DNGros

ELECTION TODAY
Base frac dem: 0.5156333333333334
Winning maps:
Unique sets 255
states, P(D_wins_states), P(D_win_states ∩ D_win_nationally), P(D_win_states | D_win_nationally), P(D_win_nationally | D_win_states)
{MI}, 58.0%, 48.0%, 93.0%, 82.7%
{PA}, 53.2%, 47.4%, 91.8%, 89.0%
{WI}, 56.7%, 46.4%, 90.0%, 81.8%
{MI, PA}, 46.4%, 44.8%, 86.9%, 96.6%
{MI, WI}, 48.4%, 44.1%, 85.6%, 91.2%

DNGros

{
  "attention_probs_dropout_prob": 0.1, 
  "directionality": "bidi", 
  "hidden_act": "gelu", 
  "hidden_dropout_prob": 0.1, 
  "hidden_size": 1024, 
  "initializer_range": 0.02, 
  "intermediate_size": 4096, 
  "max_position_embeddings": 512, 
  "num_attention_heads": 16, 

DNGros

from pathlib import Path
from tqdm import tqdm
import pandas as pd

# Get the manifests from gs://cubert/20200621_Python/github_python_minus_ethpy150open_deduplicated_manifest
cubert_py_manifests_root = Path("REPLACE_WITH_PATH_WHERE_DOWNLOAD_MANIFESTS")

def get_manifest_df(num_files_limit: int = None) -> pd.DataFrame:
    all_manifests = list(cubert_py_manifests_root.iterdir())
    taken_manifests = all_manifests[:min(num_files_limit or 999999, len(all_manifests))]

DNGros

"""This code tries to address the desire to have different forwards with
different signatures and supporting static analysis / IDE hinting.

Example:
    class MyModule(MultiforwardTorchModule):
        @add_hooks
        def forward_train(
            hidden_state: torch.Tensor,
            teacher_force_seq: List[str]
            # training specific args...