.24M parameters, VIT encoder + TROCR decoder in a VisionEncoderDecoder architecture) fine-tuned for traditional Tibetan pecha scans with PechaBridge.
 
-> **Important:** This model requires a custom BDRC-style grayscale preprocessing pipeline (adaptive binarization, background normalisation, aspect-preserving resize and padding). Using `AutoImageProcessor` directly will apply standard ImageNet normalisation and produce poor results. **Use PechaBridge's `batch-ocr` CLI** for correct end-to-end inference.
+> **Important:** This checkpoint expects PechaBridge's `gray` preprocessing followed by a fixed `256×1024` resize. The `gray` pipeline uses the minimum RGB channel as grayscale and does not binarize the image. Use the PechaBridge CLI for the supported end-to-end path; raw image input or a generic `AutoTokenizer` path will not reproduce the training pipeline.
 
 ## Recommended usage — PechaBridge CLI
 
 ```bash
-# 1. Clone PechaBridge and install dependencies
 git clone https://github.com/CodexAITeam/PechaBridge.git && cd PechaBridge
 pip install -r requirements.txt
-
-# 2. Download this model (and the line segmentation model)
 python cli.py download-models
 
-# 3a. Run batch OCR on a folder of pecha page images
 python cli.py batch-ocr \
     --ocr-model     models/ocr/PechaBridgeOCR \
     --line-model    models/line_segmentation/PechaBridgeLineSegmentation.pt \
     --layout-engine yolo_line \
     --ocr-engine    donut \
     --input-dir     /path/to/pecha/images
-
-# 3b. Or use the BDRC layout engine (auto-downloads BDRC line models)
-python cli.py batch-ocr \
-    --ocr-model     models/ocr/PechaBridgeOCR \
-    --layout-engine bdrc_line \
-    --ocr-engine    donut \
-    --input-dir     /path/to/pecha/images
-
-# 3c. Download + OCR a Staatsbibliothek zu Berlin pecha in one command
-python cli.py batch-ocr \
-    --ppn           337138764X \
-    --ocr-model     models/ocr/PechaBridgeOCR \
-    --layout-engine bdrc_line \
-    --ocr-engine    donut
 ```
 
-Each image produces a `.txt` transcript and an `*_overlay.jpg` with detected line boxes drawn on the source image.
+Each page image produces a `.txt` transcript and an `*_overlay.jpg` preview.
 
