Upload folder using huggingface_hub

README.md

@@ -0,0 +1,161 @@
+---
+language: fr
+license: mit
+tags:
+- deberta-v2
+- token-classification
+base_model: almanach/camembertav2-base
+datasets:
+- Rhapsodie
+metrics:
+- las
+- upos
+model-index:
+- name: almanach/camembertav2-base-rhapsodie
+  results:
+  - task:
+      type: token-classification
+      name: Part-of-Speech Tagging
+    dataset:
+      type: Rhapsodie
+      name: Rhapsodie
+    metrics:
+    - name: upos
+      type: upos
+      value: 0.97769
+      verified: false
+  - task:
+      type: token-classification
+      name: Dependency Parsing
+    dataset:
+      type: Rhapsodie
+      name: Rhapsodie
+    metrics:
+    - name: las
+      type: las
+      value: 0.84562
+      verified: false
+---
+
+# Model Card for almanach/camembertav2-base-rhapsodie
+
+almanach/camembertav2-base-rhapsodie is a deberta-v2 model for token classification. It is trained on the Rhapsodie dataset for the task of Part-of-Speech Tagging and Dependency Parsing.
+ The model achieves an f1 score of  on the Rhapsodie dataset.
+
+The model is part of the almanach/camembertav2-base family of model finetunes.
+
+## Model Details
+
+### Model Description
+
+- **Developed by:** Wissam Antoun (Phd Student at Almanach, Inria-Paris)
+- **Model type:** deberta-v2
+- **Language(s) (NLP):** French
+- **License:** MIT
+- **Finetuned from model :** almanach/camembertav2-base
+
+### Model Sources
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** https://github.com/WissamAntoun/camemberta
+- **Paper:** https://arxiv.org/abs/2411.08868
+
+## Uses
+
+The model can be used for token classification tasks in French for Part-of-Speech Tagging and Dependency Parsing.
+
+## Bias, Risks, and Limitations
+
+The model may exhibit biases based on the training data. The model may not generalize well to other datasets or tasks. The model may also have limitations in terms of the data it was trained on.
+
+
+## How to Get Started with the Model
+
+You can use the models directly with the hopsparser library in server mode https://github.com/hopsparser/hopsparser/blob/main/docs/server.md
+
+
+## Training Details
+
+### Training Procedure
+
+Model trained with the [hopsparser](https://github.com/hopsparser/hopsparser) library on the Rhapsodie dataset.
+
+
+#### Training Hyperparameters
+
+```yml
+# Layer dimensions
+mlp_input: 1024
+mlp_tag_hidden: 16
+mlp_arc_hidden: 512
+mlp_lab_hidden: 128
+# Lexers
+lexers:
+  - name: word_embeddings
+    type: words
+    embedding_size: 256
+    word_dropout: 0.5
+  - name: char_level_embeddings
+    type: chars_rnn
+    embedding_size: 64
+    lstm_output_size: 128
+  - name: fasttext
+    type: fasttext
+  - name: camembertav2_base_p2_17k_last_layer
+    type: bert
+    model: /scratch/camembertv2/runs/models/camembertav2-base-bf16/post/ckpt-p2-17000/pt/discriminator/
+    layers: [11]
+    subwords_reduction: "mean"
+# Training hyperparameters
+encoder_dropout: 0.5
+mlp_dropout: 0.5
+batch_size: 8
+epochs: 64
+lr:
+  base: 0.00003
+  schedule:
+    shape: linear
+    warmup_steps: 100
+
+```
+
+#### Results
+
+**UPOS:** 0.97769
+**LAS:** 0.84562
+
+## Technical Specifications
+
+### Model Architecture and Objective
+
+deberta-v2 custom model for token classification.
+
+## Citation
+
+**BibTeX:**
+
+```bibtex
+@misc{antoun2024camembert20smarterfrench,
+      title={CamemBERT 2.0: A Smarter French Language Model Aged to Perfection},
+      author={Wissam Antoun and Francis Kulumba and Rian Touchent and Éric de la Clergerie and Benoît Sagot and Djamé Seddah},
+      year={2024},
+      eprint={2411.08868},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={https://arxiv.org/abs/2411.08868},
+}
+
+@inproceedings{grobol:hal-03223424,
+    title = {Analyse en dépendances du français avec des plongements contextualisés},
+    author = {Grobol, Loïc and Crabbé, Benoît},
+    url = {https://hal.archives-ouvertes.fr/hal-03223424},
+    booktitle = {Actes de la 28ème Conférence sur le Traitement Automatique des Langues Naturelles},
+    eventtitle = {TALN-RÉCITAL 2021},
+    venue = {Lille, France},
+    pdf = {https://hal.archives-ouvertes.fr/hal-03223424/file/HOPS_final.pdf},
+    hal_id = {hal-03223424},
+    hal_version = {v1},
+}
+
+```

camembertav2_base_p2_17k_last_layer.yaml

@@ -0,0 +1,32 @@
+# Layer dimensions
+mlp_input: 1024
+mlp_tag_hidden: 16
+mlp_arc_hidden: 512
+mlp_lab_hidden: 128
+# Lexers
+lexers:
+  - name: word_embeddings
+    type: words
+    embedding_size: 256
+    word_dropout: 0.5
+  - name: char_level_embeddings
+    type: chars_rnn
+    embedding_size: 64
+    lstm_output_size: 128
+  - name: fasttext
+    type: fasttext
+  - name: camembertav2_base_p2_17k_last_layer
+    type: bert
+    model: /scratch/camembertv2/runs/models/camembertav2-base-bf16/post/ckpt-p2-17000/pt/discriminator/
+    layers: [11]
+    subwords_reduction: "mean"
+# Training hyperparameters
+encoder_dropout: 0.5
+mlp_dropout: 0.5
+batch_size: 8
+epochs: 64
+lr:
+  base: 0.00003
+  schedule:
+    shape: linear
+    warmup_steps: 100

model/config.json

@@ -0,0 +1 @@
+{"mlp_input": 1024, "mlp_tag_hidden": 16, "mlp_arc_hidden": 512, "mlp_lab_hidden": 128, "biased_biaffine": true, "default_batch_size": 8, "encoder_dropout": 0.5, "extra_annotations": {}, "labels": ["acl", "acl:relcl", "advcl", "advcl:cleft", "advmod", "amod", "appos", "aux", "aux:caus", "aux:pass", "case", "cc", "ccomp", "compound", "conj", "cop", "csubj", "csubj:pass", "dep", "dep:comp", "det", "discourse", "dislocated", "expl:subj", "fixed", "flat", "iobj", "mark", "nmod", "nmod:appos", "nsubj", "nsubj:caus", "nsubj:pass", "nummod", "obj", "obj:lvc", "obl:agent", "obl:arg", "obl:mod", "parataxis", "parataxis:insert", "parataxis:parenth", "punct", "reparandum", "root", "vocative", "xcomp"], "mlp_dropout": 0.5, "tagset": ["ADJ", "ADP", "ADV", "AUX", "CCONJ", "DET", "INTJ", "NOUN", "NUM", "PRON", "PROPN", "PUNCT", "SCONJ", "VERB", "X"], "lexers": {"word_embeddings": "words", "char_level_embeddings": "chars_rnn", "fasttext": "fasttext", "camembertav2_base_p2_17k_last_layer": "bert"}, "multitask_loss": "sum"}

model/lexers/camembertav2_base_p2_17k_last_layer/config.json

@@ -0,0 +1 @@
+{"layers": [11], "subwords_reduction": "mean", "weight_layers": false}

model/lexers/camembertav2_base_p2_17k_last_layer/model/config.json

@@ -0,0 +1,41 @@
+{
+  "_name_or_path": "/scratch/camembertv2/runs/models/camembertav2-base-bf16/post/ckpt-p2-17000/pt/discriminator/",
+  "architectures": [
+    "DebertaV2Model"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 1,
+  "conv_act": "gelu",
+  "conv_kernel_size": 0,
+  "embedding_size": 768,
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-07,
+  "max_position_embeddings": 1024,
+  "max_relative_positions": -1,
+  "model_name": "camembertav2-base-bf16",
+  "model_type": "deberta-v2",
+  "norm_rel_ebd": "layer_norm",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "pooler_dropout": 0,
+  "pooler_hidden_act": "gelu",
+  "pooler_hidden_size": 768,
+  "pos_att_type": [
+    "p2c",
+    "c2p"
+  ],
+  "position_biased_input": false,
+  "position_buckets": 256,
+  "relative_attention": true,
+  "share_att_key": true,
+  "torch_dtype": "float32",
+  "transformers_version": "4.44.2",
+  "type_vocab_size": 0,
+  "vocab_size": 32768
+}

model/lexers/camembertav2_base_p2_17k_last_layer/model/special_tokens_map.json

@@ -0,0 +1,51 @@
+{
+  "bos_token": {
+    "content": "[CLS]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "cls_token": {
+    "content": "[CLS]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "[SEP]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "[MASK]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "[PAD]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sep_token": {
+    "content": "[SEP]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "[UNK]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

