File size: 359 Bytes
ef44a3b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 | # Core — exact versions used in the paper experiments torch==2.11.0 sentencepiece==0.2.1 numpy==2.4.2 # AWS (SageMaker + S3) boto3==1.42.93 sagemaker==3.10.0 # HuggingFace (external datasets and baselines) datasets>=2.19.0 transformers>=4.40.0 peft>=0.10.0 # Evaluation scikit-learn>=1.3.0 # GGUF export (optional, for llama.cpp / Ollama) # gguf>=0.6.0 |