MCPcopy Create free account
hub / github.com/PAIR-code/lit / fit_transform

Method fit_transform

lit_nlp/lib/caching.py:252–274  ·  view source on GitHub ↗

Cache projections from ProjectorModel dimensionality reducers.

(self, inputs: Iterable[JsonDict])

Source from the content-addressed store, hash-verified

250 ##
251 # For internal use
252 def fit_transform(self, inputs: Iterable[JsonDict]):
253 """Cache projections from ProjectorModel dimensionality reducers."""
254 wrapped = self.wrapped
255 if not isinstance(wrapped, lit_model.ProjectorModel):
256 raise TypeError(
257 "Attempted to call fit_transform() on a non-ProjectorModel."
258 )
259
260 inputs_as_list = list(inputs)
261 cache_keys = [self.key_fn(d) for d in inputs_as_list]
262 if (none_keys := [k for k in cache_keys if k is None]):
263 logging.warning(
264 "Attmepting to cache %d (of %d) where the cache key is None "
265 "- this can be from a missing or empty example id. These"
266 " will be recomputed on subsequent attempts.",
267 len(none_keys),
268 len(cache_keys),
269 )
270 outputs = list(wrapped.fit_transform(inputs_as_list))
271 with self._cache.lock:
272 for cache_key, output in zip(cache_keys, outputs):
273 self._cache.put(output, cache_key)
274 return outputs
275
276 def predict(self,
277 inputs: Iterable[JsonDict],

Callers

nothing calls this directly

Calls 2

key_fnMethod · 0.95
putMethod · 0.80

Tested by

no test coverage detected