model/lexers/camembertav2_base_p2_17k_last_layer/model/tokenizer_config.json

@@ -0,0 +1,57 @@
+{
+  "add_prefix_space": true,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "[SEP]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "[UNK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "[MASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "[CLS]",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "eos_token": "[SEP]",
+  "errors": "replace",
+  "mask_token": "[MASK]",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "tokenizer_class": "RobertaTokenizer",
+  "trim_offsets": true,
+  "unk_token": "[UNK]"
+}

model/lexers/char_level_embeddings/config.json

@@ -0,0 +1 @@
+{"char_embeddings_dim": 64, "output_dim": 128, "special_tokens": ["<root>"], "charset": ["<pad>", "<special>", "!", "'", ",", "-", ".", ":", "?", "A", "B", "C", "D", "E", "F", "G", "H", "I", "J", "K", "L", "M", "N", "O", "P", "Q", "R", "S", "T", "U", "V", "W", "X", "Y", "Z", "a", "b", "c", "d", "e", "f", "g", "h", "i", "j", "k", "l", "m", "n", "o", "p", "q", "r", "s", "t", "u", "v", "w", "x", "y", "z", "~", "\u00c0", "\u00c9", "\u00e0", "\u00e2", "\u00e4", "\u00e7", "\u00e8", "\u00e9", "\u00ea", "\u00ee", "\u00ef", "\u00f4", "\u00f9", "\u00fb", "\u00fc", "\u0153", "\u2026"]}

model/lexers/fasttext/config.json

@@ -0,0 +1 @@
+{"special_tokens": ["<root>"]}

model/lexers/fasttext/fasttext_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e83b5b1c84a99cda52e5f1a5bcf1d389ead407063609663f3d6fd50a4104df7
+size 800313178

