Upload README.md

README.md

@@ -1,5 +1,54 @@
+
+This is the converted model from Unbabel/wmt23-cometkiwi-da
+
+1) Just kept the weights/bias keys()
+2) Renamed the keys to match the original Facebook/XLM-roberta-XL
+3) kept the layer_wise_attention / estimator layers
+
+Because of a hack in HF's code I had to rename the "layerwise_attention.gamma" key to "layerwise_attention.gam"
+
+I changed the config.json key "layer_transformation" from sparsemax to softmax because there is a bug in COMET since the flag is not passed, the actual function used is the default which is softmax.
+
+Usage:
+
+```
+from transformers import XLMRobertaTokenizer, XLMRobertaTokenizerFast, AutoModel
+tokenizer = XLMRobertaTokenizerFast.from_pretrained("vince62s/wmt23-cometkiwi-da-roberta-xl", trust_remote_code=True)
+model = AutoModel.from_pretrained("vince62s/wmt23-cometkiwi-da-roberta-xl", trust_remote_code=True)
+
+text = "Hello world!</s></s>Bonjour le monde"
+encoded_text = tokenizer(text, return_tensors='pt')
+print(encoded_text)
+output = model(**encoded_text)
+print(output[0])
+
+{'input_ids': tensor([[    0, 35378,  8999,    38,     2,     2, 84602,    95, 11146,     2]]), 'attention_mask': tensor([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1]])}
+tensor([[0.8217]], grad_fn=<AddmmBackward0>)
+
+```
+
+Let's double check with the original code from Unbabel Comet:
+
+```
+from comet import download_model, load_from_checkpoint
+model = load_from_checkpoint("/home/vincent/Downloads/cometkiwi23/checkpoints/model.ckpt") # this is the Unbabel checkpoint
+data = [{"mt": "Hello world!", "src": "Bonjour le monde"}]
+output = model.predict(data, gpus=0)
+print(output)
+
+Prediction([('scores', [0.8216837048530579]), ('system_score', 0.8216837048530579)])
+```
+
+
+
+
+
+
 ---
-language: 
+extra_gated_heading: Acknowledge license to accept the repository
+extra_gated_button_content: Acknowledge license
+pipeline_tag: translation
+language:
 - multilingual
 - af
 - am
@@ -61,7 +110,7 @@ language:
 - my
 - ne
 - nl
-- no
+- 'no'
 - om
 - or
 - pa
@@ -94,96 +143,78 @@ language:
 - xh
 - yi
 - zh
-license: mit
+license: cc-by-nc-sa-4.0
+library_name: transformers
 ---
 
-# XLM-RoBERTa-XL (xlarge-sized model) 
+This is a [COMET](https://github.com/Unbabel/COMET) quality estimation model: It receives a source sentence and the respective translation and returns a score that reflects the quality of the translation.
 
-XLM-RoBERTa-XL model pre-trained on 2.5TB of filtered CommonCrawl data containing 100 languages. It was introduced in the paper [Larger-Scale Transformers for Multilingual Masked Language Modeling](https://arxiv.org/abs/2105.00572) by Naman Goyal, Jingfei Du, Myle Ott, Giri Anantharaman, Alexis Conneau and first released in [this repository](https://github.com/pytorch/fairseq/tree/master/examples/xlmr). 
+# Paper
 
-Disclaimer: The team releasing XLM-RoBERTa-XL did not write a model card for this model so this model card has been written by the Hugging Face team.
+[CometKiwi: IST-Unbabel 2022 Submission for the Quality Estimation Shared Task](https://aclanthology.org/2022.wmt-1.60) (Rei et al., WMT 2022)
 
-## Model description
+# License:
 
-XLM-RoBERTa-XL is a extra large multilingual version of RoBERTa. It is pre-trained on 2.5TB of filtered CommonCrawl data containing 100 languages. 
+cc-by-nc-sa-4.0
 
-RoBERTa is a transformers model pretrained on a large corpus in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labeling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts.
+# Usage (unbabel-comet)
 
-More precisely, it was pretrained with the Masked language modeling (MLM) objective. Taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence.
+Using this model requires unbabel-comet to be installed:
 
-This way, the model learns an inner representation of 100 languages that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled sentences for instance, you can train a standard classifier using the features produced by the XLM-RoBERTa-XL model as inputs.
+```bash
+pip install --upgrade pip  # ensures that pip is current 
+pip install "unbabel-comet>=2.0.0"
+```
 
-## Intended uses & limitations
+Make sure you acknowledge its License and Log in into Hugging face hub before using:
 
-You can use the raw model for masked language modeling, but it's mostly intended to be fine-tuned on a downstream task. See the [model hub](https://huggingface.co/models?search=xlm-roberta-xl) to look for fine-tuned versions on a task that interests you.
+```bash
+huggingface-cli login
+# or using an environment variable
+huggingface-cli login --token $HUGGINGFACE_TOKEN
+```
 
-Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation, you should look at models like GPT2.
+Then you can use it through comet CLI:
 
-## Usage
+```bash
+comet-score -s {source-input}.txt -t {translation-output}.txt --model Unbabel/wmt22-cometkiwi-da
+```
 
