Update Readme.md

README.md

@@ -6,16 +6,14 @@ tags:
 - generated_from_trainer
 metrics:
 - accuracy
-- f1
+- f1 (macro)
+- top3 accuracy
 model-index:
 - name: roberta-base-with-tweet-eval-emoji-full
   results: []
 ---
 
-<!-- This model card has been generated automatically according to the information the Trainer had access to. You
-should probably proofread and complete it, then remove this comment. -->
-
-# roberta-base-with-tweet-eval-emoji-full
+# Model description
 
 This model is a fine-tuned version of [roberta-base](https://huggingface.co/roberta-base) on the tweet_eval/emoji dataset.
 It achieves the following results on the evaluation set:
@@ -24,19 +22,84 @@ It achieves the following results on the evaluation set:
 - F1: 0.3314
 - Top3 Accuracy: 0.6504
 
-## Model description
-
-More information needed
-
-## Intended uses & limitations
-
-More information needed
-
-## Training and evaluation data
-
-More information needed
+## Example of classification
+
+```python
+import torch
+from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline
+from peft import PeftModel
+
+# Specify the same model ID pushed or trained locally
+MODEL_ID = "roberta-base-tweet-emoji-lora"
+
+# Load tokenizer + model
+device = 0 if torch.cuda.is_available() else -1
+tok = AutoTokenizer.from_pretrained(MODEL_ID)
+base_model = AutoModelForSequenceClassification.from_pretrained(
+    "FacebookAI/roberta-base",
+    num_labels=20,
+    ignore_mismatched_sizes=True,
+)
+model = PeftModel.from_pretrained(base_model, MODEL_ID).eval()
+
+pipe = pipeline(
+    task="text-classification",
+    model=model,
+    tokenizer=tok,
+    return_all_scores=True,
+    function_to_apply="softmax",
+    device=device,
+)
+
+# Map label IDs to emojis
+id2label = {
+    0: "❤",
+    1: "😍",
+    2: "😂",
+    3: "💕",
+    4: "🔥",
+    5: "😊",
+    6: "😎",
+    7: "✨",
+    8: "💙",
+    9: "😘",
+    10: "📷",
+    11: "🇺🇸",
+    12: "☀",
+    13: "💜",
+    14: "😉",
+    15: "💯",
+    16: "😁",
+    17: "🎄",
+    18: "📸",
+    19: "😜",
+}
+
+
+def predict_emojis(text, top_k=2):
+    """
+    Predict top k emojis for the given text.
+
+    Args:
+        text (str): Input string.
+        k (int): Number of top emojis to return.
+
+    Returns:
+        str: Space-separated top k emojis.
+    """
+    probs = pipe(text)[0]
+    top = sorted(probs, key=lambda x: x["score"], reverse=True)[:top_k]
+    return " ".join(id2label[int(d["label"].split("_")[-1])] for d in top)
+
+print(predict_emojis("Sunny days!"))
+```
+
+Output:
+
+```
+😎 ☀
+```
 
-## Training procedure
 
 ### Training hyperparameters
 
@@ -54,7 +117,7 @@ The following hyperparameters were used during training:
 
 ### Training results
 
-| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1     | Top3 Accuracy |
+| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 (Macro)     | Top3 Accuracy |
 |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:|:-------------:|
 | 2.1291        | 1.0   | 352  | 2.4306          | 0.219    | 0.2126 | 0.405         |
 | 2.1083        | 2.0   | 704  | 2.3812          | 0.2356   | 0.2411 | 0.43          |