MCPcopy
hub / github.com/huggingface/transformers / main

Function main

examples/pytorch/text-classification/run_xnli.py:192–442  ·  view source on GitHub ↗
()

Source from the content-addressed store, hash-verified

190
191
192def main():
193 # See all possible arguments in src/transformers/training_args.py
194 # or by passing the --help flag to this script.
195 # We now keep distinct sets of args, for a cleaner separation of concerns.
196
197 parser = HfArgumentParser((ModelArguments, DataTrainingArguments, TrainingArguments))
198 model_args, data_args, training_args = parser.parse_args_into_dataclasses()
199
200 # Setup logging
201 logging.basicConfig(
202 format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
203 datefmt="%m/%d/%Y %H:%M:%S",
204 handlers=[logging.StreamHandler(sys.stdout)],
205 )
206
207 if training_args.should_log:
208 # The default of training_args.log_level is passive, so we set log level at info here to have that default.
209 transformers.utils.logging.set_verbosity_info()
210
211 log_level = training_args.get_process_log_level()
212 logger.setLevel(log_level)
213 datasets.utils.logging.set_verbosity(log_level)
214 transformers.utils.logging.set_verbosity(log_level)
215 transformers.utils.logging.enable_default_handler()
216 transformers.utils.logging.enable_explicit_format()
217
218 # Log on each process the small summary:
219 logger.warning(
220 f"Process rank: {training_args.local_process_index}, device: {training_args.device}, n_gpu: {training_args.n_gpu}, "
221 + f"distributed training: {training_args.parallel_mode.value == 'distributed'}, 16-bits training: {training_args.fp16}"
222 )
223 logger.info(f"Training/evaluation parameters {training_args}")
224
225 # Set seed before initializing model.
226 set_seed(training_args.seed)
227
228 # In distributed training, the load_dataset function guarantees that only one local process can concurrently
229 # download the dataset.
230 # Downloading and loading xnli dataset from the hub.
231 if training_args.do_train:
232 if model_args.train_language is None:
233 train_dataset = load_dataset(
234 "xnli",
235 model_args.language,
236 split="train",
237 cache_dir=model_args.cache_dir,
238 token=model_args.token,
239 )
240 else:
241 train_dataset = load_dataset(
242 "xnli",
243 model_args.train_language,
244 split="train",
245 cache_dir=model_args.cache_dir,
246 token=model_args.token,
247 )
248 label_list = train_dataset.features["label"].names
249

Callers 1

run_xnli.pyFile · 0.70

Calls 15

trainMethod · 0.95
save_modelMethod · 0.95
evaluateMethod · 0.95
predictMethod · 0.95
is_world_process_zeroMethod · 0.95
HfArgumentParserClass · 0.90
set_seedFunction · 0.90
TrainerClass · 0.90
get_process_log_levelMethod · 0.80
setLevelMethod · 0.80

Tested by

no test coverage detected