From 05704d0eb2fa59fe727652465b07db40bcb06c38 Mon Sep 17 00:00:00 2001 From: Yuren Hao Date: Wed, 8 Apr 2026 22:06:05 -0500 Subject: Initial release: GAP framework - Full pipeline: variant generation, multi-judge verification, evaluation - Loaders for OpenAI / Anthropic / Google / xAI / OpenRouter / vLLM - Framework-level mechanism analyses: paired structural overlap, repairability rescue, self-correction probe, cross-model agreement, topic x problem-type interaction - Unicode -> bare-LaTeX cleaner + audit + spot-check - Mirrors https://huggingface.co/datasets/blackhao0426/PutnamGAP --- putnam-bench-anon/requirements-local.txt | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+) create mode 100644 putnam-bench-anon/requirements-local.txt (limited to 'putnam-bench-anon/requirements-local.txt') diff --git a/putnam-bench-anon/requirements-local.txt b/putnam-bench-anon/requirements-local.txt new file mode 100644 index 0000000..fbcab3f --- /dev/null +++ b/putnam-bench-anon/requirements-local.txt @@ -0,0 +1,27 @@ +# Requirements for local model support (GPU recommended) +# Base requirements +-r requirements-minimal.txt + +# PyTorch with CUDA support (install separately with specific CUDA version) +# pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124 + +# HuggingFace transformers ecosystem +transformers>=4.30.0 +accelerate>=0.20.0 +tokenizers>=0.13.0 + +# Optional: VLLM for high-performance inference +vllm>=0.3.0 + +# GPU monitoring and management +nvidia-ml-py3>=11.0.0 + +# Model quantization and optimization +bitsandbytes>=0.39.0 + +# Additional utilities for local models +safetensors>=0.3.0 +huggingface-hub>=0.16.0 + +# Progress bars +tqdm>=4.60.0 \ No newline at end of file -- cgit v1.2.3