 ## Advanced: standalone Python usage
 
-If you need to call the model directly from Python, you must apply the BDRC-style preprocessing manually before passing pixel values to the model. See `pechabridge/ocr/preprocess_bdrc.py` in the PechaBridge repository for the full preprocessing pipeline (`gray` pipeline was used for this checkpoint).
+The tokenizer is backed by BoSentencePiece plus PechaBridge-specific special tokens. Load it with PechaBridge's adapter rather than `AutoTokenizer`.
 
 ```python
-# Minimal example — BDRC gray preprocessing applied explicitly
+from pathlib import Path
 import torch
-from transformers import VisionEncoderDecoderModel, AutoTokenizer
-from pechabridge.ocr.preprocess_bdrc import (
-    BDRCPreprocessConfig,
-    preprocess_image_bdrc,       # returns grayscale PIL Image (mode 'L')
-    bdrc_image_to_normalized_tensor,  # grayscale PIL → float32 HW in [-1, 1]
-)
+from huggingface_hub import snapshot_download
 from PIL import Image
-import numpy as np
-
-model = VisionEncoderDecoderModel.from_pretrained("TibetanCodexAITeam/PechaBridgeOCR")
-tokenizer = AutoTokenizer.from_pretrained("TibetanCodexAITeam/PechaBridgeOCR")
-
-# 1. Load the line crop as RGB
+from transformers import AutoImageProcessor, VisionEncoderDecoderModel
+from pechabridge.ocr.preprocess_bdrc import BDRCPreprocessConfig, preprocess_image_bdrc
+from pechabridge.ocr.sentencepiece_tokenizer_adapter import load_sentencepiece_tokenizer
+
+model_dir = Path(snapshot_download("TibetanCodexAITeam/PechaBridgeOCR"))
+model = VisionEncoderDecoderModel.from_pretrained(model_dir).eval()
+image_processor = AutoImageProcessor.from_pretrained(model_dir, use_fast=False)
+tokenizer = load_sentencepiece_tokenizer(model_dir)
 image = Image.open('line_crop.png').convert('RGB')
-
-# 2. Apply BDRC preprocessing:
-#    - converts to grayscale (luma by default)
-#    - optionally normalises background, binarises, pads/resizes
-#    Returns a grayscale PIL Image (mode 'L')
-cfg = BDRCPreprocessConfig.ocr_line_defaults()  # defaults used during training
-gray_pil = preprocess_image_bdrc(image, cfg)
-
-# 3. Normalise to float32 in [-1, 1] and build a 3-channel tensor
-#    (model expects C=3; replicate the single gray channel)
-gray_hw = bdrc_image_to_normalized_tensor(image, cfg)  # shape: (H, W), float32
-pixel_values = torch.tensor(gray_hw).unsqueeze(0).unsqueeze(0)  # (1, 1, H, W)
-pixel_values = pixel_values.expand(-1, 3, -1, -1)               # (1, 3, H, W)
-
-generated_ids = model.generate(pixel_values)
+cfg = BDRCPreprocessConfig.vit_defaults()
+cfg = BDRCPreprocessConfig.from_dict({
+    **cfg.to_dict(), 'binarize': False, 'gray_mode': 'min_rgb',
+})
+prepared = preprocess_image_bdrc(image=image, config=cfg).convert('RGB')
+pixel_values = image_processor(images=prepared, return_tensors='pt').pixel_values
+with torch.inference_mode():
+    generated_ids = model.generate(pixel_values)
 text = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+print(text)
 ```
 
 ## Model details
 
-- **Checkpoint**: `checkpoint-154000` (step 154000)
+- **Checkpoint**: `checkpoint-184000` (step 184000)
 - **Image preprocessing pipeline**: `gray`
-- **Repro bundle included**: yes — preprocessing config is in `repro/`
-- **Architecture**: Swin Transformer encoder (hidden_size=768, 12 layers) + BART decoder (d_model=1024, 12 layers), ~200M parameters total
+- **Input geometry**: fixed `256×1024` resize, RGB tensor normalized with mean/std `0.5`
+- **Architecture**: VIT encoder + TROCR decoder, 323.24M parameters
+- **Repro configuration included**: yes — preprocessing, generation, normalization, and metric definitions are in `repro/`
 - **Training framework**: [PechaBridge](https://github.com/CodexAITeam/PechaBridge)
 - **Training data**: Tibetan pecha line images from OpenPecha and BDRC collections
+
+## Evaluation
+
+| Checkpoint | Internal validation CER | Valid samples |
+|---|---:|---:|
+| `checkpoint-184000` | 0.5751% | 241 |
+
+CER is the global character edit distance divided by total normalized reference length. This is a training-time internal validation result, not an independent cross-collection benchmark.
+
+### External multi-source evaluation
+
+On a separate 3,306-sample evaluation set, mean CER was **0.80%**. **1.63%** of samples exceeded 10% CER and **0.57%** exceeded 20% CER.
+
+| Source dataset | Samples | Mean CER |
+|---|---:|---:|
+| OCR-Norbuketaka | 2,350 | 0.43% |
+| OCR-Google_Books | 812 | 1.59% |
+| OCR-Lhasakanjur | 118 | 1.92% |
+| OCR-Drutsa | 14 | 3.54% |
+| OCR-Betsug | 12 | 3.86% |
+
+This externally supplied result summary was not independently reproduced from artifacts included in this release.
+
+## Intended use and limitations
+
+- Intended for OCR of individual Tibetan pecha text-line crops.
+- Page-level use requires a separate line-segmentation/layout stage.
+- Performance may degrade on other scripts, modern book layouts, handwriting, severe blur, unusual scan colors, or collections not represented during training.
+- Outputs should be reviewed before scholarly, archival, or other high-impact use.
+
+## License
+
+MIT. See the PechaBridge repository for the project license.

