Update README.md

README.md

@@ -1,3 +1,59 @@
+---
+model_name: mistral-base
+tokenizer_name: mistral-base
+lora:
+  r: 32
+  alpha: 64
+  dropout: 0.05
+  target_modules:
+  - q_proj
+  - k_proj
+  - v_proj
+  - o_proj
+  - gate_proj
+  - up_proj
+  - down_proj
+training:
+  output_dir: ./lora-out
+  batch_size: 4
+  micro_batch_size: 1
+  gradient_accumulation_steps: 4
+  learning_rate: 0.0002
+  weight_decay: 0
+  warmup_ratio: 0.03
+  num_epochs: 3
+  max_grad_norm: 1
+  fp16: true
+dataset:
+  path: ./data
+  field: text
+  shuffle: true
+  num_workers: 4
+generation:
+  bos_token_id: 1
+  eos_token_id: 2
+  max_new_tokens: 256
+  temperature: 0.7
+  top_p: 0.9
+logging:
+  steps: 10
+  save_steps: 200
+license: apache-2.0
+language:
+- en
+base_model:
+- mistralai/Mistral-7B-Instruct-v0.2
+pipeline_tag: text-generation
+tags:
+- merge
+- lora
+---
+
+
+
+
+
+
 # Mistral + LoRA Fine-Tuning
 
 Lightweight fine-tuning setup for adapting a Mistral-architecture model using LoRA.  
@@ -50,4 +106,4 @@ print(tok.decode(out[0]))
 - `tokenizer_config.json` / `special_tokens_map.json` – tokenizer metadata  
 
 ## License
-MIT
+MIT