From 52b93837f91d460f3ccf558151cf8ecc1a23a0c6 Mon Sep 17 00:00:00 2001
From: Christian Medina <37550954+cmedinasoriano@users.noreply.github.com>
Date: Thu, 2 Jul 2026 22:00:15 -0400
Subject: [PATCH] fix: use NF4 naming convention for quantized model

---
 quantize_model.py                     | 2 +-
 training/configs/ornith-35b-lora.yaml | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/quantize_model.py b/quantize_model.py
index c1d1f3c..78219fd 100644
--- a/quantize_model.py
+++ b/quantize_model.py
@@ -41,7 +41,7 @@ if __name__ == "__main__":
     import argparse
     parser = argparse.ArgumentParser()
     parser.add_argument("--model-path", type=str, default="/data/models/Ornith-1.0-35B")
-    parser.add_argument("--output-path", type=str, default="/data/models/Ornith-1.0-35B-4bit-nf4")
+    parser.add_argument("--output-path", type=str, default="/data/models/Ornith-1.0-35B-nf4")
     args = parser.parse_args()
     
     quantize_model(args.model_path, args.output_path)
diff --git a/training/configs/ornith-35b-lora.yaml b/training/configs/ornith-35b-lora.yaml
index 2cf394e..86d5988 100644
--- a/training/configs/ornith-35b-lora.yaml
+++ b/training/configs/ornith-35b-lora.yaml
@@ -1,7 +1,7 @@
 # LoRA Training Configuration for Ornith-1.0-35B
 # Dataset: cyron_summary_lora_dataset (20k examples)
 
-base_model: /data/models/Ornith-1.0-35B
+base_model: /data/models/Ornith-1.0-35B-nf4
 model_type: Qwen3_5MoeForCausalLM
 tokenizer_type: AutoTokenizer