MCPcopy Create free account
hub / github.com/THUDM/GLM / process_line

Method process_line

data_utils/corpora.py:453–466  ·  view source on GitHub ↗
(self, data, tokenizer, tokenize)

Source from the content-addressed store, hash-verified

451 info.put(source_dict)
452
453 def process_line(self, data, tokenizer, tokenize):
454 source = data["meta"].get("pile_set_name", None)
455 text = data.get("text", None)
456 if source and text:
457 if source in self.filtered_sources:
458 return [], [], None
459 elif source in self.downsample_sources and random.random() > self.downsample_sources[source]:
460 return [], [], None
461 else:
462 prompt, text = self.process_sample("", tokenizer, tokenize), self.process_sample(text, tokenizer,
463 tokenize)
464 return [prompt], [text], source
465 else:
466 return [], [], None
467
468
469class Stories(PromptReader):

Callers 2

tokenize_workerMethod · 0.95
tokenize_workerMethod · 0.45

Calls 1

process_sampleMethod · 0.80

Tested by

no test coverage detected