Prepare for open-source release

.github/workflows/yapf.yml

@@ -24,8 +24,7 @@ jobs:
     - name: Install dependencies
       run: |
         python -m pip install --upgrade pip
-        pip install yapf==0.32.0
-        pip install toml==0.10.2
+        python -m pip install -r requirements-dev.txt
     - name: Running yapf
       run: |
-        yapf --diff --recursive vllm --exclude 'vllm/model_executor/parallel_utils/**'
+        yapf --diff --recursive sarathi --exclude 'sarathi/model_executor/parallel_utils/**'

.gitignore

@@ -123,6 +123,9 @@ celerybeat.pid
 .env
 .venv
 env/
+env_sarathi/
+env_flashinfer/
+env_flashinfer_2/
 venv/
 ENV/
 env.bak/
@@ -181,5 +184,16 @@ _build/
 # Custom
 datasets/**/*.json
 datasets/**/*.svg
+old
 outputs
 wandb
+benchmark_output
+benchmark_output_old
+benchmark_output_old_1
+offline_inference_output
+capacity_search_output*
+isoqps_output*
+profiling_output
+env_vidur
+high_tail_latency_output
+profiling_outputs

.gitmodules

@@ -0,0 +1,4 @@
+[submodule "third_party/flash-attention"]
+	path = third_party/flash-attention
+	url = https://github.com/skrider/flash-attention.git
+	commit = 7968148214850ede9fa4bd515316c643ebd8ae83

.pylintrc

@@ -8,7 +8,7 @@
 [MASTER]
 
 # Files or directories to be skipped. They should be base names, not paths.
-ignore=docs,parallel_utils
+ignore=parallel_utils
 
 # Files or directories matching the regex patterns are skipped. The regex
 # matches against base names, not paths.
@@ -85,15 +85,15 @@ disable=abstract-method,
         intern-builtin,
         invalid-str-codec,
         locally-disabled,
-        logging-fstring-interpolation,  # added by vLLM
-        logging-not-lazy,  # added by vLLM
+        logging-fstring-interpolation,  # added by Sarathi
+        logging-not-lazy,  # added by Sarathi
         long-builtin,
         long-suffix,
         map-builtin-not-iterating,
         misplaced-comparison-constant,
-        missing-class-docstring,  # TODO (vLLM): enable
+        missing-class-docstring,  # TODO (Sarathi): enable
         missing-function-docstring,
-        missing-module-docstring,  # TODO (vLLM): enable
+        missing-module-docstring,  # TODO (Sarathi): enable
         metaclass-assignment,
         next-method-called,
         next-method-defined,
@@ -151,11 +151,6 @@ disable=abstract-method,
         wrong-import-order,
         xrange-builtin,
         zip-builtin-not-iterating,
-	protected-access,
-	unsubscriptable-object,
-	invalid-all-object,
-	redefined-builtin,
-	consider-using-in
 
 
 [REPORTS]
@@ -269,7 +264,7 @@ generated-members=
 [FORMAT]
 
 # Maximum number of characters on a single line.
-max-line-length=180
+max-line-length=80
 
 # TODO(https://github.com/PyCQA/pylint/issues/3352): Direct pylint to exempt
 # lines made too long by directives to pytype.