Update README.md

README.md

@@ -1,7 +1,7 @@
 # Instruction from Wanwan Feng
-# 🚀 VCC Processing Workflow
 
 ## **Step 1 — Generate Embeddings for All Input Files**
+## This step shows the creation of this SE600M-embedding dataset from the existing training dataset provided by the Arc Institute for the Virtual Cell Challenge.
 ```python
 import os
 os.environ['MPLBACKEND'] = 'Agg'
@@ -43,6 +43,7 @@ for filename in FILENAMES_TO_PROCESS:
 ---
 
 ## **Step 2 — Train Model**
+## This step shows how to use the generated SE600M embedding dataset to train a State Transition (ST) model.
 ```bash
 state tx train   data.kwargs.toml_config_path="vci_pretrain/starter.toml"   data.kwargs.embed_key="X_state"   data.kwargs.num_workers=4   data.kwargs.batch_col="batch_var"   data.kwargs.pert_col="target_gene"   data.kwargs.cell_type_key="cell_type"   data.kwargs.control_pert="non-targeting"   data.kwargs.perturbation_features_file="vci_pretrain/ESM2_pert_features.pt"   training.max_steps=8000   training.ckpt_every_n_steps=1000   model=tahoe_sm   wandb.tags="[first_run]"   output_dir="competition_state"   name="first_run"
 ```