Update README.md

README.md

@@ -5,6 +5,16 @@ license: apache-2.0
 ---
 This is unsloth/llama-3-8b-Instruct trained on the Replete-AI/code-test-dataset using the code bellow with unsloth and google colab with under 15gb of vram. This training was complete in about 40 minutes total.
 
+For anyone that is new to coding and training Ai, all your really have to edit is
+
+1. (max_seq_length = 8192) To match the max tokens of the dataset or model you are using
+2. (model_name = "unsloth/llama-3-8b-Instruct",) Change what model you are finetuning, this setup is specifically for llama-3-8b
+3. (alpaca_prompt =) Change the prompt format, this one is setup to meet llama-3-8b-instruct format, but match it to your specifications. 
+4. (dataset = load_dataset("Replete-AI/code-test-dataset", split = "train")) What dataset you are using from huggingface
+5. (model.push_to_hub_merged("rombodawg/test_dataset_Codellama-3-8B", tokenizer, save_method = "merged_16bit", token = ""))
+    For the above you need to change "rombodawg" to your Hugginface name, "test_dataset_Codellama-3-8B" to the model name you want saved as, and in token = "" you need to put your huggingface write token so the model can be saved.  
+
+
 ```Python
 %%capture
 import torch
@@ -142,6 +152,7 @@ trainer = SFTTrainer(
 )
 ```
 
+
 ```Python
 trainer_stats = trainer.train()
 model.save_pretrained_merged("model", tokenizer, save_method = "merged_16bit",)