Update app.py

app.py

@@ -31,7 +31,7 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 
 # model_id = "mistralai/Mistral-7B-v0.3"
 
-model_id = "CohereForAI/aya-23-35B"
+model_id = "CohereForAI/aya-23-8B"
 
 
 tokenizer = AutoTokenizer.from_pretrained(
@@ -46,9 +46,9 @@ with init_empty_weights():
                                              torch_dtype=torch.float16,
                                               # torch_dtype=torch.fl,
                                              # attn_implementation="flash_attention_2",
-                                             # low_cpu_mem_usage=True,
+                                             low_cpu_mem_usage=True,
                                              # llm_int8_enable_fp32_cpu_offload=True,
-                                             # device_map='cuda',
+                                             device_map='cuda',
                                              
                                             )