Update README.md

README.md

@@ -1,62 +1,130 @@
 ---
-base_model: Qwen/Qwen2.5-32B-Instruct
+language: en
+license: other
 library_name: peft
-model_name: warm_lora_redacted_v2_anoym
+base_model: Qwen/Qwen2.5-32B-Instruct
+pipeline_tag: text-generation
 tags:
-- base_model:adapter:Qwen/Qwen2.5-32B-Instruct
 - lora
-- sft
+- peft
 - transformers
-- trl
-licence: license
-pipeline_tag: text-generation
+- qwen2.5
+- conversational
 ---
 
-# Model Card for warm_lora_redacted_v2_anoym
+# DeepSupport Warm LoRA ❤️‍🩹
+
+This repository provides a LoRA adapter for DeepSupport Warm, an emotional-holding companion that offers gentle reflection and warm support without rushing into what to do next.
+
+- **Base model:** `Qwen/Qwen2.5-32B-Instruct`
+- **This repo:** LoRA adapter
+
+> Recommended: use this adapter together with the official base model.
+
+---
 
-This model is a fine-tuned version of [Qwen/Qwen2.5-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-32B-Instruct).
-It has been trained using [TRL](https://github.com/huggingface/trl).
+## What it does ✨
+
+DeepSupport Warm is designed to help users feel held and less alone in the moment:
+
+- Validate and name feelings without judging  
+- Stay with emotion first, before problem-solving  
+- Offer gentle grounding and a small next step only if the user wants it  
+
+---
+
+## Quick start 🚀
+
+### 1) Install
+
+```bash
+pip install -U "transformers>=4.40" peft accelerate safetensors
+```
 
-## Quick start
+### 2) Load base model and LoRA adapter
 
 ```python
-from transformers import pipeline
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM
+from peft import PeftModel
 
-question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
-generator = pipeline("text-generation", model="None", device="cuda")
-output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
-print(output["generated_text"])
+base_id = "Qwen/Qwen2.5-32B-Instruct"
+lora_id = "Yukyin/deepsupport-warm-lora-oss"
+
+tokenizer = AutoTokenizer.from_pretrained(base_id, trust_remote_code=True)
+base = AutoModelForCausalLM.from_pretrained(
+    base_id,
+    torch_dtype=torch.float16,
+    device_map="auto",
+    trust_remote_code=True,
+)
+
+model = PeftModel.from_pretrained(base, lora_id)
+model.eval()
+
+messages = [
+    {"role": "user", "content": "我最近压力很大，感觉自己一直在被否定。"},
+]
+
+inputs = tokenizer.apply_chat_template(
+    messages,
+    tokenize=True,
+    add_generation_prompt=True,
+    return_tensors="pt",
+).to(model.device)
+
+with torch.no_grad():
+    out = model.generate(
+        inputs,
+        max_new_tokens=256,
+        do_sample=True,
+        temperature=0.85,
+        top_p=0.9,
+        repetition_penalty=1.12,
+        no_repeat_ngram_size=4,
+    )
+
+print(tokenizer.decode(out[0], skip_special_tokens=True))
 ```
 
-## Training procedure
+---
 
- 
+## Training data and release notes 📊
 
+- This OSS LoRA adapter is trained on de-identified versions of the original data.
+- The original internal LoRA adapter was trained on non-de-identified data and cannot be open-sourced at this time.
+- More details and examples are provided in the [GitHub repo](https://github.com/Yukyin/DeepSupport).
 
-This model was trained with SFT.
+---
 
-### Framework versions
+## Safety and privacy ⚠️
 
-- PEFT 0.18.0
-- TRL: 0.24.0
-- Transformers: 4.57.3
-- Pytorch: 2.8.0
-- Datasets: 3.6.0
-- Tokenizers: 0.22.1
+This project is intended for supportive conversation only.  
+It does not provide professional advice, diagnosis, or therapy. Please seek qualified professional help when needed.
 
-## Citations
+---
 
+## License 📜
 
+This adapter is released for noncommercial use. See the [GitHub repo](https://github.com/Yukyin/DeepSupport) for the full license text and commercial licensing terms.
+
+---
+
+## Citation 📚
 
-Cite TRL as:
-    
 ```bibtex
-@misc{vonwerra2022trl,
-	title        = {{TRL: Transformer Reinforcement Learning}},
-	author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
-	year         = 2020,
-	journal      = {GitHub repository},
-	publisher    = {GitHub},
-	howpublished = {\url{https://github.com/huggingface/trl}}
+@software{deepsupport_warm_2026,
+  author  = {Yuyan Chen},
+  title   = {DeepSupport Warm: An emotional-holding companion for supportive dialogue},
+  year    = {2026},
+  version = {oss},
+  url     = {\url{https://github.com/Yukyin/DeepSupport/DeepSupport_Warm}}
 }
-```
+```
+
+---
+
+## Links
+
+- GitHub: https://github.com/Yukyin/DeepSupport
+- LoRA adapter: https://huggingface.co/Yukyin/deepsupport-warm-lora-oss