ede

The v2 emulators are the emulators used in The Atacama Cosmology Telescope: DR6 Constraints on Extended Cosmological Models (2025).

These emulators were then used in Impact of ACT DR6 and DESI DR2 for Early Dark Energy and the Hubble tension by Poulin et al (2025).

They are the highest accuracy class-based emulators available on this organization and match the camb-based Jense et al. (2024) Emulators in LCDM well under 0.1-$\sigma$ where $\sigma$ is the typical 68%CL on parameter constraints using ACT DR6 data.

The emulators can be used in LCDM, mnu-LCDM, w-CDM and Neff-LCDM and combinations of these.

They emulate distances, TT, TE, EE, BB and P(k).

File formats: _v2.npz and _v2_plain.npz

Each emulator is provided in two on-disk formats with identical weights:

• <name>_v2.npz — the original CosmoPower output. A single arr_0 object array holding a pickled cosmopower_NN / cosmopower_PCAplusNN instance. Loading requires cosmopower to be importable (which transitively imports tensorflow / keras), since pickle has to reconstruct tensorflow.python.trackable.data_structures.ListWrapper references. Drop-in compatible with the CosmoPower loader API.

• <name>_v2_plain.npz — the same weights, re-saved with a flat np.savez_compressed(weights_.0=…, weights_.1=…, …, weights_.n=N, biases_.0=…, …) layout. Loads with allow_pickle=False and zero TensorFlow / CosmoPower dependencies; pure numpy. The flat-key convention uses the format <field>.<index> for list-valued fields (weights_, biases_, alphas_, betas_) plus a <field>.n sentinel for the count, and preserves scalars / 1-D arrays as-is.

The plain variant is ~8 % smaller (due to per-array compression of all the numeric blocks rather than the pickle protocol's blob compression).

When to prefer which

• Pipelines that already use CosmoPower — keep using _v2.npz; no change.

• Pipelines that load weights into a different framework (e.g. JAX, PyTorch, raw numpy) and don't want to install TensorFlow just to deserialise — use _v2_plain.npz. Example loader (~20 lines):

  import numpy as np
  npz = np.load("TT_v2_plain.npz", allow_pickle=False)
  def unflatten(saved):
      out = {}; groups = {}
      for k in saved.files:
          if "." in k:
              base, idx = k.rsplit(".", 1)
              if idx == "n": continue
              groups.setdefault(base, {})[int(idx)] = saved[k]
          else:
              v = saved[k]
              out[k] = v.item() if v.ndim == 0 and v.dtype.kind in "iuf" else v
      for base, idx_map in groups.items():
          out[base] = [idx_map[i] for i in sorted(idx_map)]
      return out
  d = unflatten(npz)
  # d["weights_"], d["biases_"], d["alphas_"], d["betas_"], d["param_train_mean"], ...