Upload folder using huggingface_hub

README.md

@@ -0,0 +1,117 @@
+---
+language: en
+license: apache-2.0
+tags:
+- sentiment-analysis
+- text-classification
+- bert
+- sst2
+- transformers
+datasets:
+- glue
+metrics:
+- accuracy
+- f1
+- precision
+- recall
+pipeline_tag: text-classification
+---
+
+# Fine-tuned BERT for Sentiment Analysis on SST-2
+
+This model is a **fine-tuned version of BERT (`bert-base-uncased`)** specifically designed for **binary sentiment classification** of English text, achieving state-of-the-art performance on the Stanford Sentiment Treebank v2 (SST-2) benchmark.
+
+## Model Description
+
+The model was created to demonstrate the practical application of **transfer learning** in Natural Language Processing (NLP). While the base BERT model has a deep understanding of general English language structure, it was not originally trained to detect sentiment. This fine-tuning process adapts BERT's powerful contextual embeddings to the specialized task of determining whether a given sentence expresses a **positive** or **negative** opinion.
+
+### Key Technical Details
+*   **Architecture:** BERT-base-uncased with a sequence classification head (2 output neurons).
+*   **Training Approach:** The pre-trained BERT layers were gently tuned while the newly added classification layer was trained from scratch over 3 epochs.
+*   **Framework:** PyTorch with the Hugging Face Transformers library.
+
+## Intended Use & Limitations
+
+### ✅ Intended Use
+This model is optimal for classifying the sentiment of **short English texts**, particularly:
+*   Movie or product reviews
+*   Social media posts (opinions)
+*   Customer feedback snippets
+
+### ⚠️ Limitations
+*   **Domain Specificity:** Performance may degrade on texts far outside the movie review domain (e.g., technical, financial, or medical jargon).
+*   **Binary Scope:** It is designed for positive/negative classification and does not detect neutral sentiment or more complex emotions.
+*   **Language:** Works only with English text.
+
+## Training Data
+
+The model was fine-tuned on the **Stanford Sentiment Treebank v2 (SST-2)** dataset from the GLUE benchmark.
+
+| Dataset Split | Number of Examples |
+| :--- | :--- |
+| **Training** | 67,349 |
+| **Validation** | 872 |
+| **Test** | 1,821 |
+
+**Example from the dataset:**
+*   **Sentence:** *"contains no wit , only labored gags"*
+*   **Label:** `0` (Negative)
+
+## Training Procedure & Hyperparameters
+
+The model was trained for **3 epochs** using the following configuration:
+
+| Hyperparameter | Value |
+| :--- | :--- |
+| **Learning Rate** | 2e-5 |
+| **Batch Size** | 16 |
+| **Optimizer** | AdamW |
+| **Weight Decay** | 0.01 |
+| **Warmup Steps** | 0 |
+| **Max Sequence Length** | 128 |
+
+The training leveraged the Hugging Face `Trainer` API for efficient optimization and evaluation.
+
+## Evaluation Results
+
+The model's performance was evaluated on the SST-2 **validation set**, yielding the following metrics:
+
+### 📊 Overall Performance
+| Metric | Score |
+| :--- | :--- |
+| **Accuracy** | **92.55%** |
+| **F1-Score (Macro Avg)** | **0.93** |
+| **Precision (Negative)** | 0.93 |
+| **Recall (Positive)** | 0.94 |
+
+### 📈 Training Progress
+| Epoch | Training Loss | Validation Loss | Validation Accuracy |
+| :--- | :--- | :--- | :--- |
+| 1 | 0.1760 | 0.2400 | 92.43% |
+| 2 | 0.1240 | 0.3320 | 91.63% |
+| 3 | **0.0704** | **0.3400** | **92.55%** |
+
+### Confusion Matrix (Validation Set, n=872)
+| | Predicted Negative | Predicted Positive |
+| :--- | :---: | :---: |
+| **Actual Negative** | **391** (TN) | 37 (FP) |
+| **Actual Positive** | 27 (FN) | **417** (TP) |
+
+## Live Inference Examples
+
+The model correctly classifies clear examples and shows nuanced understanding of ambiguous text:
+
+| Input Sentence | Predicted Label | Confidence (Negative, Positive) |
+| :--- | :--- | :--- |
+| *"The movie was fantastic!"* | **Positive** | [0.0002, 0.9998] |
+| *"I hated every minute of this film."* | **Negative** | [0.9994, 0.0006] |
+| *"It was okay, nothing special."* | **Positive** | [0.4308, 0.5692] |
+
+*Note: The third example shows low confidence, appropriately reflecting the neutral sentiment of the input.*
+
+## Conclusion
+
+This project successfully demonstrates how **transfer learning** with a foundation model like BERT can efficiently create a high-performance, specialized classifier. With minimal training time and data, the fine-tuned model achieves competitive results on a standard NLP benchmark, making it suitable for real-world sentiment analysis applications.
+
+---
+*Model card generated using best practices from the [Hugging Face Model Card Guidebook](https://huggingface.co/docs/hub/model-cards).*

config.json

@@ -0,0 +1,34 @@
+{
+  "architectures": [
+    "BertForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "negative",
+    "1": "positive"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "negative": 0,
+    "positive": 1
+  },
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "problem_type": "single_label_classification",
+  "transformers_version": "4.57.3",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 30522
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aef0ffb596aefbef130424a3b9b3c4276d3a16d62eeb09706a77c1bab427cbf2
+size 437958648

special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

tokenizer_config.json

@@ -0,0 +1,56 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100": {
+      "content": "[UNK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "101": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "102": {
+      "content": "[SEP]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "103": {
+      "content": "[MASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "clean_up_tokenization_spaces": false,
+  "cls_token": "[CLS]",
+  "do_lower_case": true,
+  "extra_special_tokens": {},
+  "mask_token": "[MASK]",
+  "model_max_length": 512,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "unk_token": "[UNK]"
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bfb7ae18e386369f751c5808ed9a9704f57a9f9600bc4ea5aa7a31eaedf8b1d1
+size 5841