Simplify frontend language (#9029)

benchmark/gsm8k/bench_sglang.py

@@ -7,7 +7,7 @@ import time
 
 import numpy as np
 
-from sglang.api import set_default_backend
+from sglang.lang.api import set_default_backend
 from sglang.test.test_utils import (
     add_common_sglang_args_and_parse,
     dump_bench_raw_result,

benchmark/hellaswag/bench_sglang.py

@@ -5,7 +5,7 @@ import time
 
 import numpy as np
 
-from sglang.api import set_default_backend
+from sglang.lang.api import set_default_backend
 from sglang.test.test_utils import (
     add_common_sglang_args_and_parse,
     select_sglang_backend,

python/pyproject.toml

@@ -47,10 +47,10 @@ runtime_common = [
     "sentencepiece",
     "soundfile==0.13.1",
     "scipy",
-    "torchao==0.9.0",
-    "transformers==4.55.0",
     "timm==1.0.16",
     "tiktoken",
+    "torchao==0.9.0",
+    "transformers==4.55.0",
     "uvicorn",
     "uvloop",
     "xgrammar==0.1.22",
@@ -84,6 +84,9 @@ srt_hip = [
     "petit_kernel==0.0.2",
 ]
 
+# CPU: torch wheel for CPU needs to be installed from https://download.pytorch.org/whl/cpu
+srt_cpu = ["sglang[runtime_common]", "einops"]
+
 # xpu is not enabled in public vllm and torch whl,
 # need to follow https://docs.vllm.ai/en/latest/getting_started/xpu-installation.htmlinstall vllm
 srt_xpu = ["sglang[runtime_common]"]
@@ -92,8 +95,6 @@ srt_xpu = ["sglang[runtime_common]"]
 # https://docs.vllm.ai/en/latest/getting_started/gaudi-installation.html
 srt_hpu = ["sglang[runtime_common]"]
 
-# CPU: torch wheel for CPU needs to be installed from https://download.pytorch.org/whl/cpu
-srt_cpu = ["sglang[runtime_common]", "einops"]
 # https://vllm-ascend.readthedocs.io/en/latest/installation.html
 srt_npu = ["sglang[runtime_common]"]
 
@@ -112,12 +113,12 @@ test = [
     "sentence_transformers",
     "pytest",
 ]
-all = ["sglang[srt]", "sglang[openai]", "sglang[anthropic]", "sglang[litellm]", "sglang[torch_memory_saver]", "sglang[decord]"]
-all_hip = ["sglang[srt_hip]", "sglang[openai]", "sglang[anthropic]", "sglang[litellm]", "sglang[decord]"]
-all_xpu = ["sglang[srt_xpu]", "sglang[openai]", "sglang[anthropic]", "sglang[litellm]", "sglang[decord]"]
-all_hpu = ["sglang[srt_hpu]", "sglang[openai]", "sglang[anthropic]", "sglang[litellm]", "sglang[decord]"]
-all_cpu = ["sglang[srt_cpu]", "sglang[openai]", "sglang[anthropic]", "sglang[litellm]", "sglang[decord]"]
-all_npu = ["sglang[srt_npu]", "sglang[openai]", "sglang[anthropic]", "sglang[litellm]", "sglang[decord]"]
+all = ["sglang[srt]", "sglang[openai]", "sglang[anthropic]", "sglang[torch_memory_saver]", "sglang[decord]"]
+all_hip = ["sglang[srt_hip]", "sglang[openai]", "sglang[anthropic]", "sglang[decord]"]
+all_xpu = ["sglang[srt_xpu]", "sglang[openai]", "sglang[anthropic]", "sglang[decord]"]
+all_hpu = ["sglang[srt_hpu]", "sglang[openai]", "sglang[anthropic]", "sglang[decord]"]
+all_cpu = ["sglang[srt_cpu]", "sglang[openai]", "sglang[anthropic]", "sglang[decord]"]
+all_npu = ["sglang[srt_npu]", "sglang[openai]", "sglang[anthropic]", "sglang[decord]"]
 
 dev = ["sglang[all]", "sglang[test]"]
 dev_hip = ["sglang[all_hip]", "sglang[test]"]

python/sglang/README.md

@@ -5,7 +5,7 @@
 - `srt`: The backend engine for running local models. (SRT = SGLang Runtime).
 - `test`: The test utilities.
 - `api.py`: The public APIs.
-- `bench_offline_throughput.py`: Benchmark the throughput in the offline mode.
+- `bench_offline_throughput.py`: Benchmark the performance in the offline mode.
 - `bench_one_batch.py`: Benchmark the latency of running a single static batch without a server.
 - `bench_one_batch_server.py`: Benchmark the latency of running a single batch with a server.
 - `bench_serving.py`: Benchmark online serving with dynamic requests.

python/sglang/__init__.py

 # SGLang public APIs
 
 # Frontend Language APIs
-from sglang.api import (
+from sglang.global_config import global_config
+from sglang.lang.api import (
     Engine,
     Runtime,
     assistant,
@@ -25,13 +26,13 @@ from sglang.api import (
     user_end,
     video,
 )
-from sglang.global_config import global_config
 from sglang.lang.backend.runtime_endpoint import RuntimeEndpoint
 from sglang.lang.choices import (
     greedy_token_selection,
     token_length_normalized,
     unconditional_likelihood_normalized,
 )
+from sglang.srt.entrypoints.engine import Engine
 from sglang.utils import LazyImport
 from sglang.version import __version__
 

python/sglang/test/few_shot_gsm8k.py

@@ -12,7 +12,7 @@ import time
 
 import numpy as np
 
-from sglang.api import set_default_backend
+from sglang.lang.api import set_default_backend
 from sglang.lang.backend.runtime_endpoint import RuntimeEndpoint
 from sglang.utils import download_and_cache_file, dump_state_text, read_jsonl
 

python/sglang/test/few_shot_gsm8k_engine.py

@@ -8,7 +8,7 @@ import time
 import numpy as np
 
 import sglang as sgl
-from sglang.api import set_default_backend
+from sglang.lang.api import set_default_backend
 from sglang.lang.backend.runtime_endpoint import RuntimeEndpoint
 from sglang.utils import download_and_cache_file, dump_state_text, read_jsonl