model/lexers/word_embeddings/config.json

@@ -0,0 +1 @@
+{"embeddings_dim": 256, "unk_word": "<unk>", "vocabulary": ["!", ",", "-ce", "-chose", "-il", "-ils", "-l\u00e0", "-moi", "-m\u00eame", "-t-il", "-vous", ".", ":", "<root>", "<unk>", "?", "A", "Abeba", "Abidal", "Addis", "Aden", "Afrique", "Agutte", "Ag\u00fcero", "Albert", "Aligre", "Alsace", "Alsace-Lorraine", "Al\u00e9sia", "Anelka", "Angleterre", "Antech", "Antoine", "Arabie", "Arago", "Argentin", "Argentins", "Avennes", "A~", "B", "Bachelard", "Barthes", "Bastille", "Belgique", "Bel~", "Besse", "Bixente", "Branly", "Brochet", "CCJ", "CNRS", "CRDT", "Cannes", "Carrisou", "Cavalaire", "Censier", "Chambre", "Champagne", "Champollion", "Champon", "Chancellerie", "Chaplin", "Charlie", "Charlot", "Chavant", "Chroniques", "Ch~", "Claudel", "Clauwaert", "Commerce", "Dati", "Dazs", "Derni\u00e8res", "Diarra", "Diderot", "Dominique", "Dubedout", "D\u00e9p\u00eache", "Eclair", "Edouard", "Emmanu~", "Erikson", "Ethiopie", "Eug\u00e8ne", "Europe", "Facebook", "Famibook", "Faubourg", "Ferran", "Figaro", "Fillon", "Fnac", "France", "Francis", "Fran\u00e7ais", "Fran\u00e7aises", "Fran\u00e7ois", "Gabi", "Gaga", "Gago", "Gallas", "Gambetta", "Gargia", "Gibert", "Gide", "Giesbert", "Gourcuff", "Grande", "Grenette", "Grenoble", "Guti\u00e9rrez", "Gu~", "G\u00e9n\u00e9ral", "Habitat", "Haruki", "Haute", "Heinze", "Henri", "Henry", "Hilton", "Homme", "Honneur", "Hubert", "Hugo", "H\u00e4agen", "I", "Industrie", "Inter", "Internet", "Jaur\u00e8s", "Jean", "Jean-Christophe", "Joffrin", "Journal", "Jules", "K", "Kafka", "Kenya", "Klein", "L", "L'", "La", "Lafayette", "Laurent", "Le", "Les", "Lesdigui\u00e8res", "Libre", "Lib\u00e9ration", "Lionel", "Loisirs", "Lyon", "L\u00e9gion", "MDE", "MacDo", "Maison", "Malraux", "Mandanda", "Maradona", "Marne", "Marsabit", "Marseille", "Mascherano", "Massimo", "Mauriac", "Maxi", "Messi", "Mex\u00e8s", "Midi", "Mitterrand", "Monfreid", "Mougeotte", "Mref", "Mulhouse", "Murakami", "M~", "Nairobi", "Nation", "Nef", "Nicolas", "Nietzsche", "Nord", "Notre-Dame", "Nouvelles", "Occident", "Olkaloo", "Ora~", "PCB", "Palais", "Papa", "Pardigon", "Paris", "Parisien", "Pascale", "Paul", "Peugeot", "Picpus", "Platon", "Premier", "Prisse", "Progr\u00e8s", "Proust", "Provence", "Quatre", "Quentin", "Rachida", "Reims", "Reuilly-Diderot", "Rey", "Rodriguez", "Roger", "Roland", "Rolex", "Royal", "R\u00e9publique", "Saint", "Saint-Germain", "Saint-Jean-de-Maurienne", "Sainte-Claire", "Sarce", "Sarkozy", "Sartre", "Saules", "Seine", "Sembat", "Shoah", "Small", "Smic", "Solf\u00e9rino", "Somalie", "Sorbonne", "Sor~", "Stendhal", "Steve", "Strasbourg", "S\u00e9gol\u00e8ne", "Tarantino", "Thierry", "UMP", "Union", "VIP", "Vallier", "Val\u00e9ry", "Vaucanson", "Victor", "Vict~", "Vincennes", "Voix", "Wittgenstein", "World", "XXX", "YYY", "Yves", "Z", "Zanetti", "Zep", "a", "abominable", "abord", "aborder", "absolument", "absorb\u00e9s", "abstentionnisme", "abstenu", "accepte", "acceptons", "accessible", "accident", "accomplir", "accord", "accorde", "accordent", "accrue", "acc\u00e8s", "acc\u00e9l\u00e8re", "acc\u00e9l\u00e9rer", "acharn\u00e9e", "achats", "achetait", "ach\u00e8tent", "ach\u00e8tes", "ach\u00e8ve", "acquis", "acte", "action", "activit\u00e9", "actualit\u00e9", "actuellement", "admettre", "administration", "adolescent", "adore", "adresser", "adversaire", "affaire", "affam\u00e9e", "affectation", "affirmation", "affirme", "affirmer", "africains", "agenda", "agit", "agr\u00e9able", "ah", "ai", "aide", "aider", "aid\u00e9", "aid\u00e9e", "ailleurs", "aime", "aim\u00e9", "air", "aisthesis", "ait", "albums", "aliments", "allais", "allait", "allant", "allante", "allemand", "aller", "allez", "allions", "all\u00e9e", "alors", "amateurs", "ambiance", "ambigu", "amenait", "amie", "amis", "amnistie", "amour", "am\u00e8ne", "am\u00e9ricain", "an", "anciens", "anecdote", "anglais", "angle", "anglo", "anime", "annexe", "anniversaire", "ann\u00e9e", "ann\u00e9es", "ans", "anti-litt\u00e9rature", "antiquit\u00e9", "aper\u00e7u", "appara\u00eet", "appareillage", "appareils", "apparemment", "appartements", "appartenez", "appartenir", "appartient", "appelait", "appeler", "appelle", "appellerai", "apprendre", "appris", "approche", "appr\u00e9hension", "appr\u00eate", "appuyer", "apr\u00e8s", "arbitre", "argent", "argentin", "argentine", "argentins", "armer", "arrach\u00e9", "arrive", "arrivent", "arriver", "arrives", "arrivez", "arriv\u00e9", "arriv\u00e9e", "arri~", "arri\u00e8re", "arri\u00e8re-grand-m\u00e8re", "arri\u00e8re-plan", "arr\u00eat", "arr\u00eate", "arr\u00eater", "art", "artisans", "artiste", "artistes", "artistique", "aspect", "aspects", "aspirations", "assauts", "assez", "association", "associations", "associer", "assurer", "atmosph\u00e9riquement", "attachement", "attaquants", "attaque", "attarderai", "attendez", "attends", "attention", "attroupe", "au-dessus", "aucune", "auditeurs", "augment\u00e9", "aujourd'hui", "aupr\u00e8s", "aura", "auraient", "aurais", "aurait", "aussi", "aus~", "automatiquement", "automatisme", "automatismes", "autorit\u00e9", "autour", "autre", "autrefois", "autrement", "autres", "autrichien", "au~", "avaient", "avais", "avait", "avanc\u00e9", "avanc\u00e9es", "avant", "avec", "avenir", "aventure", "avenue", "avertissement", "aveugles", "avez", "avoir", "avons", "avoue", "ayant", "ayez", "a~", "baguette", "baguettes", "bah", "balader", "baladeur", "balaye", "ballon", "bande", "banlieue", "banque", "baraquements", "bas\u00e9e", "battus", "beau", "beaucoup", "bel", "belles", "ben", "besoin", "bien", "biens", "bio", "bless\u00e9e", "bleu", "bling", "bois", "bon", "bonne", "bonnes", "bons", "boude", "bougez", "boulanger", "boulevard", "bouleverse", "boulot", "bourgeoise", "bourgeoisie", "bourrasque", "bout", "boutique", "bou\u00e9e", "bo\u00eete", "branch\u00e9", "branch\u00e9e", "branch\u00e9s", "bras", "brasseries", "bref", "brioches", "brouhaha", "brousse", "br\u00fblante", "bulles", "bureau", "bus", "but", "buteur", "b~", "b\u00e2timent", "b\u00e9n\u00e9ficient", "b\u00e9n\u00e9voles", "b\u00e9r\u00e9zina", "c'", "c'est-\u00e0-dire", "cabine", "cadre", "cadres", "caetera", "caf\u00e9s", "camion", "camionnette", "camp", "campagne", "cantonales", "cap", "capable", "capacit\u00e9", "capitales", "capter", "car", "caricature", "carrefour", "carri\u00e8re", "carton", "cartons", "cas", "casser", "cause", "ca~", "ce", "ceci", "cela", "celles", "celui", "cent", "centre", "centre-ville", "cents", "cercle", "cercles", "certain", "certaine", "certainement", "certaines", "certains", "ces", "cet", "cette", "ceux", "chacun", "chacune", "chaleur", "chaleureusement", "chambre", "champ", "chance", "change", "changeant", "changeants", "changement", "changements", "changer", "chang\u00e9", "chaque", "charg\u00e9s", "charni\u00e8re", "chass\u00e9s", "chats", "chaussures", "chef", "chemins", "cher", "cherche", "chercher", "chers", "chez", "chiffres", "choisi", "chose", "choses", "ch~", "ch\u00f4meurs", "ci", "ciment", "cinglant", "cinq", "cinquantaine", "cinquante", "cin\u00e9ma", "circonscription", "circuler", "citer", "citoyens", "civils", "clair", "classe", "clavier", "client\u00e8le", "client\u00e8les", "clonage", "club", "clubbers", "cl\u00e9s", "coalis\u00e9es", "code", "cognitive", "coh\u00e9sion", "coin", "coll\u00e8ges", "com", "combat", "combats", "combien", "commandait", "commanditaire", "comme", "commence", "commencerai", "commences", "commenc\u00e9", "comment", "commen\u00e7ait", "commercial", "commer\u00e7ants", "commettre", "commissaire", "commun", "communautaire", "communautaires", "communaut\u00e9", "communications", "communiquer", "compatriotes", "complexe", "compliqu\u00e9", "compl\u00e8tement", "comporte", "composante", "comprend", "comprendre", "comprends", "compris", "compte", "comp\u00e9n\u00e8trent", "concept", "conception", "concernent", "concitoyens", "concurrence", "conditions", "confiance", "confiant", "conflit", "conf\u00e9rence", "conf\u00e9rences", "conjuguer", "connais", "connaissaient", "connaissance", "connaissent", "connaissez", "connaissiez", "connaissons", "conna\u00eet", "conna\u00eetre", "connue", "conqu\u00eate", "conseillerais", "conseils", "conserve", "consid\u00e9rable", "consid\u00e9rer", "consid\u00e9r\u00e9es", "consiste", "consommateur", "constate", "construire", "construit", "cons\u00e9quent", "contemporain", "contemporains", "continue", "continuer", "continues", "continuez", "continu\u00e9", "conti~", "contourne", "contours", "contraire", "contrat", "contre", "contre-attaques", "contre-litt\u00e9rature", "contrer", "contresens", "contribue", "contribu\u00e9", "contr\u00e9", "contr\u00f4le", "convaincu", "convictions", "convulsions", "con\u00e7ue", "coordination", "corne", "corner", "corps", "correctement", "corrompront", "corromp~", "couleurs", "couloir", "coup", "couples", "coupure", "courant", "courir", "cours", "course", "courses", "court", "courte", "cousins", "couverte", "craignent", "criminels", "crise", "critique", "croire", "crois", "croiser", "crudit\u00e9", "cruellement", "crues", "cr\u00e9ation", "cr\u00e9er", "cuisine", "culture", "culturelle", "curieux", "cyber", "c~", "c\u00e9r\u00e9monies", "c\u00f4t\u00e9", "c\u00f4t\u00e9s", "c\u0153ur", "d'", "d'ailleurs", "d'autant", "dame", "dans", "date", "davantage", "de", "dehors", "demain", "demandant", "demande", "demand\u00e9", "demie", "depuis", "dep~", "dernier", "derniers", "derri\u00e8re", "des", "descend", "descends", "descriptif", "descriptions", "design", "dessin\u00e9e", "dessous", "destruction", "deux", "deuxi\u00e8me", "devant", "devants", "devanture", "devenait", "devenu", "devient", "devra", "devraient", "devrais", "de~", "diagonale", "dictatures", "difficile", "difficiles", "diffuser", "diff\u00e9rence", "diff\u00e9rentes", "diff\u00e9rents", "dignit\u00e9", "dimension", "dimensions", "dirais", "dire", "directe", "direction", "dirigeant", "diriger", "dis", "disaient", "disciple", "disciplin\u00e9", "discours", "dise", "disent", "disiez", "disons", "dispara\u00eetre", "disparu", "dispose", "distinction", "distinctions", "district", "dis~", "dit", "dites", "diverses", "diviser", "dix", "dix-huit", "dix-huiti\u00e8me", "dix-neuf", "dix-neuvi\u00e8me", "dix-septi\u00e8me", "docteur", "docteurs", "doctors", "dois", "doit", "doivent", "domaine", "dominait", "donc", "donne", "donner", "donnerons", "donn\u00e9", "donn\u00e9es", "dont", "dormaient", "dortoir", "doute", "douter", "douze", "droit", "droite", "droites", "droits", "du", "dur", "d~", "d\u00e8s", "d\u00e9bats", "d\u00e9but", "d\u00e9butant", "d\u00e9buts", "d\u00e9b~", "d\u00e9charger", "d\u00e9cha\u00eenement", "d\u00e9chiffrer", "d\u00e9cid\u00e9", "d\u00e9cid\u00e9ment", "d\u00e9cisions", "d\u00e9concentr\u00e9", "d\u00e9coration", "d\u00e9coratives", "d\u00e9cor\u00e9es", "d\u00e9crire", "d\u00e9fendre", "d\u00e9fense", "d\u00e9fensive", "d\u00e9ferlante", "d\u00e9figur\u00e9s", "d\u00e9finir", "d\u00e9finit", "d\u00e9finition", "d\u00e9fricher", "d\u00e9gage", "d\u00e9gagement", "d\u00e9garni", "d\u00e9gradent", "d\u00e9j\u00e0", "d\u00e9late", "d\u00e9licat", "d\u00e9lit", "d\u00e9mocratie", "d\u00e9mocratique", "d\u00e9montrer", "d\u00e9m\u00e9nagements", "d\u00e9nomm\u00e9", "d\u00e9nonce", "d\u00e9part", "d\u00e9pass\u00e9", "d\u00e9pass\u00e9s", "d\u00e9pend", "d\u00e9penses", "d\u00e9placements", "d\u00e9route", "d\u00e9sagr\u00e9gation", "d\u00e9saveu", "d\u00e9sert", "d\u00e9signe", "d\u00e9sir", "d\u00e9sormais", "d\u00e9tente", "d\u00e9truits", "d\u00e9velopp\u00e9e", "d\u00fb", "effectivement", "effet", "efforcerait", "eh", "elle", "elle-m\u00eame", "elles", "embarque", "emb\u00eatant", "emmen\u00e9", "emm\u00e9nagements", "empire", "emploi", "employ\u00e9", "emprise", "emprunte", "en", "encha\u00eene", "encore", "endeuill\u00e9es", "endormies", "endroit", "enfance", "enfant", "enfants", "enfin", "enfuit", "engagent", "enjeu", "ennui", "enseignants", "ensemble", "ensuite", "entendait", "entendre", "enthousiasme", "entraide", "entre", "entreprises", "entrer", "envie", "environ", "environnement", "envoy\u00e9", "erreur", "erreurs", "es", "esprit", "esp\u00e8ce", "esp\u00e9rance", "esp\u00e9rez", "essai", "essais", "essayer", "essentielles", "est", "esth\u00e9tique", "et", "eu", "euh", "eus", "eux", "exactement", "excellent", "exceptionnel", "except\u00e9", "exclamation", "excusez", "exemple", "exercice", "exerc\u00e9", "exigence", "exigences", "existe", "exist\u00e9", "explication", "explications", "explique", "expliquer", "exploser", "exprimer", "exp~", "exp\u00e9die", "exp\u00e9rience", "exp\u00e9rimente", "extraordinaire", "extr\u00eamement", "ex~", "ex\u00e9cutif", "fabrication", "fabrique", "fabriquer", "face", "facile", "faiblesse", "faire", "fais", "faisait", "faisant", "faisiez", "fait", "faites", "fallait", "fallu", "fameux", "familial", "famille", "familles", "fantaisie", "fantastique", "fassiez", "fatalit\u00e9", "faudrait", "fausse", "faut", "faute", "fautes", "fa\u00e7on", "femme", "femmes", "fen\u00eatres", "ferm\u00e9", "ferm\u00e9e", "ferm\u00e9s", "feront", "festival", "feux", "fff", "fid\u00e8le", "fiert\u00e9", "fig\u00e9", "fille", "fils", "fin", "finalement", "financier", "fini", "finissent", "fi\u00e8vre", "flagrant", "flics", "flou", "floue", "flying", "foi", "foire", "fois", "fonci\u00e8rement", "fonctions", "fond", "fondamentale", "font", "fontaine", "force", "forc\u00e9ment", "forme", "former", "formes", "formidable", "fort", "forte", "forts", "for\u00eats", "foss\u00e9", "foule", "fragilit\u00e9", "frais", "franc", "franchira", "franchit", "fran\u00e7ais", "fran\u00e7aise", "fran\u00e7aises", "frappe", "fraternel", "fraternelle", "friandises", "front", "fronti\u00e8re", "fr\u00e9quenter", "fut", "futur", "f~", "f\u00e9minins", "f\u00e9tiches", "gaillard", "galeries", "garantissent", "garantit", "garde", "gardons", "gare", "gar\u00e7on", "gauche", "gazon", "gens", "gifle", "globalement", "goulag", "gouvernement", "go\u00fbt\u00e9", "grand", "grande", "grandes", "grandir", "grands", "grands-parents", "gratuits", "grec", "gros", "gr\u00e8ve", "gr\u00eale", "guerre", "guerres", "guillemets", "g\u00e2teaux", "g\u00e9nie", "g\u00e9n\u00e9ral", "g\u00e9n\u00e9rale", "g\u00e9n\u00e9raliste", "g\u00e9n\u00e9ration", "g\u00e9n\u00e9rations", "g\u00e9n\u00e9raux", "g\u00e9n\u00e9tiques", "g\u00e9ographe", "habiller", "habill\u00e9e", "habit", "habitudes", "halle", "hasard", "haut", "haute", "hein", "heure", "heures", "heureuse", "heureusement", "heureux", "heurte", "heurtent", "hier", "histoire", "histoires", "historique", "historiques", "homme", "hommes", "honneur", "honteux", "horaire", "horaires", "horreur", "huit", "humaine", "humanit\u00e9", "humilit\u00e9", "humour", "h\u00e9donistes", "h\u00e9donistique", "h\u00e9do~", "h\u00e9ritiers", "h\u00e9sitations", "h\u00f4pital", "ici", "identit\u00e9", "id\u00e9e", "il", "ils", "image", "imagine", "immeuble", "immobilier", "imm\u00e9diat", "imm\u00e9diatement", "importance", "important", "importante", "importantes", "importe", "impose", "impression", "imprimantes", "inconscience", "incontournable", "incroyable", "indiqu\u00e9", "ind\u00e9boulonnable", "ind\u00e9pendant", "informaticien", "informatique", "informe", "infos", "initiative", "inquiets", "installent", "installer", "instant", "instants", "instaur\u00e9", "institution", "institutionnel", "institutions", "instruction", "intellectuels", "intenable", "international", "internet", "interpr\u00e9tations", "interroger", "intersection", "intervenir", "intervenu", "int\u00e9ressait", "int\u00e9ressant", "int\u00e9ressants", "int\u00e9ress\u00e9e", "int\u00e9rieur", "int\u00e9rieurs", "int\u00e9r\u00eat", "inventer", "invitation", "invitent", "invit\u00e9e", "irr\u00e9elles", "itin\u00e9raires", "j'", "jalonnent", "jamais", "jardin