MCPcopy Create free account
hub / github.com/SalesforceAIResearch/perfcodegen / infer_one_chat

Method infer_one_chat

src/model.py:247–265  ·  view source on GitHub ↗
(self, dialog, n = 10)

Source from the content-addressed store, hash-verified

245
246
247 def infer_one_chat(self, dialog, n = 10):
248 prompt_tokens = self.format_chats(dialog)[0]
249 try:
250 if n <= 10:
251 outputs = self.model.generate(prompt_token_ids = prompt_tokens, sampling_params=self.sampling_params)
252 predictions = []
253 for o in outputs[0].outputs:
254 predictions.append(o.text)
255 else:
256 iters = math.ceil(n/10)
257 predictions = []
258 for i in range(0, iters):
259 outputs = self.model.generate(prompt, self.sampling_params)
260 for o in outputs[0].outputs:
261 predictions.append(o.text)
262 return predictions
263 except Exception as e:
264 logger.error("Error occurred with reason {} for prompt:\n{}".format(str(e), dialog[0]["content"]))
265 return str(e)
266
267
268

Callers 1

infer_local_modelFunction · 0.95

Calls 2

format_chatsMethod · 0.95
generateMethod · 0.80

Tested by

no test coverage detected