A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 09:45 UTC
fix: Add type specifications for CLI arguments in tensorizer options (#18314)
CYJiang committed
May 19, 2025 at 14:42 UTC
275c5daeb0048c3b3f359bb5d9478b1e75e02857
1 file changed
+5
vllm/model_executor/model_loader/tensorizer.py
+5
@@ -214,6 +214,7 @@ class TensorizerArgs:
214
215
group.add_argument(
216
"--tensorizer-uri",
217
+ type=str,
218
help="Path to serialized model tensors. Can be a local file path,"
219
" or an HTTP(S) or S3 URI.",
220
)
@@ -226,6 +227,7 @@ class TensorizerArgs:
227
)
228
group.add_argument(
229
"--encryption-keyfile",
230
+ type=str,
231
default=None,
232
help="The file path to a binary file containing a binary key to "
233
"use for decryption. Can be a file path or S3 network URI.")
@@ -239,18 +241,21 @@ class TensorizerArgs:
241
"and model size. This greatly increases performance.")
242
group.add_argument(
243
"--s3-access-key-id",
244
+ type=str,
245
default=None,
246
help="The access key for the S3 bucket. Can also be set via the "
247
"S3_ACCESS_KEY_ID environment variable.",
248
)
249
group.add_argument(
250
"--s3-secret-access-key",
251
+ type=str,
252
default=None,
253
help="The secret access key for the S3 bucket. Can also be set via "
254
"the S3_SECRET_ACCESS_KEY environment variable.",
255
)
256
group.add_argument(
257
"--s3-endpoint",
258
+ type=str,
259
default=None,
260
help="The endpoint for the S3 bucket. Can also be set via the "
261
"S3_ENDPOINT_URL environment variable.",