", "jardins", "je", "jet", "jeu", "jeune", "jeunes", "jeunesse", "jeu~", "jouaient", "joue", "jouer", "joueur", "joueurs", "jouissance", "jour", "journaux", "journ\u00e9e", "jours", "jou\u00e9", "joyeux", "jugement", "jusqu'", "juste", "justement", "justice", "j~", "kilom\u00e8tres", "kiosque", "l'", "l'on", "la", "laboratoire", "laisse", "langage", "langue", "laquelle", "larcin", "large", "largement", "le", "leaders", "leadership", "lecteur", "lecture", "lequel", "les", "leur", "leurs", "le\u00e7on", "libert\u00e9", "libert\u00e9s", "libre", "licenciement", "lieu", "ligne", "lignes", "limites", "lire", "lisais", "lisez", "lit", "litigieuse", "littoral", "litt\u00e9raire", "litt\u00e9ralement", "litt\u00e9rature", "litt\u00e9r~", "livraison", "livre", "livres", "li\u00e9", "li\u00e9e", "li\u00e9es", "li\u00e9s", "local", "location", "logiciels", "loin", "lointain", "loisir", "long", "longe", "longer", "longez", "longtemps", "lorgne", "lorsqu'", "lorsque", "louper", "loyers", "lu", "lui", "lui-m\u00eame", "lus", "lyc\u00e9e", "lyc\u00e9en", "lyc\u00e9es", "l~", "l\u00e0", "l\u00e0-bas", "l\u00e0-dedans", "l\u00e0-dessus", "l\u00e2chent", "l\u00e2cher", "l\u00e2ch\u00e9", "l\u00e8guerons", "l\u00e9g\u00e8ret\u00e9", "m'", "ma", "machin", "machine", "magasin", "magnifique", "main", "maintenance", "maintenant", "maintenan~", "maintenir", "maint~", "maires", "mais", "majorit\u00e9", "mal", "malade", "maladie", "maladive", "malheur", "malheureusement", "mang\u00e9", "manifestations", "manipulations", "mani\u00e8re", "manque", "manqu\u00e9", "man\u00e8ge", "man\u0153uvre", "marchandises", "marche", "march\u00e9", "mari\u00e9e", "marquage", "marque", "marqu\u00e9", "marrant", "masculins", "masques", "match", "matchs", "maternelles", "matin", "mati\u00e8re", "mat\u00e9rielles", "mauvais", "mauvaise", "ma~", "ma\u00eetrisait", "ma\u00eetris\u00e9e", "me", "meilleures", "membres", "men\u00e9e", "merci", "mes", "message", "mesure", "mesures", "mesurons", "mettre", "mh", "mi-temps", "micro-processeurs", "midi", "mien", "mieux", "milieu", "militaires", "mille", "millions", "mine", "ministre", "minute", "minutes", "miraculeux", "mise", "mises", "mobile", "mobilisation", "mobilis\u00e9s", "moderne", "modernit\u00e9", "modifier", "moi", "moins", "moment", "moments", "mon", "monde", "mondiales", "mondialisation", "monsieur", "monter", "montes", "montre", "montrer", "monuments", "morbide", "morceau", "mort", "mortels", "mor~", "mot", "motif", "motivations", "mourir", "mouton", "mouvements", "moyenne", "moyennes", "moyens", "mule", "mulot", "municipales", "musicale", "musique", "mus\u00e9e", "mythe", "m~", "m\u00e8ne", "m\u00e8nent", "m\u00e8tres", "m\u00e8~", "m\u00e9decin", "m\u00e9decine", "m\u00e9decins", "m\u00e9diane", "m\u00e9diateur", "m\u00e9lange", "m\u00e9moire", "m\u00e9tier", "m\u00e9tiers", "m\u00e9tro", "m\u00e9tropole", "m\u00e9t~", "m\u00e9~", "m\u00eale", "m\u00eame", "n'", "nation", "national", "nations", "naturel", "naturellement", "naturels", "naufrage", "na\u00eetre", "na\u00effs", "ne", "neuf", "ni", "niche", "nier", "niveau", "ni~", "noir", "nom", "nombre", "nombreuses", "nombreux", "nomination", "noms", "non", "nord", "normalement", "nos", "notamment", "notion", "notre", "nous", "nouveau", "nouveaux", "nouvel", "nouvelle", "nouvelles", "nuages", "nuanc\u00e9e", "nue", "nul", "num\u00e9rique", "n\u00e9anmoins", "n\u00e9cessaire", "n\u00e9e", "n\u00e9gation", "n\u00e9gociations", "n\u00e9s", "obligatoire", "oblig\u00e9e", "oblig\u00e9s", "obstacles", "obstinent", "obstruction", "obs\u00e8de", "occasion", "occupe", "occup\u00e9es", "offensif", "officiers", "oh", "oiseau", "on", "ont", "oppose", "opposer", "optimiste", "or", "ordinateur", "ordinateurs", "ordre", "ordres", "organiser", "organis\u00e9e", "orientent", "orient\u00e9es", "orient\u00e9s", "oser", "ou", "ouais", "oublie", "oubli\u00e9", "ouest", "ouh", "oui", "outrage", "outre-mer", "ouverte", "ouverture", "ouvrir", "o\u00f9", "page", "pain", "paix", "par", "paraissait", "para\u00eet", "para\u00eetre", "parc", "parce", "pardon", "parents", "parfois", "parisienne", "parlais", "parlant", "parle", "parler", "parlerai", "parlez", "parl\u00e9", "parmi", "part", "partageons", "partager", "partant", "partent", "partez", "parti", "participation", "participer", "particulier", "particuli\u00e8rement", "partie", "partiels", "partir", "partis", "pas", "passage", "passait", "passant", "passa~", "passe", "passent", "passer", "passerelle", "passes", "passez", "passion", "passionnant", "pass\u00e9", "patrimoine", "patron", "pauvre", "pauvret\u00e9", "pays", "peine", "peinture", "penalty", "pendant", "pensais", "pensait", "pensant", "pensante", "pense", "penser", "penseur", "pensez", "pensum", "pens\u00e9e", "pens\u00e9es", "pen~", "people", "peoples", "perdre", "perdu", "performance", "permanence", "permet", "permettent", "personnage", "personnages", "perte", "peser", "petit", "petite", "petites", "petits", "peu", "peuples", "peur", "peut", "peut-\u00eatre", "peut~", "peuvent", "peux", "pff", "philo", "philosophe", "philosophes", "philosophie", "philosophique", "photos", "phrase", "phrases", "ph\u00e9nom\u00e8ne", "pied", "pieds", "piquer", "piscine", "pi\u00e8ge", "place", "places", "plaisaient", "plaisir", "plant\u00e9", "plan\u00e8te", "plateau", "plateaux", "plein", "plong\u00e9", "plupart", "plus", "plusieurs", "plut\u00f4t", "point", "points", "poitrine", "polarisaient", "police", "policier", "policiers", "politique", "politiques", "pompiers", "ponctuel", "ponctuels", "porche", "portes", "pose", "poser", "posez", "possession", "poss\u00e9dait", "poste", "postes", "poteau", "pour", "pourquoi", "pourra", "pourrais", "pourrait", "pourriez", "poursuivez", "poursuivre", "pourtant", "pouss\u00e9", "pouvais", "pouvait", "pouvez", "pouvoir", "po\u00e9sie", "po\u00ef\u00e9sis", "pratiquant", "pratique", "pratiquement", "pratiquez", "pra~", "premier", "premi\u00e8re", "prend", "prendre", "prendront", "prends", "presqu'", "presque", "presse", "preuve", "primaire", "principal", "principalement", "priori", "pris", "prise", "privil\u00e9giait", "privil\u00e9gi\u00e9s", "priv\u00e9e", "prix", "probl\u00e8me", "probl\u00e8mes", "prochainement", "proches", "production", "productivit\u00e9", "produit", "professeurs", "profite", "profiter", "profond\u00e9ment", "programmation", "programme", "programmes", "progr\u00e8s", "projets", "prolonge", "prolongement", "promener", "pronostique", "propos", "proposons", "propos\u00e9s", "propre", "protestation", "province", "pr\u00e8s", "pr\u00e9alable", "pr\u00e9carit\u00e9", "pr\u00e9cieux", "pr\u00e9cipite", "pr\u00e9cise", "pr\u00e9cision", "pr\u00e9cis\u00e9ment", "pr\u00e9c\u00e9dent", "pr\u00e9d\u00e9cesseurs", "pr\u00e9fecture", "pr\u00e9fet", "pr\u00e9gnante", "pr\u00e9liminaires", "pr\u00e9occupent", "pr\u00e9pa", "pr\u00e9sent", "pr\u00e9sentes", "pr\u00e9sident", "pr\u00e9sidentielle", "pr\u00e9vient", "pr\u00e9voir", "pr\u00e9vu", "pu", "public", "publics", "publiques", "pudeur", "pudique", "puis", "puisqu'", "puisque", "puissamment", "puissance", "puisse", "p~", "p\u00e2tisserie", "p\u00e2tissier", "p\u00e8re", "p\u00e8sent", "p\u00e9riph\u00e9riques", "qu'", "quai", "quais", "quand", "quarante", "quarante-huit", "quarante-neuf", "quaranti\u00e8me", "quartier", "quartiers", "quar~", "quatorze", "quatorzi\u00e8me", "quatre", "quatre-vingt-dix-neuf", "quatre-vingt-onze", "quatre-vingt-six", "que", "quel", "quelle", "quelqu'un", "quelque", "quelquefois", "quelques", "question", "qui", "quinze", "quoi", "quoique", "quotidienne", "racont\u00e9", "rails", "raison", "rappelez", "rapport", "rapproche", "rassembl\u00e9e", "ratatin\u00e9", "rattraper", "rayon", "rebondit", "recevoir", "recherche", "recherch\u00e9e", "reconnaissez", "reconna\u00eet", "reconna\u00eetre", "records", "recueil", "reculer", "redis", "reflet", "refus\u00e9es", "regarde", "regarder", "regardez", "rejoindre