Revert: TODO's

Author: rahul-tuli

examples/llama-3/sft.yaml

@@ -1,6 +1,4 @@
-base_model: "nm-testing/llama2.c-stories42M-gsm8k-sparse-only-uncompressed"
-# TODO: change to
-# base_model: neuralmagic/Sparse-Llama-3.1-8B-2of4
+base_model: neuralmagic/Sparse-Llama-3.1-8B-2of4
 
 plugins:
   - axolotl.integrations.llmcompressor_sft.SFTPlugin
@@ -27,7 +25,7 @@ wandb_watch:
 wandb_name:
 wandb_log_model:
 
-# gradient_accumulation_steps: 8
+gradient_accumulation_steps: 8
 micro_batch_size: 1
 num_epochs: 1
 optimizer: paged_adamw_8bit