Upload README.md

README.md

@@ -0,0 +1,70 @@
+---
+# For reference on model card metadata, see the spec: https://github.com/huggingface/hub-docs/blob/main/modelcard.md?plain=1
+# Doc / guide: https://huggingface.co/docs/hub/model-cards
+license: mit
+language:
+- cs
+---
+# Model Card for small-e-czech-binary-online-risks-cs
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+This model is fine-tuned for binary text classification of Online Risks in Instant Messenger dialogs of Adolescents in Czech. 
+
+## Model Description
+
+The model was fine-tuned on a dataset of Czech Instant Messenger dialogs of Adolescents. The classification is binary and the model outputs probablities for labels {0,1}: Online Risks present or not.
+
+- **Developed by:** Anonymous
+- **Language(s):** cs
+- **Finetuned from:** Seznam/small-e-czech
+
+## Model Sources
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** https://github.com/justtherightsize/supportive-interactions-and-risks
+- **Paper:** Stay tuned!
+
+## Usage
+Here is how to use this model to classify a context-window of a dialogue:
+
+```python
+import numpy as np
+from transformers import AutoTokenizer, AutoModelForSequenceClassification
+
+# Prepare input texts. This model is fine-tuned for Czech
+test_texts = ['Utterance1;Utterance2;Utterance3']
+
+# Load the model and tokenizer
+model = AutoModelForSequenceClassification.from_pretrained(
+    'justtherightsize/small-e-czech-binary-online-risks-cs', num_labels=2).to("cuda")
+
+tokenizer = AutoTokenizer.from_pretrained(
+    'justtherightsize/small-e-czech-binary-online-risks-cs',
+    use_fast=False, truncation_side='left')
+assert tokenizer.truncation_side == 'left'
+
+# Define helper functions
+def get_probs(text, tokenizer, model):
+    inputs = tokenizer(text, padding=True, truncation=True, max_length=256,
+                       return_tensors="pt").to("cuda")
+    outputs = model(**inputs)
+    return outputs[0].softmax(1)
+
+def preds2class(probs, threshold=0.5):
+    pclasses = np.zeros(probs.shape)
+    pclasses[np.where(probs >= threshold)] = 1
+    return pclasses.argmax(-1)
+
+def print_predictions(texts):
+    probabilities = [get_probs(
+        texts[i], tokenizer, model).cpu().detach().numpy()[0]
+                     for i in range(len(texts))]
+    predicted_classes = preds2class(np.array(probabilities))
+    for c, p in zip(predicted_classes, probabilities):
+        print(f'{c}: {p}')
+
+# Run the prediction
+print_predictions(test_texts)
+```