", "rejoint", "relativement", "religieuse", "religieuses", "relire", "relis", "rel\u00e2chement", "rel\u00e8vement", "remarque", "remercie", "remis", "remontant", "remonte", "remonter", "remontez", "rempla\u00e7ante", "renaissance", "rencontre", "rencontrent", "rencontres", "rencontr\u00e9", "rencontr\u00e9s", "rendez", "rendre", "renoncement", "rentre", "rentrent", "rentrer", "rentres", "rentr\u00e9", "rentr\u00e9e", "rentr\u00e9s", "renverse", "repartie", "repartis", "repensant", "repos", "repousse", "repousser", "reprends", "reprenez", "repris", "repr\u00e9senter", "rep\u00e9rer", "respect", "respecter", "respectueuse", "responsabilit\u00e9", "responsabilit\u00e9s", "responsable", "responsables", "ressemble", "ressentie", "ressentir", "ressentons", "ressort", "restaient", "restant", "reste", "restent", "rest\u00e9e", "retombez", "retour", "retournent", "retourner", "retrait", "retrouve", "retrouver", "revenant", "revenir", "revenu", "revenue", "reviennent", "reviens", "revient", "re\u00e7oit", "rh\u00e9torique", "ricanement", "riche", "ridicule", "rien", "rivage", "romans", "romantique", "rond-point", "rose", "route", "rue", "rues", "rythme", "r\u00e8gle", "r\u00e8gles", "r\u00e9action", "r\u00e9aliser", "r\u00e9cemment", "r\u00e9cents", "r\u00e9ception", "r\u00e9ciproquement", "r\u00e9cup\u00e9rer", "r\u00e9cup\u00e9r\u00e9", "r\u00e9flexion", "r\u00e9forme", "r\u00e9formes", "r\u00e9f\u00e9rence", "r\u00e9gion", "r\u00e9gions", "r\u00e9gl\u00e9", "r\u00e9habiliter", "r\u00e9paration", "r\u00e9pond", "r\u00e9pondre", "r\u00e9ponse", "r\u00e9publicain", "r\u00e9seaux", "r\u00e9serv\u00e9", "r\u00e9sultat", "r\u00e9sumer", "r\u00e9tablir", "r\u00e9ussirons", "r\u00eaver", "r\u00eaveur", "r\u00f4le", "s'", "sa", "sacr\u00e9es", "sais", "sait", "salaires", "sanction", "sanctionn\u00e9e", "sandwich", "sans", "sant\u00e9", "sauf", "saura", "saute", "sauvage", "sauvagement", "sauve", "sauvetage", "savaient", "savez", "savoir", "savons", "sav~", "sa~", "science", "scientifique", "scolaire", "scolarit\u00e9", "scrutin", "sc\u00e8ne", "se", "second", "seconde", "seigneur", "seizi\u00e8me", "selon", "semaines", "semblait", "semblant", "semble", "semblerait", "semblez", "sens", "sensation", "sentiment", "sen~", "sept", "septi\u00e8me", "sera", "serait", "seras", "serez", "serons", "serveur", "service", "services", "ses", "set", "seul", "seule", "seulement", "sexe", "sexualit\u00e9", "se~", "si", "siffl\u00e9", "signale", "signes", "sillonnant", "simple", "simplement", "singulier", "sinon", "site", "sites", "situation", "situe", "situer", "sixi\u00e8me", "si\u00e8cle", "si\u00e8cles", "social", "sociale", "socialiste", "sociaux", "sociologie", "soci\u00e9t\u00e9", "soci\u00e9t\u00e9s", "soi", "soient", "soir", "soit", "soixante-cinq", "soixante-quatre", "soixante-quinze", "sol", "soleil", "solidaire", "solidarit\u00e9", "solidarit\u00e9s", "solide", "solution", "sommes", "son", "sont", "sors", "sort", "sorte", "sorti", "sortie", "sortir", "soucis", "souffrance", "souffrent", "souhaite", "soulignent", "souligner", "souris", "sous", "soutien", "souvenez", "souvenir", "souvenirs", "souvent", "souviens", "sovi\u00e9tique", "spectateur", "spontan\u00e9", "sport", "sp\u00e9cialement", "sp\u00e9cialiste", "sp\u00e9cialis\u00e9", "sp\u00e9cialis\u00e9s", "square", "stade", "stationner", "stations", "statut", "stopp\u00e9", "stressant", "su", "successives", "succ\u00e8s", "sud", "suffit", "suffrage", "suis", "suite", "suivez", "suivi", "suivre", "sujet", "sur", "surface", "surnaturelles", "surtout", "su~", "symbolis\u00e9", "sympathie", "syst\u00e8me", "s~", "s\u00e9curit\u00e9", "s\u00e9duction", "s\u00e9parable", "s\u00e9quence", "s\u00e9rieux", "s\u00fbr", "s\u00fbre", "s\u00fbrement", "tandis", "tant", "tard", "tchater", "techniquement", "tel", "telle", "tellement", "telles", "tels", "temps", "temp\u00eate", "tendance", "tendu", "tendue", "tenir", "tentent", "tenu", "terme", "termine", "terrasses", "terre", "territoriale", "test", "th\u00e9matiques", "th\u00e9oriciens", "th\u00e9orie", "th\u00e9oris\u00e9", "tienne", "tient", "tir", "tire", "tirer", "tissu", "toi", "tol\u00e9rance", "tombe", "tombent", "tomber", "tomb\u00e9e", "tontons", "tornade", "totalement", "totalitaires", "touche", "toucher", "touchons", "touch\u00e9s", "toujours", "tour", "tourisme", "tourments", "tournait", "tourne", "tourner", "tournes", "tournez", "tous", "tout", "toute", "toutes", "tou~", "trace", "traces", "tradition", "trag\u00e9dies", "train", "traiter", "traiteur", "trajet", "tram", "trams", "tranquille", "transform\u00e9", "transporte", "transports", "travail", "travaillais", "travaille", "travailler", "travaillez", "travaill\u00e9", "travaux", "travers", "traverse", "traverser", "traverses", "traversez", "treize", "trente", "trente-deux", "trente-et-un", "trente-neuf", "triomphe", "triste", "trois", "troisi\u00e8me", "tromper", "trop", "trottoir", "trou", "trouve", "trouver", "trouverai", "trouv\u00e9", "truc", "trucs", "tr\u00e8s", "tu", "type", "t~", "t\u00e2che", "t\u00e9l\u00e9phone", "t\u00e9l\u00e9ph\u00e9rique", "t\u00e9l\u00e9vis\u00e9", "t\u00eate", "t\u00f4t", "un", "une", "unie", "uniquement", "univers", "universel", "uns", "usine", "utile", "utilisateur", "utilisateurs", "utilise", "utiliser", "utilises", "utilis\u00e9", "u~", "va", "vacances", "vachement", "vague", "vais", "valeurs", "variations", "varie", "vari\u00e9", "vas", "vaut", "venait", "venez", "venir", "vente", "venue", "verra", "verras", "verrez", "verront", "vers", "version", "vert", "vertueuse", "veulent", "veut", "veux", "victimes", "victoire", "vid\u00e9os", "vie", "vieil", "vieille", "vieillissent", "viennent", "viens", "vient", "vif", "ville", "villes", "vingt", "vingt-cinq", "vingt-deux", "vingt-deuxi\u00e8me", "vingt-et-uni\u00e8me", "vingt-neuf", "vingt-trois", "vingt~", "visiblement", "visualise", "vite", "vitrine", "vitrines", "vivant", "vivante", "vivent", "vivons", "vivre", "voies", "voil\u00e0", "voir", "vois", "voisin", "voisins", "voit", "voiture", "voitures", "vol", "vole", "voler", "volontaire", "volont\u00e9", "vol\u00e9", "vont", "vos", "vote", "votre", "voudrais", "voudrait", "voudrons", "voulais", "voulait", "voulez", "vouloir", "voulu", "vous", "vous-m\u00eame", "voyaient", "voyait", "voyez", "voyons", "vrai", "vraiment", "vu", "vue", "vues", "v~", "v\u00e9cu", "v\u00e9ritable", "v\u00e9rit\u00e9", "y", "yeux", "ziki", "z\u00e9ro", "\u00c0", "\u00c9tat", "\u00c9tienne", "\u00c9toile", "\u00e0", "\u00e2g\u00e9e", "\u00e2me", "\u00e7a", "\u00e7~", "\u00e9bahie", "\u00e9bauche", "\u00e9carte", "\u00e9changer", "\u00e9chapper", "\u00e9cho", "\u00e9choient", "\u00e9clairages", "\u00e9clectique", "\u00e9cole", "\u00e9coles", "\u00e9crasante", "\u00e9crire", "\u00e9cris", "\u00e9crit", "\u00e9crite", "\u00e9criture", "\u00e9crivain", "\u00e9crivains", "\u00e9crivant", "\u00e9crivent", "\u00e9ditorial", "\u00e9ducation", "\u00e9galement", "\u00e9galisation", "\u00e9gards", "\u00e9glise", "\u00e9lan", "\u00e9lecteurs", "\u00e9lectorales", "\u00e9lectronique", "\u00e9loge", "\u00e9lus", "\u00e9l\u00e9ments", "\u00e9minemment", "\u00e9motion", "\u00e9mu", "\u00e9nergiquement", "\u00e9nerv\u00e9", "\u00e9ner~", "\u00e9norme", "\u00e9norm\u00e9ment", "\u00e9poque", "\u00e9poques", "\u00e9pouvantable", "\u00e9pouvantablement", "\u00e9preuve", "\u00e9prouve", "\u00e9prouver", "\u00e9puise", "\u00e9quateur", "\u00e9quipe", "\u00e9taient", "\u00e9tais", "\u00e9tait", "\u00e9tant", "\u00e9thiopien", "\u00e9thique", "\u00e9tiez", "\u00e9tonnement", "\u00e9trange", "\u00e9tranger", "\u00e9trangers", "\u00e9tudes", "\u00e9tudiant", "\u00e9tu~", "\u00e9t~", "\u00e9t\u00e9", "\u00e9videmment", "\u00e9vidence", "\u00e9vident", "\u00e9viter", "\u00e9volution", "\u00e9voqu\u00e9", "\u00e9v\u00e8nements", "\u00e9~", "\u00eates", "\u00eatre", "\u00eele", "\u0153uvre", "\u0153uvres", "\u2026"], "word_dropout": 0.5}

