Topic Classifier v2 Added

feat: Push updated Topic Classifier model with eval_loss 0.0233, eval_accuracy 0.9908, eval_f1 0.9908, CORPORATE_DOCUMENTS precision 1.00, FINANCIAL precision 0.95, HARMFUL precision 0.95, MEDICAL precision 0.99, accuracy 0.99, macro avg F1 0.97, weighted avg F1 0.99, support 4565 samples

README.md

@@ -0,0 +1,128 @@
+
+# Topic Classifier
+
+This repository contains the Topic Classifier model developed by DAXA.AI. The Topic Classifier is a machine learning model designed to categorize text documents across various domains, such as corporate documents, financial texts, harmful content, and medical documents.
+
+## Model Details
+
+### Model Description
+
+The Topic Classifier is a BERT-based model, fine-tuned from the `distilbert-base-uncased` model. It is intended for categorizing text into specific topics, including "CORPORATE_DOCUMENTS," "FINANCIAL," "HARMFUL," and "MEDICAL." This model streamlines text classification tasks across multiple sectors, making it suitable for various business use cases.
+
+- **Developed by:** DAXA.AI
+- **Funded by:** Open Source
+- **Model type:** Text classification
+- **Language(s):** English
+- **License:** MIT
+- **Fine-tuned from:** `distilbert-base-uncased`
+
+### Model Sources
+
+- **Repository:** [https://huggingface.co/daxa-ai/topic-classifier](https://huggingface.co/daxa-ai/Topic-Classifier-2)
+- **Demo:** [https://huggingface.co/spaces/daxa-ai/Topic-Classifier-2](https://huggingface.co/spaces/daxa-ai/Topic-Classifier-2)
+
+## Usage
+
+### How to Get Started with the Model
+
+To use the Topic Classifier in your Python project, you can follow the steps below:
+
+```python
+# Import necessary libraries
+from transformers import AutoTokenizer, AutoModelForSequenceClassification
+import torch
+import joblib
+from huggingface_hub import hf_hub_url, cached_download
+
+# Load the tokenizer and model
+tokenizer = AutoTokenizer.from_pretrained("daxa-ai/topic-classifier")
+model = AutoModelForSequenceClassification.from_pretrained("daxa-ai/topic-classifier")
+
+# Example text
+text = "Please enter your text here."
+encoded_input = tokenizer(text, return_tensors='pt')
+output = model(**encoded_input)
+
+# Apply softmax to the logits
+probabilities = torch.nn.functional.softmax(output.logits, dim=-1)
+
+# Get the predicted label
+predicted_label = torch.argmax(probabilities, dim=-1)
+
+# URL of your Hugging Face model repository
+REPO_NAME = "daxa-ai/topic-classifier"
+
+# Path to the label encoder file in the repository
+LABEL_ENCODER_FILE = "label_encoder.joblib"
+
+# Construct the URL to the label encoder file
+url = hf_hub_url(REPO_NAME, filename=LABEL_ENCODER_FILE)
+
+# Download and cache the label encoder file
+filename = cached_download(url)
+
+# Load the label encoder
+label_encoder = joblib.load(filename)
+
+# Decode the predicted label
+decoded_label = label_encoder.inverse_transform(predicted_label.numpy())
+
+print(decoded_label)
+```
+
+## Training Details
+
+### Training Data
+
+The training dataset consists of 29,286 entries, categorized into four distinct labels. The distribution of these labels is presented below:
+
+| Document Type       | Instances |
+| ------------------- | --------- |
+| CORPORATE_DOCUMENTS | 17,649    |
+| FINANCIAL           | 3,385     |
+| HARMFUL             | 2,388     |
+| MEDICAL             | 5,864     |
+
+### Evaluation
+
+#### Testing Data & Metrics
+
+The model was evaluated on a dataset consisting of 4,565 entries. The distribution of labels in the evaluation set is shown below:
+
+| Document Type       | Instances |
+| ------------------- | --------- |
+| CORPORATE_DOCUMENTS | 3,051     |
+| FINANCIAL           | 409       |
+| HARMFUL             | 246       |
+| MEDICAL             | 859       |
+
+The evaluation metrics include precision, recall, and F1-score, calculated for each label:
+
+| Document Type       | Precision | Recall | F1-Score | Support |
+| ------------------- | --------- | ------ | -------- | ------- |
+| CORPORATE_DOCUMENTS | 1.00      | 1.00   | 1.00     | 3,051   |
+| FINANCIAL           | 0.95      | 0.96   | 0.96     | 409     |
+| HARMFUL             | 0.95      | 0.95   | 0.95     | 246     |
+| MEDICAL             | 0.99      | 1.00   | 0.99     | 859     |
+| Accuracy            |           |        | 0.99     | 4,565   |
+| Macro Avg           | 0.97      | 0.98   | 0.97     | 4,565   |
+| Weighted Avg        | 0.99      | 0.99   | 0.99     | 4,565   |
+
+#### Test Data Evaluation Results
+
+The model's evaluation results are as follows:
+
+- **Evaluation Loss:** 0.0233
+- **Accuracy:** 0.9908
+- **Precision:** 0.9909
+- **Recall:** 0.9908
+- **F1-Score:** 0.9908
+- **Evaluation Runtime:** 30.1149 seconds
+- **Evaluation Samples Per Second:** 151.586
+- **Evaluation Steps Per Second:** 2.391
+
+## Conclusion
+
+The Topic Classifier achieves high accuracy, precision, recall, and F1-score, making it a reliable model for categorizing text across the domains of corporate documents, financial content, harmful content, and medical texts. The model is optimized for immediate deployment and works efficiently in real-world applications.
+
+For more information or to try the model yourself, check out the public space [here](https://huggingface.co/spaces/daxa-ai/Topic-Classifier-2).

config.json

@@ -0,0 +1,36 @@
+{
+  "_name_or_path": "distilbert-base-uncased",
+  "activation": "gelu",
+  "architectures": [
+    "DistilBertForSequenceClassification"
+  ],
+  "attention_dropout": 0.1,
+  "dim": 768,
+  "dropout": 0.1,
+  "hidden_dim": 3072,
+  "id2label": {
+    "0": "CORPORATE_DOCUMENTS",
+    "1": "FINANCIAL",
+    "2": "HARMFUL",
+    "3": "MEDICAL"
+  },
+  "initializer_range": 0.02,
+  "label2id": {
+    "CORPORATE_DOCUMENTS": 0,
+    "FINANCIAL": 1,
+    "HARMFUL": 2,
+    "MEDICAL": 3
+  },
+  "max_position_embeddings": 512,
+  "model_type": "distilbert",
+  "n_heads": 12,
+  "n_layers": 6,
+  "pad_token_id": 0,
+  "qa_dropout": 0.1,
+  "seq_classif_dropout": 0.2,
+  "sinusoidal_pos_embds": false,
+  "tie_weights_": true,
+  "torch_dtype": "float32",
+  "transformers_version": "4.45.1",
+  "vocab_size": 30522
+}

label_encoder.joblib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ecc34413f18d00dd522f2996ce202a485c39fc1e0def340590a6469914332400
+size 582

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01349bd229a507099512340ff61bf05d9a05fc96556d78f49f9338025ff60fa7
+size 267860714

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 +1,55 @@
+{
+  "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,
+  "mask_token": "[MASK]",
+  "model_max_length": 512,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "DistilBertTokenizer",
+  "unk_token": "[UNK]"
+}