ViuAI commited on
Commit
71e7e36
·
verified ·
1 Parent(s): d90246b

Upload runners/RTX_5090_BLACKWELL_PRETRAIN.txt with huggingface_hub

Browse files
runners/RTX_5090_BLACKWELL_PRETRAIN.txt CHANGED
@@ -13,16 +13,19 @@
13
  # • Auto-uploads ckpt_latest.pt to Hugging Face in the background every 250 steps.
14
  # ==============================================================================
15
 
16
- !pip install -q torch transformers datasets tokenizers huggingface_hub numpy hf_transfer
 
 
17
 
18
  import os, sys, shutil
19
  import torch
20
  from huggingface_hub import HfApi, hf_hub_download
21
 
22
- # Real-time unbuffered logging, Rust-accelerated downloads & memory allocator protection
23
  os.environ["PYTHONUNBUFFERED"] = "1"
24
  os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True"
25
  os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
 
26
 
27
  HF_TOKEN = os.environ.get("HF_TOKEN") or ''.join(['hf_','ssyCVhuny','XxjGdqKp','VLPpkmWK','FrrMOIFbg'])
28
  os.environ["HF_TOKEN"] = HF_TOKEN
 
13
  # • Auto-uploads ckpt_latest.pt to Hugging Face in the background every 250 steps.
14
  # ==============================================================================
15
 
16
+ # Ensure PyTorch with CUDA 12.8 (sm_120 Blackwell support) is installed
17
+ !pip install -q --upgrade --index-url https://download.pytorch.org/whl/cu128 torch
18
+ !pip install -q transformers datasets tokenizers huggingface_hub numpy hf_transfer
19
 
20
  import os, sys, shutil
21
  import torch
22
  from huggingface_hub import HfApi, hf_hub_download
23
 
24
+ # Real-time unbuffered logging, Rust-accelerated downloads & Blackwell compatibility
25
  os.environ["PYTHONUNBUFFERED"] = "1"
26
  os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True"
27
  os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
28
+ os.environ["TORCH_CUDA_ARCH_LIST"] = "12.0;9.0+PTX"
29
 
30
  HF_TOKEN = os.environ.get("HF_TOKEN") or ''.join(['hf_','ssyCVhuny','XxjGdqKp','VLPpkmWK','FrrMOIFbg'])
31
  os.environ["HF_TOKEN"] = HF_TOKEN