model/weights.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5dbece0dcb3522f0cce9de74d3847d9b1d9e6f8b4f4951892b62400a2450232
+size 1739319276

train.log

@@ -0,0 +1,112 @@
+[hops] 2024-09-23 16:26:02.103 | INFO     | Initializing a parser from /workspace/configs/exp_camembertv2/camembertav2_base_p2_17k_last_layer.yaml
+[hops] 2024-09-23 16:26:02.119 | INFO     | Generating a FastText model from the treebank
+[hops] 2024-09-23 16:26:02.124 | INFO     | Training fasttext model
+[hops] 2024-09-23 16:26:07.221 | INFO     | Start training on cuda:0
+[hops] 2024-09-23 16:26:07.224 | WARNING  | You're using a RobertaTokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.
+[hops] 2024-09-23 16:26:17.032 | INFO     | Epoch 0: train loss 3.0198	dev loss 2.5813	dev tag acc 06.78%	dev head acc 24.04%	dev deprel acc 34.03%
+[hops] 2024-09-23 16:26:17.033 | INFO     | New best model: head accuracy 24.04% > 0.00%
+[hops] 2024-09-23 16:26:28.781 | INFO     | Epoch 1: train loss 2.3534	dev loss 1.8865	dev tag acc 16.02%	dev head acc 34.55%	dev deprel acc 58.81%
+[hops] 2024-09-23 16:26:28.782 | INFO     | New best model: head accuracy 34.55% > 24.04%
+[hops] 2024-09-23 16:26:41.177 | INFO     | Epoch 2: train loss 1.8324	dev loss 1.4990	dev tag acc 33.81%	dev head acc 51.15%	dev deprel acc 68.77%
+[hops] 2024-09-23 16:26:41.178 | INFO     | New best model: head accuracy 51.15% > 34.55%
+[hops] 2024-09-23 16:26:53.654 | INFO     | Epoch 3: train loss 1.4805	dev loss 1.2183	dev tag acc 60.65%	dev head acc 64.19%	dev deprel acc 77.73%
+[hops] 2024-09-23 16:26:53.655 | INFO     | New best model: head accuracy 64.19% > 51.15%
+[hops] 2024-09-23 16:27:05.915 | INFO     | Epoch 4: train loss 1.2220	dev loss 1.0230	dev tag acc 64.64%	dev head acc 70.79%	dev deprel acc 81.67%
+[hops] 2024-09-23 16:27:05.916 | INFO     | New best model: head accuracy 70.79% > 64.19%
+[hops] 2024-09-23 16:27:18.205 | INFO     | Epoch 5: train loss 1.0437	dev loss 0.9197	dev tag acc 71.84%	dev head acc 73.47%	dev deprel acc 83.84%
+[hops] 2024-09-23 16:27:18.206 | INFO     | New best model: head accuracy 73.47% > 70.79%
+[hops] 2024-09-23 16:27:30.479 | INFO     | Epoch 6: train loss 0.9163	dev loss 0.8243	dev tag acc 76.67%	dev head acc 76.32%	dev deprel acc 85.66%
+[hops] 2024-09-23 16:27:30.480 | INFO     | New best model: head accuracy 76.32% > 73.47%
+[hops] 2024-09-23 16:27:43.020 | INFO     | Epoch 7: train loss 0.8055	dev loss 0.7605	dev tag acc 77.97%	dev head acc 77.86%	dev deprel acc 86.18%
+[hops] 2024-09-23 16:27:43.021 | INFO     | New best model: head accuracy 77.86% > 76.32%
+[hops] 2024-09-23 16:27:55.200 | INFO     | Epoch 8: train loss 0.7197	dev loss 0.6791	dev tag acc 80.11%	dev head acc 80.47%	dev deprel acc 87.75%
+[hops] 2024-09-23 16:27:55.201 | INFO     | New best model: head accuracy 80.47% > 77.86%
+[hops] 2024-09-23 16:28:07.158 | INFO     | Epoch 9: train loss 0.6403	dev loss 0.6609	dev tag acc 82.00%	dev head acc 81.19%	dev deprel acc 88.10%
+[hops] 2024-09-23 16:28:07.159 | INFO     | New best model: head accuracy 81.19% > 80.47%
+[hops] 2024-09-23 16:28:20.202 | INFO     | Epoch 10: train loss 0.5750	dev loss 0.6266	dev tag acc 86.48%	dev head acc 81.53%	dev deprel acc 88.46%
+[hops] 2024-09-23 16:28:20.203 | INFO     | New best model: head accuracy 81.53% > 81.19%
+[hops] 2024-09-23 16:28:32.873 | INFO     | Epoch 11: train loss 0.5237	dev loss 0.5865	dev tag acc 88.83%	dev head acc 82.58%	dev deprel acc 88.99%
+[hops] 2024-09-23 16:28:32.874 | INFO     | New best model: head accuracy 82.58% > 81.53%
+[hops] 2024-09-23 16:28:44.747 | INFO     | Epoch 12: train loss 0.4711	dev loss 0.6140	dev tag acc 91.18%	dev head acc 83.19%	dev deprel acc 88.91%
+[hops] 2024-09-23 16:28:44.748 | INFO     | New best model: head accuracy 83.19% > 82.58%
+[hops] 2024-09-23 16:28:57.334 | INFO     | Epoch 13: train loss 0.4285	dev loss 0.5683	dev tag acc 92.55%	dev head acc 83.75%	dev deprel acc 89.79%
+[hops] 2024-09-23 16:28:57.335 | INFO     | New best model: head accuracy 83.75% > 83.19%
+[hops] 2024-09-23 16:29:09.635 | INFO     | Epoch 14: train loss 0.3936	dev loss 0.5409	dev tag acc 93.76%	dev head acc 84.73%	dev deprel acc 90.03%
+[hops] 2024-09-23 16:29:09.636 | INFO     | New best model: head accuracy 84.73% > 83.75%
+[hops] 2024-09-23 16:29:21.930 | INFO     | Epoch 15: train loss 0.3572	dev loss 0.5503	dev tag acc 94.38%	dev head acc 84.95%	dev deprel acc 89.94%
+[hops] 2024-09-23 16:29:21.931 | INFO     | New best model: head accuracy 84.95% > 84.73%
+[hops] 2024-09-23 16:29:34.223 | INFO     | Epoch 16: train loss 0.3330	dev loss 0.5319	dev tag acc 95.01%	dev head acc 85.68%	dev deprel acc 90.38%
+[hops] 2024-09-23 16:29:34.224 | INFO     | New best model: head accuracy 85.68% > 84.95%
+[hops] 2024-09-23 16:29:46.529 | INFO     | Epoch 17: train loss 0.3041	dev loss 0.5143	dev tag acc 95.49%	dev head acc 85.63%	dev deprel acc 90.86%
+[hops] 2024-09-23 16:29:56.855 | INFO     | Epoch 18: train loss 0.2777	dev loss 0.5152	dev tag acc 95.62%	dev head acc 86.44%	dev deprel acc 90.64%
+[hops] 2024-09-23 16:29:56.856 | INFO     | New best model: head accuracy 86.44% > 85.68%
+[hops] 2024-09-23 16:30:09.439 | INFO     | Epoch 19: train loss 0.2620	dev loss 0.5347	dev tag acc 96.02%	dev head acc 86.35%	dev deprel acc 90.86%
+[hops] 2024-09-23 16:30:20.466 | INFO     | Epoch 20: train loss 0.2412	dev loss 0.5382	dev tag acc 96.25%	dev head acc 86.75%	dev deprel acc 91.09%
+[hops] 2024-09-23 16:30:20.467 | INFO     | New best model: head accuracy 86.75% > 86.44%
+[hops] 2024-09-23 16:30:32.804 | INFO     | Epoch 21: train loss 0.2277	dev loss 0.5043	dev tag acc 96.67%	dev head acc 87.73%	dev deprel acc 91.17%
+[hops] 2024-09-23 16:30:32.805 | INFO     | New best model: head accuracy 87.73% > 86.75%
+[hops] 2024-09-23 16:30:45.266 | INFO     | Epoch 22: train loss 0.2125	dev loss 0.5428	dev tag acc 96.52%	dev head acc 87.02%	dev deprel acc 91.30%
+[hops] 2024-09-23 16:30:55.802 | INFO     | Epoch 23: train loss 0.1977	dev loss 0.5209	dev tag acc 96.56%	dev head acc 87.84%	dev deprel acc 91.54%
+[hops] 2024-09-23 16:30:55.803 | INFO     | New best model: head accuracy 87.84% > 87.73%
+[hops] 2024-09-23 16:31:08.582 | INFO     | Epoch 24: train loss 0.1878	dev loss 0.5292	dev tag acc 96.89%	dev head acc 87.50%	dev deprel acc 91.75%
+[hops] 2024-09-23 16:31:19.329 | INFO     | Epoch 25: train loss 0.1775	dev loss 0.5573	dev tag acc 96.98%	dev head acc 87.34%	dev deprel acc 91.44%
+[hops] 2024-09-23 16:31:30.072 | INFO     | Epoch 26: train loss 0.1669	dev loss 0.5689	dev tag acc 97.13%	dev head acc 87.85%	dev deprel acc 91.70%
+[hops] 2024-09-23 16:31:30.073 | INFO     | New best model: head accuracy 87.85% > 87.84%
+[hops] 2024-09-23 16:31:42.885 | INFO     | Epoch 27: train loss 0.1614	dev loss 0.5304	dev tag acc 97.27%	dev head acc 87.78%	dev deprel acc 91.76%
