Dev/add delphi

algorithms/delphi.py

@@ -0,0 +1,38 @@
+from transformers import T5Config, T5ForConditionalGeneration, T5Tokenizer
+import torch
+import os
+
+class Delphi:
+    def __init__(self, device="cuda", model="large"):  # device parameter instead of device_id
+        self.device = torch.device(device)  # use the provided device
+        print(f"Delphi device: {self.device}", model)
+
+        if model == "large":
+            self.MODEL_LOCATION = "../models/delphi-large"
+            self.MODEL_BASE = "t5-large"
+
+        elif model == "11b":
+            self.MODEL_LOCATION = "../models/delphi-11b"
+            self.MODEL_BASE = "t5-11b"
+
+        else:
+            raise ValueError("Model should be either 'large' or '11b'")
+
+        # verify that the model exists
+        if not os.path.exists(self.MODEL_LOCATION):
+            raise ValueError("Model file does not exist: {}".format(self.MODEL_LOCATION))
+
+        self.load_model()
+
+    def load_model(self):
+        self.model = T5ForConditionalGeneration.from_pretrained(self.MODEL_LOCATION)
+        self.model.to(self.device)
+        self.tokenizer = T5Tokenizer.from_pretrained(self.MODEL_BASE, model_max_length=512)
+
+    def run_inference(self, input_string):
+        input_ids = self.tokenizer(input_string, return_tensors='pt').to(self.device).input_ids
+        outputs = self.model.generate(input_ids, max_length=200)
+
+        decoded_outputs = self.tokenizer.decode(outputs[0])
+
+        return decoded_outputs

baseline_system_local_files.py

@@ -4,11 +4,14 @@
 
 from algorithms.llm_baseline import LLMBaseline
 from algorithms.llama_index import LlamaIndex
+from algorithms.delphi import Delphi
 from utils.enums import ProbeType
 from prompt_engineering.common import build_casualties_string, prepare_prompt
 from similarity_measures.bert import force_choice_with_bert
 
 
+
+
 def main():
     parser = argparse.ArgumentParser(
         description="Simple LLM baseline system running against local files")
@@ -95,6 +98,9 @@ def run_baseline_system_local_filepath(
         algorithm = LlamaIndex(
             device="cuda", model_name=model,
             **algorithm_kwargs_parsed)
+    elif algorithm == "delphi":
+        algorithm = Delphi(
+        device="cuda", model=model)
 
     algorithm.load_model()