Implement gpt3 greedy_until

lm_eval/models/gpt3.py

@@ -25,6 +25,7 @@ class GPT3LM(LM):
 
     MAX_LENGTH = 2048
     REQ_CHUNK_SIZE = 64
+    MAX_GEN_TOKS = 256
 
     def __init__(self, engine, truncate=False):
         """
@@ -48,9 +49,6 @@ class GPT3LM(LM):
         return cls(engine=args.get("engine", "davinci"))
 
     def loglikelihood(self, requests):
-        import openai
-        res = []
-
         for chunk in tqdm(utils.chunks(requests, self.REQ_CHUNK_SIZE)):
             inps = []
             ctxlens = []
@@ -78,5 +76,23 @@ class GPT3LM(LM):
         return res
 
     def greedy_until(self, requests):
-        # TODO: implement
-        pass
\ No newline at end of file
+        import openai
+        res = []
+
+        for context, until in tqdm(requests):
+            context_enc = self.tokenizer.encode(context)
+            inp = context_enc[-(self.MAX_LENGTH - self.MAX_GEN_TOKS):]
+            ctxlen = len(context_enc) - max(0, len(context_enc) - (self.MAX_LENGTH - self.MAX_GEN_TOKS))
+
+            response = openai.Completion.create(
+                engine=self.engine,
+                prompt=[inp],
+                max_tokens=self.MAX_GEN_TOKS, 
+                temperature=0.,
+                logprobs=10,
+            )
+
+            res.append(response.choices[0]['text'])
+        
+        return res
+