|
58 | 58 |
|
59 | 59 | # TODO: remove when https://github.com/huggingface/transformers/issues/28096 is addressed |
60 | 60 | 'RajuKandasamy/tamillama_tiny_30m', |
| 61 | + |
| 62 | + # TODO: remove when need for trust_remote_code can be addressed in CI |
| 63 | + 'monologg/kobert', |
61 | 64 | ] |
62 | 65 |
|
63 | 66 | MAX_TESTS = { |
@@ -269,15 +272,18 @@ def generate_tokenizer_tests(): |
269 | 272 | tokenizer = AutoTokenizer.from_pretrained( |
270 | 273 | tokenizer_name, |
271 | 274 | use_fast=False, |
| 275 | + trust_remote_code=True, |
272 | 276 | ) |
273 | 277 | decoder_tokenizer = AutoTokenizer.from_pretrained( |
274 | 278 | tokenizer_name, |
275 | 279 | use_fast=True, |
| 280 | + trust_remote_code=True, |
276 | 281 | ) |
277 | 282 |
|
278 | 283 | else: |
279 | 284 | decoder_tokenizer = tokenizer = AutoTokenizer.from_pretrained( |
280 | | - tokenizer_name) |
| 285 | + tokenizer_name, |
| 286 | + trust_remote_code=True) |
281 | 287 |
|
282 | 288 | except (KeyError, EnvironmentError): |
283 | 289 | # If a KeyError/EnvironmentError is raised from the AutoTokenizer, it |
@@ -329,6 +335,7 @@ def generate_tokenizer_tests(): |
329 | 335 |
|
330 | 336 | # TODO: Remove once https://github.com/huggingface/transformers/pull/26678 is fixed |
331 | 337 | use_fast='llama' not in tokenizer_id, |
| 338 | + trust_remote_code=True, |
332 | 339 | ) |
333 | 340 | tokenizer_results = [] |
334 | 341 | for key in TOKENIZERS_WITH_CHAT_TEMPLATES[tokenizer_id]: |
@@ -363,7 +370,7 @@ def generate_config_tests(): |
363 | 370 | print(' -', config_name) |
364 | 371 | try: |
365 | 372 | # Load config |
366 | | - config = AutoConfig.from_pretrained(config_name) |
| 373 | + config = AutoConfig.from_pretrained(config_name, trust_remote_code=True) |
367 | 374 | except Exception: |
368 | 375 | # Something went wrong, skip this config |
369 | 376 | continue |
|
0 commit comments