Spaces:
Running
Running
File size: 12,114 Bytes
5f923cd | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 | # Copyright 2025 The ODML Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import io
import os
import pathlib
import subprocess
from absl.testing import absltest
from litert_lm.runtime.proto import llm_metadata_pb2
from litert_lm.schema.py import litertlm_core
from litert_lm.schema.py import litertlm_peek
from python import runfiles
_TOML_TEMPLATE = """
# A template for testing the TOML parser.
[system_metadata]
entries = [
{ key = "author", value_type = "String", value = "The ODML Authors" }
]
[[section]]
# Section 0: LlmMetadataProto
section_type = "LlmMetadata"
data_path = "{LLM_METADATA_PATH}"
[[section]]
# Section 1: HF_Tokenizer
section_type = "HF_Tokenizer"
data_path = "{HF_TOKENIZER_PATH}"
[[section]]
# Section 2: TFLiteModel (Embedder)
section_type = "TFLiteModel"
model_type = "EMBEDDER"
data_path = "{EMBEDDER_PATH}"
[[section]]
# Section 3: TFLiteModel (Prefill/Decode)
section_type = "TFLiteModel"
model_type = "PREFILL_DECODE"
backend_constraint = "GPU"
data_path = "{PREFILL_DECODE_PATH}"
additional_metadata = [
{ key = "License", value_type = "String", value = "Example" }
]
"""
class LiteRTLMBuilderCLITest(absltest.TestCase):
def setUp(self):
super().setUp()
self.temp_dir = self.create_tempdir().full_path
def _create_placeholder_file(self, filename: str, content: bytes) -> str:
filepath = os.path.join(self.temp_dir, filename)
with litertlm_core.open_file(filepath, "wb") as f:
f.write(content)
return filepath
def _get_command_path(self) -> str:
"""Returns the path to the command binary."""
r = runfiles.Create()
return r.Rlocation(
os.path.join(
os.path.dirname(os.path.abspath(__file__)), "litertlm_builder_cli"
)
)
def _run_command(self, *args) -> str:
"""Runs the command with the given arguments."""
output_path = os.path.join(self.temp_dir, "litertlm.litertlm")
command = [
self._get_command_path(),
*args,
"output",
"--path",
output_path,
]
try:
subprocess.run(command, check=True, capture_output=True)
except subprocess.CalledProcessError as e:
print("command stdout:\n", e.stdout.decode("utf-8"))
print("command stderr:\n", e.stderr.decode("utf-8"))
raise e
return output_path
def _peek_litertlm_file(self, path: str) -> str:
"""Peeks the litertlm file and returns the string representation."""
stream = io.StringIO()
litertlm_peek.peek_litertlm_file(path, self.temp_dir, stream)
return stream.getvalue()
def test_system_metadata(self):
"""Tests that system metadata can be added correctly."""
args = ["system_metadata", "--str", "key1", "value1"]
output_path = self._run_command(*args)
self.assertTrue(os.path.exists(output_path))
ss = self._peek_litertlm_file(output_path)
self.assertIn("Key: key1, Value (String): value1", ss)
self.assertIn("Sections (0)", ss)
def test_llm_metadata(self):
"""Tests that LLM metadata can be added from a binary proto file."""
llm_metadata = llm_metadata_pb2.LlmMetadata(max_num_tokens=123)
bin_proto = llm_metadata.SerializeToString()
metadata_path = self._create_placeholder_file("llm.pb", bin_proto)
args = [
"system_metadata",
"--int",
"my_key",
"23",
"llm_metadata",
"--path",
metadata_path,
]
output_path = self._run_command(*args)
self.assertTrue(os.path.exists(output_path))
ss = self._peek_litertlm_file(output_path)
self.assertIn("max_num_tokens: 123", ss)
self.assertIn("Sections (1)", ss)
def test_tflite_model(self):
"""Tests that a TFLite model can be added correctly."""
