Validation Metrics
loss: 0.1564033180475235
f1_macro: 0.9023266184854538
f1_micro: 0.9275
f1_weighted: 0.9281147770697895
precision_macro: 0.8944987578959265
precision_micro: 0.9275
precision_weighted: 0.9308721399366291
recall_macro: 0.9135199509056998
recall_micro: 0.9275
recall_weighted: 0.9275
accuracy: 0.9275
Exemple d'utilisation
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
# Définir le nom du modèle et le token d'accès
model_name = "TPM-28/emotion-FR-camembert"
access_token = "<HF_token>"
# Charger le tokenizer et le modèle
tokenizer = AutoTokenizer.from_pretrained(model_name, use_auth_token=access_token)
model = AutoModelForSequenceClassification.from_pretrained(model_name, use_auth_token=access_token)
# Définir les classes
classes = ["anger", "fear", "joy", "love", "sadness", "surprise"]
def classify_text(text):
# Tokenizer le texte
inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=512)
# Effectuer l'inférence
with torch.no_grad():
outputs = model(**inputs)
# Obtenir les prédictions
probabilities = torch.nn.functional.softmax(outputs.logits, dim=-1)
prediction = torch.argmax(probabilities, dim=-1)
# Obtenir la classe prédite et sa probabilité
predicted_class = classes[prediction.item()]
confidence = probabilities[0][prediction].item()
return predicted_class, confidence
# Exemple d'utilisation
text_to_classify = "je suis vraiment content"
predicted_class, confidence = classify_text(text_to_classify)
print(f"Texte : {text_to_classify}")
print(f"Classe prédite : {predicted_class}")
print(f"Confiance : {confidence:.2f}")
- Downloads last month
- 0
Model tree for TPM-28/emotion-FR-camembert
Base model
almanach/camembert-base