Upload Script Reproduction checkpoint for NLP4DH 2026

README.md

@@ -0,0 +1,70 @@
+---
+language:
+- egy
+- de
+tags:
+- translation
+- ancient-egyptian
+- hieroglyphics
+- contamination-study
+- nlp4dh
+license: mit
+base_model: facebook/m2m100_418M
+pipeline_tag: translation
+---
+
+# Script Reproduction — Hieroglyphic-to-German Translation
+
+This model is part of the paper **"Data Contamination in Neural Machine Translation of Ancient Egyptian Hieroglyphics"** (NLP4DH 2026).
+
+## Model Description
+
+M2M-100 (418M) retrained using the original train.py script from the hiero-transformer repository with default hyperparameters (epochs=20, batch_size=16, lr=3e-5). This represents the closest replication of the original training procedure.
+
+**Task:** Ancient Egyptian hieroglyphics (Gardiner notation) → German translation
+
+## Performance
+
+| Subset | BLEU |
+|--------|------|
+| All (n=50) | 42.2 |
+| Contaminated (n=16) | 77.5 |
+| Clean (n=34) | 33.8 |
+
+> **Important:** The "All" and "Contaminated" BLEU scores are inflated due to target-side data contamination (32% of test targets appear in training). The **Clean** score represents genuine translation quality on decontaminated samples.
+
+## Usage
+
+```python
+from transformers import M2MForConditionalGeneration, M2MTokenizer
+
+model = M2MForConditionalGeneration.from_pretrained("bumblelbee/hiero-m2m100-script-reproduction")
+tokenizer = M2MTokenizer.from_pretrained("bumblelbee/hiero-m2m100-script-reproduction")
+
+# Gardiner notation input (hieroglyphic transliteration)
+source = "D36 N35 G17 D21 X1 O34"
+
+tokenizer.src_lang = "ea"
+inputs = tokenizer(source, return_tensors="pt")
+generated = model.generate(**inputs, forced_bos_token_id=tokenizer.get_lang_id("de"))
+output = tokenizer.decode(generated[0], skip_special_tokens=True)
+print(output)
+```
+
+## Training Data
+
+Fine-tuned on 18,669 ea→de pairs from the Thesaurus Linguae Aegyptiae (TLA), maintained by the Berlin-Brandenburg Academy of Sciences and Humanities.
+
+## Citation
+
+```bibtex
+@inproceedings{contamination2026nlp4dh,
+  title={Data Contamination in Neural Machine Translation of Ancient Egyptian Hieroglyphics},
+  booktitle={Proceedings of the Workshop on Natural Language Processing for Digital Humanities (NLP4DH 2026)},
+  year={2026}
+}
+```
+
+## Paper Repository
+
+See the full paper, scripts, and results: [GitHub repository](https://github.com/[repository])

config.json

@@ -0,0 +1,37 @@
+{
+  "_name_or_path": "facebook/m2m100_418M",
+  "activation_dropout": 0.0,
+  "activation_function": "relu",
+  "architectures": [
+    "M2M100ForConditionalGeneration"
+  ],
+  "attention_dropout": 0.1,
+  "bos_token_id": 0,
+  "d_model": 1024,
+  "decoder_attention_heads": 16,
+  "decoder_ffn_dim": 4096,
+  "decoder_layerdrop": 0.05,
+  "decoder_layers": 12,
+  "decoder_start_token_id": 2,
+  "dropout": 0.1,
+  "early_stopping": true,
+  "encoder_attention_heads": 16,
+  "encoder_ffn_dim": 4096,
+  "encoder_layerdrop": 0.05,
+  "encoder_layers": 12,
+  "eos_token_id": 2,
+  "gradient_checkpointing": false,
+  "init_std": 0.02,
+  "is_encoder_decoder": true,
+  "max_length": 200,
+  "max_position_embeddings": 1024,
+  "model_type": "m2m_100",
+  "num_beams": 5,
+  "num_hidden_layers": 12,
+  "pad_token_id": 1,
+  "scale_embedding": true,
+  "torch_dtype": "float32",
+  "transformers_version": "4.44.0",
+  "use_cache": true,
+  "vocab_size": 128112
+}

generation_config.json

@@ -0,0 +1,11 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 0,
+  "decoder_start_token_id": 2,
+  "early_stopping": true,
+  "eos_token_id": 2,
+  "max_length": 200,
+  "num_beams": 5,
+  "pad_token_id": 1,
+  "transformers_version": "4.44.0"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c37c09c606c59e49b8f9aa214cc87d8f2e7ee9ab2c3dde5b0616ff00d74a9e1
+size 1935681888