Upload runners/NVIDIA_H200_TITAN_PRETRAIN.txt with huggingface_hub

runners/NVIDIA_H200_TITAN_PRETRAIN.txt

@@ -15,15 +15,16 @@
 #    • Auto-uploads ckpt_latest.pt to Hugging Face in the background every 250 steps.
 # ==============================================================================
 
-!pip install -q torch transformers datasets tokenizers huggingface_hub numpy
+!pip install -q torch transformers datasets tokenizers huggingface_hub numpy hf_transfer
 
 import os, sys, shutil
 import torch
 from huggingface_hub import HfApi, hf_hub_download
 
-# Real-time unbuffered logging & memory allocator protection
+# Real-time unbuffered logging, Rust-accelerated downloads & memory safety
 os.environ["PYTHONUNBUFFERED"] = "1"
 os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True"
+os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
 
 HF_TOKEN = os.environ.get("HF_TOKEN") or ''.join(['hf_','ssyCVhuny','XxjGdqKp','VLPpkmWK','FrrMOIFbg'])
 os.environ["HF_TOKEN"] = HF_TOKEN
@@ -35,8 +36,9 @@ print("=" * 80)
 print("🚀 ViuAI Sarus-500M — Launching NVIDIA H200 141GB Titan Pretraining")
 print("=" * 80)
 
-# Continuation Pretraining Hyperparameters
-os.environ["TOTAL_STEPS"] = "60000"       # 20,000 continuation steps (~10B tokens)
+# Continuation Pretraining Hyperparameters (Full Run to 60,000 Steps)
+os.environ["TOTAL_STEPS"] = "60000"       # Full 20,000 continuation steps (~10B tokens)
+os.environ["MAX_STEPS"] = "60000"         # Full run to Step 60,000
 os.environ["MAX_LR"] = "1.0e-4"           # Calibrated continuation LR to protect base weights
 os.environ["MIN_LR"] = "1.0e-5"
 os.environ["WARMUP_STEPS"] = "500"