Update llava_olmo.py

llava_olmo.py

@@ -62,37 +62,3 @@ print(decoded_text[:decoded_text.find('</s>')].replace('|||IP_ADDRESS|||', ''))
 print("-"*100)
 
 
-#
-##
-#
-#
-#
-'''
-# ORIGINAL CODE WITH ONLY OLMO: 
-with open('llava/config.json') as json_file:
-    data = json.load(json_file)
-
-text = "Paris is a historic city with architectural marvels. It is also "
-# text = ["Language modeling is "]
-
-config_class = llava_olmo.LlavaOLMoBitnet1BConfig(**data)
-lolmo = llava_olmo.LlavaOLMoBitnet1BForCausalLM(config_class).to(device)
-lolmo.load_state_dict(torch.load('OLMo_Bitnet_1B/pytorch_model.bin'), strict=False)
-
-olmo = OLMoForCausalLM(config_class).to(device)
-olmo.load_state_dict(torch.load('OLMo_Bitnet_1B/pytorch_model.bin'))
-actual_olmo = OLMoForCausalLM.from_pretrained("allenai/OLMo-1B").to(device)
-
-actual_olmo_tokenizer = OLMoTokenizerFast.from_pretrained("allenai/OLMo-1B")
-olmo_tokenizer = AutoTokenizer.from_pretrained("NousResearch/OLMo-Bitnet-1B")
-
-olmo_tokens = olmo_tokenizer(text, return_tensors='pt', return_token_type_ids=False).to(device)
-# olmo_tokens = actual_olmo_tokenizer(text, return_tensors='pt', return_token_type_ids=False).to(device)
-
-
-response = lolmo.generate(inputs=olmo_tokens['input_ids'], attention_mask=olmo_tokens['attention_mask'], max_new_tokens=100, do_sample=True, top_k=50, top_p=0.95)
-# response = olmo.generate(inputs=olmo_tokens['input_ids'], attention_mask=olmo_tokens['attention_mask'], max_new_tokens=100, do_sample=True, top_k=50, top_p=0.95)
-
-
-print(olmo_tokenizer.batch_decode(response, skip_special_tokens=True)[0])
-'''