+[hops] 2024-09-23 16:31:53.373 | INFO     | Epoch 28: train loss 0.1495	dev loss 0.5681	dev tag acc 97.40%	dev head acc 88.12%	dev deprel acc 92.03%
+[hops] 2024-09-23 16:31:53.374 | INFO     | New best model: head accuracy 88.12% > 87.85%
+[hops] 2024-09-23 16:32:05.726 | INFO     | Epoch 29: train loss 0.1408	dev loss 0.5820	dev tag acc 97.37%	dev head acc 88.11%	dev deprel acc 92.01%
+[hops] 2024-09-23 16:32:16.250 | INFO     | Epoch 30: train loss 0.1351	dev loss 0.5901	dev tag acc 97.46%	dev head acc 88.22%	dev deprel acc 92.02%
+[hops] 2024-09-23 16:32:16.251 | INFO     | New best model: head accuracy 88.22% > 88.12%
+[hops] 2024-09-23 16:32:28.631 | INFO     | Epoch 31: train loss 0.1305	dev loss 0.5919	dev tag acc 97.71%	dev head acc 88.00%	dev deprel acc 92.13%
+[hops] 2024-09-23 16:32:39.222 | INFO     | Epoch 32: train loss 0.1231	dev loss 0.5793	dev tag acc 97.65%	dev head acc 88.31%	dev deprel acc 92.27%
+[hops] 2024-09-23 16:32:39.223 | INFO     | New best model: head accuracy 88.31% > 88.22%
+[hops] 2024-09-23 16:32:52.026 | INFO     | Epoch 33: train loss 0.1161	dev loss 0.6157	dev tag acc 97.72%	dev head acc 88.26%	dev deprel acc 92.19%
+[hops] 2024-09-23 16:33:02.318 | INFO     | Epoch 34: train loss 0.1129	dev loss 0.6230	dev tag acc 97.86%	dev head acc 88.45%	dev deprel acc 92.34%
+[hops] 2024-09-23 16:33:02.319 | INFO     | New best model: head accuracy 88.45% > 88.31%
+[hops] 2024-09-23 16:33:13.887 | INFO     | Epoch 35: train loss 0.1054	dev loss 0.6268	dev tag acc 97.96%	dev head acc 88.53%	dev deprel acc 92.52%
+[hops] 2024-09-23 16:33:13.888 | INFO     | New best model: head accuracy 88.53% > 88.45%
+[hops] 2024-09-23 16:33:25.702 | INFO     | Epoch 36: train loss 0.1031	dev loss 0.6226	dev tag acc 97.94%	dev head acc 88.38%	dev deprel acc 92.19%
+[hops] 2024-09-23 16:33:35.604 | INFO     | Epoch 37: train loss 0.0954	dev loss 0.6804	dev tag acc 98.04%	dev head acc 88.48%	dev deprel acc 92.06%
+[hops] 2024-09-23 16:33:45.642 | INFO     | Epoch 38: train loss 0.0938	dev loss 0.6665	dev tag acc 98.14%	dev head acc 88.63%	dev deprel acc 92.31%
+[hops] 2024-09-23 16:33:45.642 | INFO     | New best model: head accuracy 88.63% > 88.53%
+[hops] 2024-09-23 16:33:57.304 | INFO     | Epoch 39: train loss 0.0914	dev loss 0.6795	dev tag acc 98.18%	dev head acc 88.77%	dev deprel acc 92.27%
+[hops] 2024-09-23 16:33:57.306 | INFO     | New best model: head accuracy 88.77% > 88.63%
+[hops] 2024-09-23 16:34:08.777 | INFO     | Epoch 40: train loss 0.0891	dev loss 0.6621	dev tag acc 98.14%	dev head acc 88.71%	dev deprel acc 92.48%
+[hops] 2024-09-23 16:34:18.811 | INFO     | Epoch 41: train loss 0.0813	dev loss 0.6896	dev tag acc 98.19%	dev head acc 88.63%	dev deprel acc 92.56%
+[hops] 2024-09-23 16:34:29.749 | INFO     | Epoch 42: train loss 0.0778	dev loss 0.6912	dev tag acc 98.19%	dev head acc 88.61%	dev deprel acc 92.53%
+[hops] 2024-09-23 16:34:40.303 | INFO     | Epoch 43: train loss 0.0752	dev loss 0.7193	dev tag acc 98.22%	dev head acc 88.63%	dev deprel acc 92.56%
+[hops] 2024-09-23 16:34:50.784 | INFO     | Epoch 44: train loss 0.0743	dev loss 0.7266	dev tag acc 98.16%	dev head acc 88.99%	dev deprel acc 92.50%
+[hops] 2024-09-23 16:34:50.785 | INFO     | New best model: head accuracy 88.99% > 88.77%
+[hops] 2024-09-23 16:35:03.591 | INFO     | Epoch 45: train loss 0.0730	dev loss 0.6925	dev tag acc 98.30%	dev head acc 88.74%	dev deprel acc 92.57%
+[hops] 2024-09-23 16:35:14.168 | INFO     | Epoch 46: train loss 0.0682	dev loss 0.7621	dev tag acc 98.23%	dev head acc 88.75%	dev deprel acc 92.34%
+[hops] 2024-09-23 16:35:25.058 | INFO     | Epoch 47: train loss 0.0670	dev loss 0.7358	dev tag acc 98.25%	dev head acc 88.82%	dev deprel acc 92.38%
+[hops] 2024-09-23 16:35:35.506 | INFO     | Epoch 48: train loss 0.0646	dev loss 0.7508	dev tag acc 98.24%	dev head acc 89.01%	dev deprel acc 92.52%
+[hops] 2024-09-23 16:35:35.507 | INFO     | New best model: head accuracy 89.01% > 88.99%
+[hops] 2024-09-23 16:35:47.820 | INFO     | Epoch 49: train loss 0.0637	dev loss 0.7604	dev tag acc 98.21%	dev head acc 88.84%	dev deprel acc 92.43%
+[hops] 2024-09-23 16:35:58.797 | INFO     | Epoch 50: train loss 0.0592	dev loss 0.7639	dev tag acc 98.24%	dev head acc 88.98%	dev deprel acc 92.48%
+[hops] 2024-09-23 16:36:09.269 | INFO     | Epoch 51: train loss 0.0593	dev loss 0.7748	dev tag acc 98.24%	dev head acc 88.94%	dev deprel acc 92.60%
+[hops] 2024-09-23 16:36:19.876 | INFO     | Epoch 52: train loss 0.0576	dev loss 0.7841	dev tag acc 98.29%	dev head acc 89.05%	dev deprel acc 92.40%
+[hops] 2024-09-23 16:36:19.877 | INFO     | New best model: head accuracy 89.05% > 89.01%
+[hops] 2024-09-23 16:36:31.352 | INFO     | Epoch 53: train loss 0.0571	dev loss 0.7784	dev tag acc 98.22%	dev head acc 89.02%	dev deprel acc 92.46%
+[hops] 2024-09-23 16:36:40.930 | INFO     | Epoch 54: train loss 0.0562	dev loss 0.7698	dev tag acc 98.24%	dev head acc 89.15%	dev deprel acc 92.45%
+[hops] 2024-09-23 16:36:40.931 | INFO     | New best model: head accuracy 89.15% > 89.05%
+[hops] 2024-09-23 16:36:52.477 | INFO     | Epoch 55: train loss 0.0532	dev loss 0.7898	dev tag acc 98.29%	dev head acc 89.14%	dev deprel acc 92.60%
+[hops] 2024-09-23 16:37:02.354 | INFO     | Epoch 56: train loss 0.0546	dev loss 0.7842	dev tag acc 98.27%	dev head acc 89.13%	dev deprel acc 92.63%
+[hops] 2024-09-23 16:37:12.220 | INFO     | Epoch 57: train loss 0.0498	dev loss 0.7922	dev tag acc 98.30%	dev head acc 89.13%	dev deprel acc 92.56%
+[hops] 2024-09-23 16:37:21.987 | INFO     | Epoch 58: train loss 0.0488	dev loss 0.8101	dev tag acc 98.30%	dev head acc 89.10%	dev deprel acc 92.62%
+[hops] 2024-09-23 16:37:31.749 | INFO     | Epoch 59: train loss 0.0507	dev loss 0.8213	dev tag acc 98.30%	dev head acc 89.12%	dev deprel acc 92.61%
+[hops] 2024-09-23 16:37:41.569 | INFO     | Epoch 60: train loss 0.0483	dev loss 0.8216	dev tag acc 98.32%	dev head acc 89.05%	dev deprel acc 92.66%
+[hops] 2024-09-23 16:37:51.421 | INFO     | Epoch 61: train loss 0.0488	dev loss 0.8189	dev tag acc 98.33%	dev head acc 89.11%	dev deprel acc 92.65%
+[hops] 2024-09-23 16:38:01.038 | INFO     | Epoch 62: train loss 0.0461	dev loss 0.8247	dev tag acc 98.32%	dev head acc 89.09%	dev deprel acc 92.65%
+[hops] 2024-09-23 16:38:11.711 | INFO     | Epoch 63: train loss 0.0449	dev loss 0.8253	dev tag acc 98.33%	dev head acc 89.06%	dev deprel acc 92.62%
+[hops] 2024-09-23 16:38:15.869 | WARNING  | You're using a RobertaTokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.
+[hops] 2024-09-23 16:38:22.277 | WARNING  | You're using a RobertaTokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.
+[hops] 2024-09-23 16:38:24.822 | INFO     | Metrics for Rhapsodie-camembertav2_base_p2_17k_last_layer+rand_seed=25
+ ─────────────────────────────── 
+  Split   UPOS     UAS     LAS   
+ ─────────────────────────────── 
+  Dev     98.24   89.24   85.62  
+  Test    97.77   88.66   84.56  
+ ─────────────────────────────── 
+