Add a benchmark and additional tests.

Also add a script to help running sanitizer builds, and do some cleanup.

Co-authored-by: Andrey Mikhaylov <amik@google.com>
Co-authored-by: Eugene Kliuchnikov <eustas@google.com>
Co-authored-by: Sami Boukortt <sboukortt@google.com>
Co-authored-by: Zoltan Szabadka <szabadka@google.com>

Author: 5 people

.clang-tidy

@@ -90,6 +90,7 @@ Checks: "-*,\
           -concurrency-mt-unsafe,\
           -cppcoreguidelines-avoid-c-arrays,\
           -cppcoreguidelines-avoid-const-or-ref-data-members,\
+          -cppcoreguidelines-avoid-do-while,\
           -cppcoreguidelines-avoid-goto,\
           -cppcoreguidelines-avoid-magic-numbers,\
           -cppcoreguidelines-avoid-non-const-global-variables,\

.github/workflows/build.yml

@@ -1,7 +1,11 @@
 name: build
 
 # Trigger on push, pull request, or via manual dispatch.
-on: [push, pull_request, workflow_dispatch]
+on:
+  push:
+  pull_request:
+    types: [opened, reopened, labeled, unlabeled, synchronize]
+  workflow_dispatch:
 
 jobs:
   build:

.gitignore

@@ -0,0 +1,4 @@
+.cache/
+bazel-*/
+build-*/
+python/*/__pycache__

BUILD.bazel

@@ -86,6 +86,19 @@ cc_library(
     ],
 )
 
+cc_test(
+    name = "gemma_test",
+    srcs = ["gemma_test.cc"],
+    deps = [
+        ":args",
+        ":gemma_lib",
+        ":ops",
+        "@googletest//:gtest_main",
+        "@hwy//:hwy_test_util",
+        "@hwy//:thread_pool",
+    ],
+)
+
 cc_library(
     name = "app",
     hdrs = [
@@ -132,3 +145,22 @@ cc_binary(
         "@hwy//:thread_pool",
     ],
 )
+
+cc_binary(
+    name = "benchmark",
+    srcs = [
+        "benchmark.cc",
+    ],
+    deps = [
+        ":app",
+        ":args",
+        ":gemma_lib",
+        # "//base",
+        "//compression:compress",
+        "@hwy//:hwy",
+        "@hwy//:nanobenchmark",
+        "@hwy//:profiler",
+        "@hwy//:thread_pool",
+        "@nlohmann_json//:json",
+    ],
+)

CMakeLists.txt

@@ -30,6 +30,9 @@ FetchContent_MakeAvailable(highway)
 FetchContent_Declare(sentencepiece GIT_REPOSITORY https://github.com/google/sentencepiece GIT_TAG 53de76561cfc149d3c01037f0595669ad32a5e7c EXCLUDE_FROM_ALL)
 FetchContent_MakeAvailable(sentencepiece)
 
+FetchContent_Declare(json GIT_REPOSITORY https://github.com/nlohmann/json.git GIT_TAG 9cca280a4d0ccf0c08f47a99aa71d1b0e52f8d03 EXCLUDE_FROM_ALL)
+FetchContent_MakeAvailable(json)
+
 set(SOURCES
   gemma.cc
   compression/blob_store.cc
@@ -60,17 +63,7 @@ if (WEIGHT_TYPE)
   add_definitions(-DGEMMA_WEIGHT_T=${WEIGHT_TYPE})
 endif()
 
-# Executable Target
-
-add_executable(gemma run.cc)
-target_sources(gemma PRIVATE ${SOURCES})
-set_property(TARGET gemma PROPERTY CXX_STANDARD 17)
-target_link_libraries(gemma hwy hwy_contrib sentencepiece)
-target_include_directories(gemma PRIVATE ./)
 FetchContent_GetProperties(sentencepiece)
-target_include_directories(gemma PRIVATE ${sentencepiece_SOURCE_DIR})
-target_compile_definitions(gemma PRIVATE $<$<PLATFORM_ID:Windows>:_CRT_SECURE_NO_WARNINGS NOMINMAX>)
-target_compile_options(gemma PRIVATE $<$<PLATFORM_ID:Windows>:-Wno-deprecated-declarations>)
 
 ## Library Target
 
@@ -84,11 +77,21 @@ target_include_directories(libgemma PUBLIC ${sentencepiece_SOURCE_DIR})
 target_compile_definitions(libgemma PRIVATE $<$<PLATFORM_ID:Windows>:_CRT_SECURE_NO_WARNINGS NOMINMAX>)
 target_compile_options(libgemma PRIVATE $<$<PLATFORM_ID:Windows>:-Wno-deprecated-declarations>)
 
+# Executable Target
+
+add_executable(gemma run.cc)
+target_link_libraries(gemma libgemma hwy hwy_contrib)
+
+add_executable(benchmark benchmark.cc)
+target_link_libraries(benchmark libgemma hwy hwy_contrib nlohmann_json::nlohmann_json)
+
+## Tests
 set(GEMMA_ENABLE_TESTS OFF CACHE BOOL "Enable Gemma tests")
 if (GEMMA_ENABLE_TESTS)
 
 set(GEMMA_TEST_FILES
   ops_test.cc
+  gemma_test.cc
 )
 
 foreach (TESTFILE IN LISTS GEMMA_TEST_FILES)

MODULE.bazel

@@ -20,6 +20,13 @@ http_archive(
     strip_prefix = "highway-1.1.0",
 )
 
+http_archive(
+    name = "nlohmann_json",
+    urls = ["https://github.com/nlohmann/json/archive/refs/tags/v3.11.3.zip"],
+    integrity = "sha256-BAIrBdgG61/3MCPCgLaGl9Erk+G3JnoLIqGjnsdXgGk=",
+    strip_prefix = "json-3.11.3",
+)
+
 http_archive(
     name = "com_google_sentencepiece",
     sha256 = "8409b0126ebd62b256c685d5757150cf7fcb2b92a2f2b98efb3f38fc36719754",