import numpy as np
from general_tools.utils import get_root, get_max_root
ROOT = get_root("internn")
import sys
sys.path.append(str(ROOT))
sys.path.append(str(ROOT / "data"))
print(sys.path)
from internn_utils import *
from pytorch_utils import *
from sen_loader import get_text
from error_measures import *
eps = 1e-7
def sample_to_text(sample, output):
text = [s.lower() for s in sample["text"]]
out_text = [
get_text(o.argmax(-1))[:sample["length"][i]]
for i, o in enumerate(output)
]
return text, out_text
def cer_index(sample, output, index, **kwargs):
"""
Args:
sample:
output:
index: 2D array, Batch x indices of preds
import warnings
from pathlib import Path
import yaml
from itertools import product
import sys
from general_tools.utils import get_root
LM = get_root("lm")
sys.path.append(str(LM / "slurm"))
import gen
from subprocess import Popen
baseline_configs = ["00_master.yaml"]
NAME = "02_REDO"
baseline_configs = [(LM / "configs") / b for b in baseline_configs]
variation_dict = {
"experiment_type": ["vgg_embeddings"],
"lm_model_path": ['lm/results/BASE/BERT_EXPERIMENT_TYPE.pt'],
"embedding_norm": ["softmax"],
"train_mode2": [
"single character", "multicharacter USE_CORRECT_CHAR_100",
"multicharacter MEAN_EMBEDDING_20 RANDOM_CHAR_20 USE_CORRECT_CHAR_20",
"multicharacter MEAN_EMBEDDING_80 USE_CORRECT_CHAR_20",
"multicharacter RANDOM_CHAR_80 USE_CORRECT_CHAR_20"
]
}
baseline_dict = {"max_intensity": 0}
baseline_dict = False
def cartesian_product(inp):