Update README.md

README.md

@@ -1,41 +1,42 @@
 ---
 language: en
 license: apache-2.0
+pipeline_tag: text-generation
+library_name: transformers
 tags:
 - text-generation
-- summarization
 - language-modeling
 - transformers
 - from-scratch
-pipeline_tag: text-generation
-library_name: transformers
+model_name: Genesis-100M
+---
 
-# Custom 100M Parameter Language Model
+# Genesis-100M
 
-## Overview
-A GPT-style decoder-only Transformer trained from scratch on news text
-using a custom BPE tokenizer.
+Genesis-100M is a 100M-parameter GPT-style language model trained from scratch on news text using a custom BPE tokenizer.
 
 ## Architecture
-- Layers: 12
+- Decoder-only Transformer (GPT-style)
+- 12 layers
 - Hidden size: 768
-- Heads: 12
+- Attention heads: 12
 - Context length: 512
 - Parameters: ~100M
 
 ## Training
-- Dataset: News articles (CNN/DailyMail, articles only)
-- Training method: Causal language modeling
+- Dataset: News articles (CNN/DailyMail – articles only)
+- Objective: Causal Language Modeling
 - Hardware: Google Colab GPU
 - Precision: FP16
-- Steps: 2000
+- Training steps: 2000
+- Optimizations: Gradient checkpointing, gradient accumulation
 
 ## Intended Use
 - Research
 - Educational purposes
-- Prompt-based text generation / summarization
+- Text generation experiments
 
 ## Limitations
 - Not instruction-tuned
 - Trained for limited steps
-- Outputs may be verbose or inconsistent
+- Outputs may be verbose or repetitive