Update AI_Agent/llm_adapters/hf_adapter.py

AI_Agent/llm_adapters/hf_adapter.py

@@ -1,20 +1,21 @@
+# AI_Agent/llm_adapters/hf_adapter.py
 from transformers import AutoModelForCausalLM, AutoTokenizer
 import torch
 import asyncio
 
 class HuggingFaceAdapter:
-    def __init__(self, model_name="tiiuae/falcon-7b-instruct"):
+    def __init__(self, model_name="google/gemma-3n-E2B-it"):
         self.model_name = model_name
         self.tokenizer = AutoTokenizer.from_pretrained(model_name)
         self.model = AutoModelForCausalLM.from_pretrained(
             model_name,
-            dtype=torch.float16,       # updated from torch_dtype to dtype
-            device_map="auto"          # requires accelerate
+            dtype=torch.float32,  # CPU-friendly
+            device_map=None       # CPU only
         )
 
-    async def generate(self, prompt: str, max_tokens=500):
+    async def generate(self, prompt: str, max_tokens=300):
         def _sync_generate():
-            inputs = self.tokenizer(prompt, return_tensors="pt").to(self.model.device)
+            inputs = self.tokenizer(prompt, return_tensors="pt")  # no .to(self.model.device) needed
             outputs = self.model.generate(**inputs, max_new_tokens=max_tokens)
             text = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
             return text