tflite_path = self._create_placeholder_file(
"model.tflite", b"dummy tflite content"
)
args = [
"system_metadata",
"--int",
"my_key",
"23",
"tflite_model",
"--path",
tflite_path,
"--model_type",
"prefill_decode",
"--str_metadata",
"model_version",
"1.0.1",
"--backend_constraint",
"CPU",
]
output_path = self._run_command(*args)
self.assertTrue(os.path.exists(output_path))
ss = self._peek_litertlm_file(output_path)
self.assertIn("Sections (1)", ss)
self.assertIn("Data Type: TFLiteModel", ss)
self.assertIn("Key: model_type, Value (String): tf_lite_prefill_decode", ss)
self.assertIn("Key: model_version, Value (String): 1.0.1", ss)
self.assertIn("Key: backend_constraint, Value (String): cpu", ss)
def test_tflite_weights(self):
"""Tests that TFLite weights can be added correctly via CLI."""
tflite_path = self._create_placeholder_file(
"model.weights", b"dummy tflite weights content"
)
args = [
"system_metadata",
"--int",
"my_key",
"23",
"tflite_weights",
"--path",
tflite_path,
"--model_type",
"prefill_decode",
"--str_metadata",
"weights_version",
"1.0.1",
]
output_path = self._run_command(*args)
self.assertTrue(os.path.exists(output_path))
ss = self._peek_litertlm_file(output_path)
self.assertIn("Sections (1)", ss)
self.assertIn("Data Type: TFLiteWeights", ss)
self.assertIn("Key: model_type, Value (String): tf_lite_prefill_decode", ss)
self.assertIn("Key: weights_version, Value (String): 1.0.1", ss)
def test_sp_tokenizer(self):
"""Tests that a SentencePiece tokenizer can be added correctly."""
sp_path = self._create_placeholder_file("sp.model", b"dummy sp content")
args = [
"system_metadata",
"--int",
"my_key",
"23",
"sp_tokenizer",
"--path",
sp_path,
"--str_metadata",
"tokenizer_version",
"1.0.1",
]
output_path = self._run_command(*args)
self.assertTrue(os.path.exists(output_path))
ss = self._peek_litertlm_file(output_path)
self.assertIn("Sections (1)", ss)
self.assertIn("Data Type: SP_Tokenizer", ss)
self.assertIn("Key: tokenizer_version, Value (String): 1.0.1", ss)
def test_hf_tokenizer(self):
"""Tests that a HuggingFace tokenizer can be added correctly."""
hf_path = self._create_placeholder_file(
"tokenizer.json", b'{"version": "1.0"}'
)
args = [
"system_metadata",
"--int",
"my_key",
"23",
"hf_tokenizer",
"--path",
hf_path,
"--str_metadata",
"tokenizer_version",
"1.0.1",
]
output_path = self._run_command(*args)
self.assertTrue(os.path.exists(output_path))
ss = self._peek_litertlm_file(output_path)
self.assertIn("Sections (1)", ss)
self.assertIn("Data Type: HF_Tokenizer_Zlib", ss)
self.assertIn("Key: tokenizer_version, Value (String): 1.0.1", ss)
def test_end_to_end(self):
"""Tests a more complex end-to-end scenario with multiple sections."""
