Upload 4 files

README.md

@@ -1,12 +1,36 @@
----
-title: LLMProject
-emoji: 📈
-colorFrom: green
-colorTo: indigo
-sdk: gradio
-sdk_version: 5.49.1
-app_file: app.py
-pinned: false
+
+# Multipurpose NLP Web App (Gradio on Hugging Face Spaces)
+
+An all‑in‑one **Translation** + **Summarization** app:
+
+- **Translation**: [facebook/m2m100_418M](https://huggingface.co/facebook/m2m100_418M) (~100 languages)
+- **Summarization**: [sshleifer/distilbart-cnn-12-6](https://huggingface.co/sshleifer/distilbart-cnn-12-6) (English)
+
+## Deploy in 3 minutes
+
+1. **Create a Space**: https://huggingface.co/new-space  
+   - **Space SDK**: `Gradio`  
+   - **Hardware**: CPU Basic is fine (bigger models may load slower).  
+2. **Upload these three files**: `app.py`, `requirements.txt`, `languages.json`.  
+3. **Commit** → The Space will build and start. No secrets needed.
+
+## Features
+- Two tabs: **Translate** and **Summarize**
+- Dropdowns for source/target languages (common ~45 languages pre‑mapped)
+- Adjustable generation lengths
+- Example inputs to test quickly
+
+## Notes
+- M2M100 requires setting `src_lang` and a `forced_bos_token_id` for the target language; handled for you in `app.py`.
+- Summarization is English‑focused for speed. If you need multilingual summaries, swap the model in `app.py` for an mT5‑based XLSum model (e.g., `csebuetnlp/mT5_multilingual_XLSum`) and add `sentencepiece` (already included).
+
+## Optional: Multilingual Summarization
+Replace the summarizer lines in `app.py` with:
+```python
+SUM_MODEL_NAME = "csebuetnlp/mT5_multilingual_XLSum"
+summarizer = pipeline("summarization", model=SUM_MODEL_NAME, tokenizer=SUM_MODEL_NAME)
+```
+
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+Built for assignments and demos. Enjoy!

