Update README.md

tuned and updated some stuff in the finetune notebook

README.md

@@ -20,13 +20,9 @@ The following `bitsandbytes` quantization config was used during training:
 - bnb_4bit_compute_dtype: float16
 ### Framework versions
 
-
 - PEFT 0.6.0.dev0
 
-- # -*- coding: utf-8 -*-
-"""bf16_sharded_Fine_Tuning_using_QLora(1).ipynb
-
-Automatically generated by Colaboratory.
+"""
 
 Original file is located at
     https://colab.research.google.com/drive/1yH0ov1ZDpun6yGi19zE07jkF_EUMI1Bf
@@ -63,6 +59,7 @@ wandb_key=["<API_KEY>"]
 wandb.init(project="<project_name>",
           name="<name>"
 )
+
 # login with API
 from huggingface_hub import login
 login()
@@ -127,11 +124,11 @@ output_dir = "./results"
 per_device_train_batch_size = 4
 gradient_accumulation_steps = 4
 optim = "paged_adamw_32bit"
-save_steps = 10
-logging_steps = 11
+save_steps = 100
+logging_steps = 10
 learning_rate = 2e-4
 max_grad_norm = 0.3
-max_steps = 10
+max_steps = 100
 warmup_ratio = 0.03
 lr_scheduler_type = "constant"
 
@@ -175,16 +172,12 @@ for name, module in trainer.model.named_modules():
 
 """## Train the model
 You're using a LlamaTokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.
-
 Now let's train the model! Simply call `trainer.train()`
 """
 
 trainer.train()
 
 """During training, the model should converge nicely as follows:
-
-![image](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/loss-falcon-7b.png)
-
 The `SFTTrainer` also takes care of properly saving only the adapters during training instead of saving the entire model.
 """