From b4538c013bbd318c887a1f1586967a49c6d506f0 Mon Sep 17 00:00:00 2001
From: Christian Medina <37550954+cmedinasoriano@users.noreply.github.com>
Date: Fri, 3 Jul 2026 00:15:39 -0400
Subject: [PATCH] fix: use 1-indexing for shard names (model-00001-of-00016)

---
 quantize_streaming.py | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/quantize_streaming.py b/quantize_streaming.py
index eac58e8..a0f1f4b 100644
--- a/quantize_streaming.py
+++ b/quantize_streaming.py
@@ -43,11 +43,11 @@ def streaming_quantize(model_path: str, output_path: str):
     
     def process_shard(idx, shard_file, gpu_id):
         """Process a single shard (called per-GPU)."""
-        shard_name = f"model-{idx:05d}-of-{len(shards):05d}.safetensors"
+        shard_name = f"model-{idx+1:05d}-of-{len(shards):05d}.safetensors"
         if (output_path / shard_name).exists():
             return
         
-        print(f"[{idx}/{len(shards)}] {Path(shard_file).name} (GPU {gpu_id})")
+        print(f"[{idx+1}/{len(shards)}] {Path(shard_file).name} (GPU {gpu_id})")
         
         # Load to assigned GPU
         state_dict = load_file(shard_file, device=f"cuda:{gpu_id}")
@@ -88,7 +88,7 @@ def streaming_quantize(model_path: str, output_path: str):
     # Find first unsaved shard
     start_idx = 0
     for idx in range(len(shards)):
-        shard_name = f"model-{idx:05d}-of-{len(shards):05d}.safetensors"
+        shard_name = f"model-{idx+1:05d}-of-{len(shards):05d}.safetensors"
         if not (output_path / shard_name).exists():
             start_idx = idx
             break