app.py

@@ -0,0 +1,107 @@
+
+import gradio as gr
+from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, pipeline
+import json
+
+# ---- Load language map ----
+with open("languages.json", "r", encoding="utf-8") as f:
+    LANG_MAP = json.load(f)
+
+LANG_NAMES = list(LANG_MAP.keys())
+
+# ---- Translation setup (M2M100 supports ~100 languages) ----
+TRANS_MODEL_NAME = "facebook/m2m100_418M"
+trans_tokenizer = AutoTokenizer.from_pretrained(TRANS_MODEL_NAME)
+trans_model = AutoModelForSeq2SeqLM.from_pretrained(TRANS_MODEL_NAME)
+
+def translate_text(text, src_name, tgt_name, max_new_tokens=200):
+    if not text.strip():
+        return ""
+    src = LANG_MAP[src_name]
+    tgt = LANG_MAP[tgt_name]
+    # Set source language for tokenizer
+    trans_tokenizer.src_lang = src
+    # Tokenize
+    encoded = trans_tokenizer(text, return_tensors="pt", truncation=True, max_length=1024)
+    # Set target language via forced BOS token
+    forced_bos_token_id = trans_tokenizer.get_lang_id(tgt)
+    generated_tokens = trans_model.generate(
+        **encoded,
+        forced_bos_token_id=forced_bos_token_id,
+        max_new_tokens=max_new_tokens,
+        num_beams=4
+    )
+    return trans_tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)[0]
+
+# ---- Summarization setup (English-focused, small & fast) ----
+SUM_MODEL_NAME = "sshleifer/distilbart-cnn-12-6"
+summarizer = pipeline("summarization", model=SUM_MODEL_NAME)
+
+def summarize_text(long_text, min_len=60, max_len=160):
+    if not long_text.strip():
+        return ""
+    # DistilBART expects English and up to 1024 tokens; truncation handled by pipeline tokenizer
+    summary = summarizer(long_text, min_length=int(min_len), max_length=int(max_len), truncation=True)
+    return summary[0]["summary_text"]
+
+# ---- Gradio UI ----
+with gr.Blocks(title="Multipurpose NLP (Translate + Summarize)") as demo:
+    gr.Markdown(
+        """
+        # 🌐 Multipurpose NLP: Translation & Summarization
+        - **Translate** between many languages using **M2M100 (418M)**
+        - **Summarize** long English text using **DistilBART-CNN**
+        - Built with **Gradio** on Hugging Face Spaces
+        """
+    )
+
+    with gr.Tab("🗣️ Translate"):
+        with gr.Row():
+            src_lang = gr.Dropdown(choices=LANG_NAMES, value="English", label="Source language")
+            tgt_lang = gr.Dropdown(choices=LANG_NAMES, value="Hindi", label="Target language")
+        input_text = gr.Textbox(lines=6, label="Input text")
+        max_toks = gr.Slider(16, 400, value=200, step=1, label="Max new tokens")
+        translate_btn = gr.Button("Translate")
+        output_text = gr.Textbox(lines=6, label="Translation", interactive=False)
+
+        translate_btn.click(
+            fn=translate_text,
+            inputs=[input_text, src_lang, tgt_lang, max_toks],
+            outputs=[output_text]
+        )
+
+        gr.Examples(
+            examples=[
+                ["Hello, how are you?", "English", "French", 60],
+                ["India is developing rapidly in science and technology.", "English", "Hindi", 80],
+                ["مرحبا كيف حالك؟", "Arabic", "English", 60],
+            ],
+            inputs=[input_text, src_lang, tgt_lang, max_toks],
+        )
+
+    with gr.Tab("📝 Summarize"):
+        long_input = gr.Textbox(lines=12, label="Long input text (English)")
+        with gr.Row():
+            min_len = gr.Slider(20, 200, value=60, step=1, label="Min summary length")
+            max_len = gr.Slider(60, 400, value=160, step=1, label="Max summary length")
+        sum_btn = gr.Button("Summarize")
+        sum_output = gr.Textbox(lines=8, label="Summary", interactive=False)
+
+        sum_btn.click(
+            fn=summarize_text,
+            inputs=[long_input, min_len, max_len],
+            outputs=[sum_output]
+        )
+
+        gr.Examples(
+            examples=[
+                ["""Artificial intelligence (AI) has seen rapid growth over the past decade. 
+                Advances in deep learning, transformers, and large-scale datasets have enabled breakthroughs across natural language processing, computer vision, and speech. 
+                However, concerns about bias, interpretability, and energy consumption persist. 
+                As AI systems are deployed in high-stakes settings, researchers and policymakers emphasize responsible AI principles such as fairness, transparency, and accountability."""],
+            ],
+            inputs=[long_input],
+        )
+
+if __name__ == "__main__":
+    demo.launch()

languages.json

@@ -0,0 +1,51 @@
+{
+  "Arabic": "ar",
+  "Bengali": "bn",
+  "Bulgarian": "bg",
+  "Chinese (Simplified)": "zh",
+  "Chinese (Traditional)": "zh",
+  "Czech": "cs",
+  "Danish": "da",
+  "Dutch": "nl",
+  "English": "en",
+  "Estonian": "et",
+  "Finnish": "fi",
+  "French": "fr",
+  "German": "de",
+  "Greek": "el",
+  "Hebrew": "he",
+  "Hindi": "hi",
+  "Hungarian": "hu",
+  "Indonesian": "id",
+  "Italian": "it",
+  "Japanese": "ja",
+  "Kannada": "kn",
+  "Korean": "ko",
+  "Malay": "ms",
+  "Marathi": "mr",
+  "Norwegian": "no",
+  "Persian (Farsi)": "fa",
+  "Polish": "pl",
+  "Portuguese": "pt",
+  "Punjabi": "pa",
+  "Romanian": "ro",
+  "Russian": "ru",
+  "Sinhala": "si",
+  "Slovak": "sk",
+  "Spanish": "es",
+  "Swedish": "sv",
+  "Tamil": "ta",
+  "Telugu": "te",
+  "Thai": "th",
+  "Turkish": "tr",
+  "Ukrainian": "uk",
+  "Urdu": "ur",
+  "Vietnamese": "vi",
+  "Swahili": "sw",
+  "Yoruba": "yo",
+  "Zulu": "zu",
+  "Malayalam": "ml",
+  "Gujarati": "gu",
+  "Nepali": "ne",
+  "Filipino (Tagalog)": "tl"
+}

requirements.txt

@@ -0,0 +1,8 @@
+
+gradio>=4.44.0
+transformers>=4.44.0
+torch>=2.2.0
+sentencepiece>=0.1.99
+sacremoses>=0.0.53
+accelerate>=0.33.0
+safetensors>=0.4.2