-You can use this model directly with a pipeline for masked language modeling:
+Or using Python:
 
 ```python
->>> from transformers import pipeline
->>> unmasker = pipeline('fill-mask', model='facebook/xlm-roberta-xl')
->>> unmasker("Europe is a <mask> continent.")
-
-[{'score': 0.08562745153903961,
-  'token': 38043,
-  'token_str': 'living',
-  'sequence': 'Europe is a living continent.'},
- {'score': 0.0799778401851654,
-  'token': 103494,
-  'token_str': 'dead',
-  'sequence': 'Europe is a dead continent.'},
- {'score': 0.046154674142599106,
-  'token': 72856,
-  'token_str': 'lost',
-  'sequence': 'Europe is a lost continent.'},
- {'score': 0.04358183592557907,
-  'token': 19336,
-  'token_str': 'small',
-  'sequence': 'Europe is a small continent.'},
- {'score': 0.040570393204689026,
-  'token': 34923,
-  'token_str': 'beautiful',
-  'sequence': 'Europe is a beautiful continent.'}]
+from comet import download_model, load_from_checkpoint
+
+model_path = download_model("Unbabel/wmt22-cometkiwi-da")
+model = load_from_checkpoint(model_path)
+data = [
+    {
+        "src": "The output signal provides constant sync so the display never glitches.",
+        "mt": "Das Ausgangssignal bietet eine konstante Synchronisation, so dass die Anzeige nie stört."
+    },
+    {
+        "src": "Kroužek ilustrace je určen všem milovníkům umění ve věku od 10 do 15 let.",
+        "mt": "Кільце ілюстрації призначене для всіх любителів мистецтва у віці від 10 до 15 років."
+    },
+    {
+        "src": "Mandela then became South Africa's first black president after his African National Congress party won the 1994 election.",
+        "mt": "その後、1994年の選挙でアフリカ国民会議派が勝利し、南アフリカ初の黒人大統領となった。"
+    }
+]
+model_output = model.predict(data, batch_size=8, gpus=1)
+print (model_output)
 ```
 
-Here is how to use this model to get the features of a given text in PyTorch:
+# Intended uses
 
-```python
-from transformers import AutoTokenizer, AutoModelForMaskedLM
+Our model is intented to be used for **reference-free MT evaluation**. 
 
-tokenizer = AutoTokenizer.from_pretrained('facebook/xlm-roberta-xl')
-model = AutoModelForMaskedLM.from_pretrained("facebook/xlm-roberta-xl")
+Given a source text and its translation, outputs a single score between 0 and 1 where 1 represents a perfect translation. 
 
-# prepare input
-text = "Replace me by any text you'd like."
-encoded_input = tokenizer(text, return_tensors='pt')
+# Languages Covered:
 
-# forward pass
-output = model(**encoded_input)
-```
+This model builds on top of InfoXLM which cover the following languages:
+
+Afrikaans, Albanian, Amharic, Arabic, Armenian, Assamese, Azerbaijani, Basque, Belarusian, Bengali, Bengali Romanized, Bosnian, Breton, Bulgarian, Burmese, Burmese, Catalan, Chinese (Simplified), Chinese (Traditional), Croatian, Czech, Danish, Dutch, English, Esperanto, Estonian, Filipino, Finnish, French, Galician, Georgian, German, Greek, Gujarati, Hausa, Hebrew, Hindi, Hindi Romanized, Hungarian, Icelandic, Indonesian, Irish, Italian, Japanese, Javanese, Kannada, Kazakh, Khmer, Korean, Kurdish (Kurmanji), Kyrgyz, Lao, Latin, Latvian, Lithuanian, Macedonian, Malagasy, Malay, Malayalam, Marathi, Mongolian, Nepali, Norwegian, Oriya, Oromo, Pashto, Persian, Polish, Portuguese, Punjabi, Romanian, Russian, Sanskri, Scottish, Gaelic, Serbian, Sindhi, Sinhala, Slovak, Slovenian, Somali, Spanish, Sundanese, Swahili, Swedish, Tamil, Tamil Romanized, Telugu, Telugu Romanized, Thai, Turkish, Ukrainian, Urdu, Urdu Romanized, Uyghur, Uzbek, Vietnamese, Welsh, Western, Frisian, Xhosa, Yiddish.
 
-### BibTeX entry and citation info
-
-```bibtex
-@article{DBLP:journals/corr/abs-2105-00572,
-  author    = {Naman Goyal and
-               Jingfei Du and
-               Myle Ott and
-               Giri Anantharaman and
-               Alexis Conneau},
-  title     = {Larger-Scale Transformers for Multilingual Masked Language Modeling},
-  journal   = {CoRR},
-  volume    = {abs/2105.00572},
-  year      = {2021},
-  url       = {https://arxiv.org/abs/2105.00572},
-  eprinttype = {arXiv},
-  eprint    = {2105.00572},
-  timestamp = {Wed, 12 May 2021 15:54:31 +0200},
-  biburl    = {https://dblp.org/rec/journals/corr/abs-2105-00572.bib},
-  bibsource = {dblp computer science bibliography, https://dblp.org}
-}
-```
+Thus, results for language pairs containing uncovered languages are unreliable!