Guide graduate students through iPEPS calculations for 2D quantum lattice models using Tenax. Covers the full pipeline: simple update (imaginary time evolution), AD-based ground-state optimization via optimize_gs_ad with CTM environments, and quasiparticle excitation spectra via compute_excitations. Supports 1-site and 2-site unit cells. Use this skill when the user mentions iPEPS, PEPS, 2D tensor networks, projected entangled pair states, corner transfer matrix (CTM), simple update, imaginary time evolution for 2D, AD optimization of tensor networks, quasiparticle excitations, or Brillouin zone dispersion. Also trigger for "2D Heisenberg ground state", "Neel order", "2D phase diagram", or any request to go beyond 1D DMRG to 2D systems.
Help users design and run performance benchmarks for Tenax algorithms across hardware backends (CPU, CUDA, TPU, Metal). Generates the correct CLI commands for python -m benchmarks.run, interprets JSON/CSV output, and helps compare results across backends and algorithm configurations. Use this skill when the user mentions benchmarks, performance, timing, speed comparison, GPU vs CPU, profiling, or wants to measure how fast DMRG/iDMRG/TRG/iPEPS runs. Also trigger for "how fast is", "performance test", "scaling study", or "which backend is faster".
Diagnose and fix errors in Tenax tensor network code running on JAX. Covers four categories: (1) shape mismatches and tensor dimension errors, (2) JAX tracing and jit compilation issues, (3) gradient and autodiff problems, and (4) convergence failures in DMRG, iDMRG, TRG, iPEPS, and other algorithms. Use this skill whenever the user pastes a Tenax traceback, mentions a runtime error involving tensors or contractions, reports NaNs or non-converging energy, asks "why is my DMRG not converging", complains about slow JAX compilation, or shares code that produces unexpected shapes. Also trigger when the user says "debug", "error", "broken", "wrong shape", "NaN", or "doesn't converge" in the context of tensor networks, Tenax, tenax, or JAX. Trigger for SymmetricTensor errors, label mismatch, FlowDirection issues, or AutoMPO build failures.
Guide graduate students through a complete DMRG ground-state calculation using Tenax, from defining the Hamiltonian through AutoMPO to analyzing results. Pedagogical and step-by-step. Covers finite DMRG, iDMRG (1D chain and infinite cylinder), and 2D cylinder DMRG. Use this skill whenever the user wants to run DMRG, find a ground state with tensor networks, set up an MPS calculation, build an MPO Hamiltonian, or asks about DMRG sweeps, bond dimensions, canonical forms, or convergence — especially in the context of Tenax or tenax. Also trigger for "ground state energy", "variational MPS", "finite DMRG", "infinite DMRG", "iDMRG", "cylinder DMRG", "AutoMPO", "Heisenberg chain", "build MPO", or "how do I use DMRG for [model]".
Help users migrate tensor network code from ITensor (Julia/C++) to Tenax. Maps ITensor concepts (Index, ITensor, MPS, MPO, AutoMPO, siteinds) to Tenax equivalents. Use this skill when the user mentions ITensor, "coming from ITensor", "ITensor equivalent", "Julia tensor networks", or asks how Tenax compares to ITensor. Also trigger for "migrate from ITensor" or "I used to use ITensor".
Help users migrate tensor network code from TeNPy to Tenax. Maps TeNPy concepts (Site, MPS, MPO, Model, Engine) to Tenax equivalents. Use this skill when the user mentions TeNPy, "coming from TeNPy", "TeNPy equivalent", or asks how Tenax compares to TeNPy. Also trigger for "migrate from TeNPy", "I used to use TeNPy", or "tenpy".
Guide users through computing physical observables from Tenax ground states: local expectation values, correlation functions, entanglement entropy, and order parameters from MPS (DMRG/iDMRG) and iPEPS states. Use this skill when the user asks "how do I compute observables", "measure correlation function", "entanglement entropy", "expectation value", "order parameter", "magnetization", "structure factor", or "what can I measure from my ground state". Also trigger for "post-DMRG analysis", "reduced density matrix", or "how to extract physics from MPS".
Teach users Tenax's core tensor operations: creating DenseTensor and SymmetricTensor, label-based contraction, truncated SVD, full SVD, eigendecomposition (eigh), QR decomposition, and the TensorNetwork graph container. Use this skill when the user asks about basic tensor operations, "how does contraction work", "how to do SVD", "eigendecomposition", "create a tensor", "what are labels", DenseTensor, contract, truncated_svd, svd, eigh, qr_decompose, or wants to understand the building blocks before running algorithms. Also trigger for "tensor basics", "label-based contraction", "how do I contract two tensors", "entanglement spectrum", or "diagonalize a density matrix".