Update docs (#12)

README.md

@@ -153,10 +153,10 @@ def image_qa(s, image_file, question):
 
 ### Constrained Decoding
 ```python
-@function
+@sgl.function
 def regular_expression_gen(s):
     s += "Q: What is the IP address of the Google DNS servers?\n"
-    s += "A: " + gen(
+    s += "A: " + sgl.gen(
         "answer",
         temperature=0,
         regex=r"((25[0-5]|2[0-4]\d|[01]?\d\d?).){3}(25[0-5]|2[0-4]\d|[01]?\d\d?)",
@@ -197,7 +197,7 @@ for out in state.text_iter():
 ## Backend: SGLang Runtime (SRT)
 The SGLang Runtime (SRT) is designed to work best with the SGLang frontend.
 However, it can also be used as a standalone API server.
-In this case, the RadixAttention can still greatly accelerate many use cases.
+In this case, the [RadixAttention](https://arxiv.org/abs/2312.07104) can still greatly accelerate many use cases.
 
 ### Usage
 Launch a server
@@ -237,7 +237,7 @@ python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port
 - Mixtral-8x7B on NVIDIA A10G, FP16, Tensor Parallelism=8
 ![mixtral_8x7b](assets/mixtral_8x7b.jpg)
 
-Learn more [here]().
+Learn more [here](docs/benchmark_results.md).
 
 ## Roadmap
 - [ ] Function call

docs/benchmark_results.md

+## Benchmark Results
+
+We tested our system on the following common LLM workloads and reported the achieved throughput:
+- **[MMLU](https://arxiv.org/abs/2009.03300)**: A 5-shot, multi-choice, multi-task benchmark.
+- **[HellaSwag](https://arxiv.org/abs/1905.07830)**: A 20-shot, multi-choice sentence completion benchmark.
+- **[ReAct Agent](https://arxiv.org/abs/2210.03629)**: An agent task using prompt traces collected from the original ReAct paper.
+- **[Tree-of-Thought](https://arxiv.org/pdf/2305.10601.pdf)**: A custom tree search-based prompt for solving GSM-8K problems.
+- **JSON Decode**: Extracting information from a Wikipedia page and outputting it in JSON format.
+- **Chat (short)**: A synthetic chat benchmark where each conversation includes 4 turns with short LLM outputs.
+- **Chat (long)**: A synthetic chat benchmark where each conversation includes 4 turns with long LLM outputs.
+- **[DSPy RAG](https://github.com/stanfordnlp/dspy)**: A retrieval-augmented generation pipeline in the DSPy tutorial.
+- **[LLaVA Bench](https://github.com/haotian-liu/LLaVA)**: Running LLaVA v1.5, a vision language model on the LLaVA-in-the-wild benchmark.
+
+We tested both Llama-7B on one NVIDIA A10G GPU (24GB) and Mixtral-8x7B on 8 NVIDIA A10G GPUs with tensor parallelism, using FP16 precision. We used vllm v0.2.5, guidance v0.1.8, and Hugging Face TGI v1.3.0 as baseline systems.
+
+
+- Llama-7B on NVIDIA A10G, FP16, Tensor Parallelism=1
+![llama_7b](../assets/llama_7b.jpg)
+
+- Mixtral-8x7B on NVIDIA A10G, FP16, Tensor Parallelism=8
+![mixtral_8x7b](../assets/mixtral_8x7b.jpg)
+
+The benchmark code is available [here](https://github.com/sgl-project/sglang/tree/main/benchmark).