Update README.md

README.md

@@ -1,7 +1,33 @@
----
-{}
----
-#### Typhoon Translate
+
+**Typhoon translate**
+
+**Typhoon translate** is a lightweight, 4-billion-parameter language model designed specifically for high-quality Thai ↔ English translation—right from your local device.
+
+Unlike general-purpose models, Typhoon Translate is fine-tuned for translation tasks and works best with dedicated prompts. Its strength lies in generating natural, fluent translations while preserving meaning and tone in both directions.
+
+Note: For optimal results, use the system prompts:
+`Translate the following text into Thai.` or
+`Translate the following text into English.`
+
+## **Performance**
+
+We used GPT-4o-mini as an "AI judge" in AlpacaEval 2.0, comparing Typhoon Translate against its own generations and other top systems.
+
+![EN -> TH performance]()
+
+![TH -> EN performance]()
+
+## **Model Description**
+
+- **Model type**: A 4B instruct decoder-only model based on Gemma3 architecture.
+- **Requirement**: transformers 4.51.1 or newer.
+- **Primary Language(s)**: Thai 🇹🇭 and English 🇬🇧
+- **License**: [Gemma License](https://github.com/google-deepmind/gemma/blob/main/LICENSE)
+
+
+## Quickstart
+
+This code snippet shows how to use the Typhoon translation model for Thai or English text generation using the transformers library with a specific prompt.
 
 ```python
 from transformers import AutoTokenizer, AutoModelForCausalLM
@@ -16,18 +42,16 @@ model = AutoModelForCausalLM.from_pretrained(
     device_map="auto",
 )
 
-### Translate Thai to English
 # messages = [
 #     {"role": "system", "content": "Translate the following text into English."},
-#     {"role": "user", "content": "ขอสูตรไก่ย่าง"},
+#     {"role": "user", "content": "ขออนุญาตสอบถามข้อมูลเพิ่มเติมได้ไหมครับ"},
 # ]
 
-### Translate English to Thai
+# ### Translate English to Thai
 messages = [
     {"role": "system", "content": "Translate the following text into Thai."},
-    {"role": "user", "content": "Hello, how are you?"},
+    {"role": "user", "content": "What is machine learning?"},
 ]
-
 input_ids = tokenizer.apply_chat_template(
     messages,
     add_generation_prompt=True,
@@ -36,11 +60,55 @@ input_ids = tokenizer.apply_chat_template(
 
 outputs = model.generate(
     input_ids,
-    max_new_tokens=512,
-    do_sample=True,
-    temperature=0.3,
+    max_new_tokens=8192,
+    temperature=0.2,
 )
 response = outputs[0][input_ids.shape[-1]:]
 print(tokenizer.decode(response, skip_special_tokens=True))
+```
+
+## Deploy as Server
+
+This section shows how to run Typhoon translate as an OpenAI-compatible API server using vllm.
+
+- SGLang:
+```base
+python3 -m sglang.launch_server scb10x/typhoon-translate-4b --context-length 16000 --dtype bfloat16
+```
+- vLLM:
+```bash
+vllm serve scb10x/typhoon-translate-4b --max-model-len 16000 --dtype bfloat16
+```
+## Best Practices
+To achieve optimal performance, we recommend the following settings:
+- Use system prompt `Translate the following text into Thai.` for English to Thai translation and `Translate the following text into English.` for Thai to English translation.
+- Set low temperature. 
+- Using an context length of 8192 tokens.
+
+## Intended Uses & Limitations
+This is a task-specific model intended to be used only with the provided prompts. It does not include any guardrails. Due to the nature of large language models (LLMs), a certain level of hallucination may occur. We recommend that developers carefully assess these risks in the context of their specific use case.
+
+
+## **Follow us**
+
+**https://twitter.com/opentyphoon**
+
+## **Support**
+
+**https://discord.gg/us5gAYmrxw**
+
+
+## **Citation**
 
+- If you find Typhoon2 useful for your work, please cite it using:
+```
+@misc{typhoon2,
+      title={Typhoon 2: A Family of Open Text and Multimodal Thai Large Language Models}, 
+      author={Kunat Pipatanakul and Potsawee Manakul and Natapong Nitarach and Warit Sirichotedumrong and Surapon Nonesung and Teetouch Jaknamon and Parinthapat Pengpun and Pittawat Taveekitworachai and Adisai Na-Thalang and Sittipong Sripaisarnmongkol and Krisanapong Jirayoot and Kasima Tharnpipitchai},
+      year={2024},
+      eprint={2412.13702},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={https://arxiv.org/abs/2412.13702}, 
+}
 ```