sp_path = self._create_placeholder_file("sp.model", b"dummy sp content")
tflite_path = self._create_placeholder_file(
"model.tflite", b"dummy tflite content"
)
llm_metadata = llm_metadata_pb2.LlmMetadata(max_num_tokens=123)
bin_proto = llm_metadata.SerializeToString()
metadata_path = self._create_placeholder_file("llm.pb", bin_proto)
args = [
"system_metadata",
"--str",
"Authors",
"ODML team",
"sp_tokenizer",
"--path",
sp_path,
"tflite_model",
"--path",
tflite_path,
"--model_type",
"embedder",
"tflite_model",
"--path",
tflite_path,
"--model_type",
"prefill_decode",
"--backend_constraint",
"GPU",
"llm_metadata",
"--path",
metadata_path,
]
output_path = self._run_command(*args)
self.assertTrue(os.path.exists(output_path))
ss = self._peek_litertlm_file(output_path)
self.assertIn("Sections (4)", ss)
self.assertIn("Data Type: SP_Tokenizer", ss)
self.assertIn("Data Type: TFLiteModel", ss)
self.assertIn("Key: model_type, Value (String): tf_lite_embedder", ss)
self.assertIn("Key: model_type, Value (String): tf_lite_prefill_decode", ss)
self.assertIn("Key: backend_constraint, Value (String): gpu", ss)
self.assertIn("Data Type: LlmMetadataProto", ss)
self.assertIn("max_num_tokens: 123", ss)
def test_toml_file(self):
"""Tests that a TOML file can be added correctly."""
hf_path = pathlib.Path(
self._create_placeholder_file("tokenizer.json", b'{"version": "1.0"}')
).as_posix()
tflite_path = pathlib.Path(
self._create_placeholder_file("model.tflite", b"dummy tflite content")
).as_posix()
llm_metadata = llm_metadata_pb2.LlmMetadata(max_num_tokens=123)
bin_proto = llm_metadata.SerializeToString()
metadata_path = pathlib.Path(
self._create_placeholder_file("llm.pb", bin_proto)
).as_posix()
toml_path = self._create_placeholder_file(
"test.toml",
_TOML_TEMPLATE.replace("{LLM_METADATA_PATH}", metadata_path)
.replace("{HF_TOKENIZER_PATH}", hf_path)
.replace("{EMBEDDER_PATH}", tflite_path)
.replace("{PREFILL_DECODE_PATH}", tflite_path)
.encode("utf-8"),
)
args = [
"toml",
"--path",
toml_path,
]
output_path = self._run_command(*args)
self.assertTrue(os.path.exists(output_path))
ss = self._peek_litertlm_file(output_path)
self.assertIn("Sections (4)", ss)
self.assertIn("Data Type: HF_Tokenizer_Zlib", ss)
self.assertIn("Data Type: TFLiteModel", ss)
self.assertIn("Key: model_type, Value (String): tf_lite_embedder", ss)
self.assertIn("Key: model_type, Value (String): tf_lite_prefill_decode", ss)
self.assertIn("Key: backend_constraint, Value (String): gpu", ss)
self.assertIn("Data Type: LlmMetadataProto", ss)
self.assertIn("max_num_tokens: 123", ss)
def test_toml_cannot_be_used_with_other_args(self):
"""Tests that a TOML file cannot be used with other args."""
tflite_path = self._create_placeholder_file(
"model.tflite", b"dummy tflite content"
)
toml_path = self._create_placeholder_file(
"test.toml",
_TOML_TEMPLATE.replace("{PREFILL_DECODE_PATH}", tflite_path).encode(
"utf-8"
),
)
args = [
"toml",
"--path",
toml_path,
"system_metadata",
"--str",
"key1",
"value1",
]
with self.assertRaises(subprocess.CalledProcessError):
self._run_command(*args)
def test_help_root(self):
"""Tests that the help command prints the correct output."""
command = [self._get_command_path(), "--help"]
output = subprocess.run(command, check=True, capture_output=True)
self.assertEqual(output.returncode, 0)
stdout = output.stdout.decode("utf-8")
self.assertIn(
"Build a LiteRT-LM file from input files and metadata", stdout
)
def test_help_subcommand(self):
"""Tests that the help command prints the correct output for subcommand."""
command = [self._get_command_path(), "system_metadata", "--help"]
output = subprocess.run(command, check=True, capture_output=True)
self.assertEqual(output.returncode, 0)
stdout = output.stdout.decode("utf-8")
self.assertIn(
"Add one or more system metadata key-value pairs to the LiteRT-LM"
" file.",
stdout,
)
if __name__ == "__main__":
absltest.main()
|