add bloom api support

prompt_lib/backends/bloom_api.py

@@ -0,0 +1,65 @@
+from cmath import log
+import requests
+from pprint import pprint
+import json
+import sys
+import logging
+from huggingface_hub import InferenceApi
+import time
+import os
+hf_token = os.getenv('HF_API_TOKEN')
+logging.basicConfig(level=logging.INFO)
+
+# adapted from https://github.com/Sentdex/BLOOM_Examples/blob/main/BLOOM_api_example.ipynb
+
+class BloomWrapper:
+    @staticmethod
+    def call_bloom_api(
+        prompt: str, 
+        max_length: int = 248, 
+        top_k: float = 0, 
+        num_beams: int = 0,
+        no_repeat_ngram_size: int = 0,
+        top_p: float = 0.9,
+        seed: int = 42, 
+        temperature: float = 0.7,
+        greedy_decoding: bool = False,
+        return_full_text: bool = False
+    ) -> str:
+
+        top_k = None if top_k == 0 else top_k
+        do_sample = False if num_beams > 0 else not greedy_decoding
+        num_beams = None if (greedy_decoding or num_beams == 0) else num_beams
+        no_repeat_ngram_size = None if num_beams is None else no_repeat_ngram_size
+        top_p = None if num_beams else top_p
+        early_stopping = None if num_beams is None else num_beams > 0
+
+        params = {
+            "max_new_tokens": max_length,
+            "top_k": top_k,
+            "top_p": top_p,
+            "temperature": temperature,
+            "do_sample": do_sample,
+            "seed": seed,
+            "early_stopping":early_stopping,
+            "no_repeat_ngram_size":no_repeat_ngram_size,
+            "num_beams":num_beams,
+            "return_full_text":return_full_text
+        }
+
+        s = time.time()
+        inference = InferenceApi("bigscience/bloom",token=hf_token)
+        response = inference(prompt, params=params)
+        return response
+
+
+
+if __name__ == "__main__":
+    response = BloomWrapper.call_bloom_api(
+        prompt="The thing that makes large language models interesting is",
+        max_length=248, 
+        num_beams=0,
+        greedy_decoding=True
+    )
+
+    print(response)