config.json

@@ -54,9 +54,9 @@
     "use_bfloat16": false,
     "use_cache": false,
     "use_learned_position_embeddings": false,
-    "vocab_size": 20008
+    "vocab_size": 20007
   },
-  "decoder_start_token_id": 20005,
+  "decoder_start_token_id": 20004,
   "dtype": "float32",
   "encoder": {
     "_name_or_path": "",
@@ -78,7 +78,10 @@
       "0": "LABEL_0",
       "1": "LABEL_1"
     },
-    "image_size": 384,
+    "image_size": [
+      256,
+      1024
+    ],
     "initializer_range": 0.02,
     "intermediate_size": 3072,
     "is_decoder": false,
@@ -111,12 +114,12 @@
     "torchscript": false,
     "use_bfloat16": false
   },
-  "eos_token_id": 20006,
+  "eos_token_id": 20005,
   "is_encoder_decoder": true,
   "model_type": "vision-encoder-decoder",
-  "pad_token_id": 20000,
+  "pad_token_id": 3,
   "tie_word_embeddings": false,
-  "transformers_version": "5.2.0",
+  "transformers_version": "5.6.2",
   "use_cache": false,
-  "vocab_size": 20008
+  "vocab_size": 20007
 }

generation_config.json

@@ -4,23 +4,22 @@
   "assistant_lookbehind": 10,
   "bad_words_ids": [
     [
-      20005
+      20004
     ]
   ],
   "bos_token_id": 0,
-  "decoder_start_token_id": 20005,
+  "decoder_start_token_id": 20004,
   "diversity_penalty": 0.0,
   "do_sample": false,
   "early_stopping": false,
   "encoder_no_repeat_ngram_size": 0,
   "encoder_repetition_penalty": 1.0,
-  "eos_token_id": 20006,
+  "eos_token_id": 20005,
   "epsilon_cutoff": 0.0,
   "eta_cutoff": 0.0,
   "length_penalty": 1.0,
   "max_length": 160,
   "min_length": 0,
-  "min_new_tokens": 0,
   "no_repeat_ngram_size": 3,
   "num_assistant_tokens": 20,
   "num_assistant_tokens_schedule": "constant",
@@ -28,18 +27,18 @@
   "num_beams": 1,
   "num_return_sequences": 1,
   "output_scores": false,
-  "pad_token_id": 20000,
+  "pad_token_id": 3,
   "remove_invalid_values": false,
   "repetition_penalty": 1.15,
   "return_dict_in_generate": false,
   "suppress_tokens": [
-    20005
+    20004
   ],
   "target_lookbehind": 10,
   "temperature": 1.0,
   "top_k": 50,
   "top_p": 1.0,
-  "transformers_version": "5.2.0",
+  "transformers_version": "5.6.2",
   "typical_p": 1.0,
   "use_cache": false
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:76f918953adde7ddf91aa4268649db486df81448838b1b01a2d072f71c2cd8d6
-size 1291653280
+oid sha256:b876faf7dac748c9a9d953f9b2dfc6d32f7b9dc006bdd285d8c687524d233266
+size 1293021344

preprocessor_config.json

@@ -1,5 +1,4 @@
 {
-  "data_format": "channels_first",
   "do_normalize": true,
   "do_pad": false,
   "do_rescale": true,
@@ -9,7 +8,7 @@
     0.5,
     0.5
   ],
-  "image_processor_type": "ViTImageProcessorFast",
+  "image_processor_type": "ViTImageProcessor",
   "image_std": [
     0.5,
     0.5,
@@ -18,7 +17,7 @@
   "resample": 2,
   "rescale_factor": 0.00392156862745098,
   "size": {
-    "height": 384,
-    "width": 384
+    "height": 256,
+    "width": 1024
   }
 }

