Update model card: transformer-style usage + local run guide

LOCAL_SETUP_GUIDE.md

@@ -0,0 +1,59 @@
+# Local Setup Guide (Laptop)
+
+This model is part of the DevaFlow project (custom D3PM, not native `transformers.AutoModel` format).
+
+## 1) Environment
+
+```bash
+python3.11 -m venv .venv
+source .venv/bin/activate
+pip install -U pip
+pip install -r requirements.txt
+```
+
+## 2) Quick Inference
+
+```python
+from inference_api import predict
+print(predict("dharmo rakṣati rakṣitaḥ"))
+```
+
+## 3) Transformer-Style Use
+
+```python
+import torch
+from config import CONFIG
+from inference import load_model, _build_tokenizers
+
+cfg = CONFIG
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+model, cfg = load_model("best_model.pt", cfg, device)
+src_tok, tgt_tok = _build_tokenizers(cfg)
+
+text = "yadā mano nivarteta viṣayebhyaḥ svabhāvataḥ"
+input_ids = torch.tensor([src_tok.encode(text)], dtype=torch.long, device=device)
+out = model.generate(
+    input_ids,
+    num_steps=cfg["inference"]["num_steps"],
+    temperature=cfg["inference"]["temperature"],
+    top_k=cfg["inference"]["top_k"],
+    repetition_penalty=cfg["inference"]["repetition_penalty"],
+    diversity_penalty=cfg["inference"]["diversity_penalty"],
+)
+ids = [x for x in out[0].tolist() if x > 4]
+print(tgt_tok.decode(ids).strip())
+```
+
+## 4) Full Project Execution
+
+For training, UI, Tasks 1–5, ablation workflow, and HF deployment, use the full project repository and run:
+
+- `python train.py`
+- `python inference.py`
+- `python app.py`
+- `python analysis/run_analysis.py --task <1|2|3|4|5|all>`
+
+Task 4 note:
+- `--phase generate_configs` first
+- train ablation checkpoints
+- then `--phase analyze`

README.md

@@ -1,20 +1,24 @@
 ---
-datasets:
-- paws/sanskrit-verses-gretil
+license: mit
 language:
 - sa
-metrics:
-- bleurt
-- character
-base_model:
-- bhsinghgrid/DevaFlow
-library_name: adapter-transformers
+- en
+tags:
+- sanskrit
+- paraphrase
+- diffusion
+- d3pm
+- pytorch
+pipeline_tag: text-generation
 ---
 
 # Sanskrit D3PM Paraphrase Model
 
 Roman/IAST Sanskrit input to Devanagari output using a D3PM cross-attention model.
 
+This is a **custom PyTorch architecture** (not a native `transformers.AutoModel` checkpoint).
+You can still use it in a transformer-like workflow (load once, pass text, get generated text) via `inference_api.py`.
+
 ## Files Included
 
 - `best_model.pt` — trained checkpoint
@@ -24,6 +28,7 @@ Roman/IAST Sanskrit input to Devanagari output using a D3PM cross-attention mode
 - `handler.py` — Hugging Face Endpoint handler
 - `model/`, `diffusion/` — architecture modules
 - `sanskrit_src_tokenizer.json`, `sanskrit_tgt_tokenizer.json` — tokenizers
+- `LOCAL_SETUP_GUIDE.md` — full laptop setup and execution guide
 
 ## Quick Local Test
 
@@ -32,6 +37,46 @@ from inference_api import predict
 print(predict("dharmo rakṣati rakṣitaḥ")["output"])
 ```
 
+## Transformer-Style Usage (Recommended)
+
+Use this model as a reusable generation object:
+
+```python
+import torch
+from config import CONFIG
+from inference import load_model, _build_tokenizers
+
+cfg = CONFIG
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+model, cfg = load_model("best_model.pt", cfg, device)
+src_tok, tgt_tok = _build_tokenizers(cfg)
+
+def generate(text: str):
+    input_ids = torch.tensor([src_tok.encode(text)], dtype=torch.long, device=device)
+    output_ids = model.generate(
+        input_ids,
+        num_steps=cfg["inference"]["num_steps"],
+        temperature=cfg["inference"]["temperature"],
+        top_k=cfg["inference"]["top_k"],
+        repetition_penalty=cfg["inference"]["repetition_penalty"],
+        diversity_penalty=cfg["inference"]["diversity_penalty"],
+    )
+    ids = [x for x in output_ids[0].tolist() if x > 4]
+    return tgt_tok.decode(ids).strip()
+
+print(generate("yadā mano nivarteta viṣayebhyaḥ svabhāvataḥ"))
+```
+
+## About `transformers` Compatibility
+
+- This repo does not expose `config.json` + `model.safetensors` in `transformers` format.
+- If you want full `AutoModel`/`pipeline` compatibility, you must create a wrapper architecture and export weights into HF Transformers conventions.
+- For production today, use:
+  - `inference_api.py` for Python apps
+  - `handler.py` for HF Inference Endpoints
+  - `space_repo/app.py` for Gradio UI
+
 ## Endpoint Payload
 
 ```json
@@ -60,4 +105,10 @@ git remote add origin https://huggingface.co/<your-username>/sanskrit-d3pm
 git add .
 git commit -m "Initial model release"
 git push -u origin main
-```
+```
+
+## Full Local Laptop Guide
+
+For complete setup (training, inference, UI, tasks 1-5, ablation, and deployment), see:
+
+- `LOCAL_SETUP_GUIDE.md`