repro/cer_step_154000.json

@@ -1,13 +0,0 @@
-{
-  "step": 154000,
-  "checkpoint_dir": "/home/ubuntu/data/PechaBridge/models/donut_full_gray_aug_restart/checkpoint-154000",
-  "repro_cer": 0.027263326125787347,
-  "repro_cer_percent": 2.7263326125787346,
-  "repro_valid_n": 300,
-  "repro_empty_ref_count": 0,
-  "repro_empty_pred_count": 0,
-  "repro_edit_sum": 1160,
-  "repro_ref_len_sum": 42548,
-  "repro_runtime": 41.041670083999634,
-  "repro_world_size": 1
-}

repro/generate_config.json

@@ -1,11 +1,11 @@
 {
-  "decoder_start_token_id": 20005,
+  "decoder_start_token_id": 20004,
   "bos_token_id": 0,
-  "eos_token_id": 20006,
-  "pad_token_id": 20000,
+  "eos_token_id": 20005,
+  "pad_token_id": 3,
   "max_length": 160,
   "max_new_tokens": null,
-  "min_new_tokens": 0,
+  "min_new_tokens": null,
   "num_beams": 1,
   "do_sample": false,
   "temperature": 1.0,
@@ -16,7 +16,7 @@
   "early_stopping": false,
   "bad_words_ids": [
     [
-      20005
+      20004
     ]
   ],
   "forced_bos_token_id": null,

repro/image_preprocess.json

@@ -2,10 +2,10 @@
   "pipeline": "gray",
   "color_space": "RGB",
   "channel_order": "channels_first",
-  "processor_class": "ViTImageProcessorFast",
+  "processor_class": "ViTImageProcessor",
   "image_processor_config": {
-    "data_format": "channels_first",
     "do_normalize": true,
+    "do_pad": false,
     "do_rescale": true,
     "do_resize": true,
     "image_mean": [
@@ -13,7 +13,7 @@
       0.5,
       0.5
     ],
-    "image_processor_type": "ViTImageProcessorFast",
+    "image_processor_type": "ViTImageProcessor",
     "image_std": [
       0.5,
       0.5,
@@ -22,14 +22,13 @@
     "resample": 2,
     "rescale_factor": 0.00392156862745098,
     "size": {
-      "height": 384,
-      "width": 384
-    },
-    "do_pad": false
+      "height": 256,
+      "width": 1024
+    }
   },
   "size": {
-    "height": 384,
-    "width": 384
+    "height": 256,
+    "width": 1024
   },
   "do_resize": true,
   "do_rescale": true,

sentencepiece.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6032a003043e791f2e573a8cce72e512cd6027bef640dcb5457920cd67724fdc
-size 911556
+oid sha256:3f59e5bebc3be675fb79cb6598a2aff557ecfc40bdb86450780771542cba3f19
+size 906758

spiece.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6032a003043e791f2e573a8cce72e512cd6027bef640dcb5457920cd67724fdc
-size 911556
+oid sha256:3f59e5bebc3be675fb79cb6598a2aff557ecfc40bdb86450780771542cba3f19
+size 906758

tokenizer.json

@@ -1,243 +0,0 @@
-{
-  "version": "1.0",
-  "truncation": null,
-  "padding": null,
-  "added_tokens": [
-    {
-      "id": 0,
-      "content": "<pad>",
-      "single_word": false,
-      "lstrip": false,
-      "rstrip": false,
-      "normalized": false,
-      "special": true
-    },
-    {
-      "id": 1,
-      "content": "<unk>",
-      "single_word": false,
-      "lstrip": false,
-      "rstrip": false,
-      "normalized": false,
-      "special": true
-    },
-    {
-      "id": 2,
-      "content": "[CLS]",
-      "single_word": false,
-      "lstrip": false,
-      "rstrip": false,
-      "normalized": false,
-      "special": true
-    },
-    {
-      "id": 3,
-      "content": "[SEP]",
-      "single_word": false,
-      "lstrip": false,
-      "rstrip": false,
-      "normalized": false,
-      "special": true
-    },
-    {
-      "id": 4,
-      "content": "[MASK]",
-      "single_word": false,
-      "lstrip": false,
-      "rstrip": false,
-      "normalized": false,
-      "special": true
-    },
-    {
-      "id": 5,
-      "content": "<s>",
-      "single_word": false,
-      "lstrip": false,
-      "rstrip": false,
-      "normalized": false,
-      "special": true
-    },
-    {
-      "id": 6,
-      "content": "</s>",
-      "single_word": false,
-      "lstrip": false,
-      "rstrip": false,
-      "normalized": false,
-      "special": true
-    }
-  ],
-  "normalizer": {
-    "type": "Sequence",
-    "normalizers": [
-      {
-        "type": "Replace",
-        "pattern": {
-          "String": "``"
-        },
-        "content": "\""
-      },
-      {
-        "type": "Replace",
-        "pattern": {
-          "String": "''"
-        },
-        "content": "\""
-      },
-      {
-        "type": "NFKD"
-      },
-      {
-        "type": "StripAccents"
-      },
-      {
-        "type": "Lowercase"
-      },
-      {
-        "type": "Replace",
-        "pattern": {
-          "Regex": " {2,}"
-        },
-        "content": " "
-      },
-      {
-        "type": "NFKD"
-      },
-      {
-        "type": "StripAccents"
-      },
-      {
-        "type": "Lowercase"
-      },
-      {
-        "type": "Replace",
-        "pattern": {
-          "Regex": " {2,}"
-        },
-        "content": " "
-      }
-    ]
-  },
-  "pre_tokenizer": {
-    "type": "Sequence",
-    "pretokenizers": [
-      {
-        "type": "WhitespaceSplit"
-      },
-      {
-        "type": "Metaspace",
-        "replacement": "▁",
-        "prepend_scheme": "always",
-        "split": true
-      }
-    ]
-  },
-  "post_processor": {
-    "type": "TemplateProcessing",
-    "single": [
-      {
-        "SpecialToken": {
-          "id": "[CLS]",
-          "type_id": 0
-        }
-      },
-      {
-        "Sequence": {
-          "id": "A",
-          "type_id": 0
-        }
-      },
-      {
-        "SpecialToken": {
-          "id": "[SEP]",
-          "type_id": 0
-        }
-      }
-    ],
-    "pair": [
-      {
-        "SpecialToken": {
-          "id": "[CLS]",
-          "type_id": 0
-        }
-      },
-      {
-        "Sequence": {
-          "id": "A",
-          "type_id": 0
-        }
-      },
-      {
-        "SpecialToken": {
-          "id": "[SEP]",
-          "type_id": 0
-        }
-      },
-      {
-        "Sequence": {
-          "id": "B",
-          "type_id": 1
-        }
-      },
-      {
-        "SpecialToken": {
-          "id": "[SEP]",
-          "type_id": 1
-        }
-      }
-    ],
-    "special_tokens": {
-      "[CLS]": {
-        "id": "[CLS]",
-        "ids": [
-          2
-        ],
-        "tokens": [
-          "[CLS]"
-        ]
-      },
-      "[SEP]": {
-        "id": "[SEP]",
-        "ids": [
-          3
-        ],
-        "tokens": [
-          "[SEP]"
-        ]
-      }
-    }
-  },
-  "decoder": {
-    "type": "Metaspace",
-    "replacement": "▁",
-    "prepend_scheme": "always",
-    "split": true
-  },
-  "model": {
-    "type": "Unigram",
-    "unk_id": 1,
-    "vocab": [
-      [
-        "<pad>",
-        0.0
-      ],
-      [
-        "<unk>",
-        0.0
-      ],
-      [
-        "[CLS]",
-        0.0
-      ],
-      [
-        "[SEP]",
-        0.0
-      ],
-      [
-        "[MASK]",
-        0.0
-      ]
-    ],
-    "byte_fallback": false
-  }
-}

tokenizer_config.json

@@ -1,6 +1,6 @@
 {
   "tokenizer_class": "SentencePieceTokenizerAdapter",
-  "source_tokenizer_dir": "/home/ubuntu/data/PechaBridge/ext/BoSentencePiece",
+  "source_tokenizer_dir": "/home/nico/Code/PechaBridge/ext/BoSentencePiece",
   "model_max_length": 1000000000000000019884624838656,
   "unk_token": "<unk>",
   "pad_token": "<pad>",