Spaces:
Runtime error
Runtime error
File size: 3,763 Bytes
8894e1f 7dd65d1 3d72bf8 8894e1f 4dd0f69 8894e1f 7dd65d1 3d72bf8 8894e1f 4a5b13d 8894e1f 4a5b13d 8894e1f 4a5b13d 3d72bf8 4a5b13d 3d72bf8 4a5b13d 3d72bf8 4a5b13d 1216e7c 3d72bf8 8894e1f 3d72bf8 8894e1f 4a5b13d 3d72bf8 4a5b13d 3d72bf8 4a5b13d 3d72bf8 4a5b13d 8894e1f 3d72bf8 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 |
import os
import numpy as np
import pandas as pd
# from inspiredco.critique import Critique
from zeno import (
ZenoOptions,
distill,
metric,
model,
ModelReturn,
DistillReturn,
MetricReturn,
)
# client = Critique(api_key=os.environ["INSPIREDCO_API_KEY"])
@model
def pred_fns(name):
def pred(df, ops):
model_df = pd.read_csv(
ops.label_path + "/{}.tsv".format(name),
sep="\t",
quoting=3,
keep_default_na=False,
)
embed_df = pd.read_csv(
ops.label_path + "/ref_embed.tsv",
sep="\t",
keep_default_na=False,
quoting=3,
)
df_join = df[["text"]].merge(
model_df[["text", "translation"]], on="text", how="left"
)
df_join = df_join.merge(embed_df, on="text", how="left")
return ModelReturn(
model_output=df_join["translation"].fillna(""),
embedding=[np.fromstring(d[1:-1], sep=",") for d in df_join["embed"]],
)
return pred
@distill
def bert_score(df, ops):
eval_dict = df[["source", ops.output_column, "reference"]].to_dict("records")
for d in eval_dict:
d["references"] = [d.pop("reference")]
d["target"] = d.pop(ops.output_column)
result = client.evaluate(
metric="bert_score", config={"model": "bert-base-uncased"}, dataset=eval_dict
)
return DistillReturn(
distill_output=[round(r["value"], 6) for r in result["examples"]]
)
@distill
def bleu(df, ops):
eval_dict = df[[ops.output_column, "reference"]].to_dict("records")
for d in eval_dict:
d["references"] = [d.pop("reference")]
d["target"] = d.pop(ops.output_column)
result = client.evaluate(
metric="bleu",
config={"smooth_method": "add_k", "smooth-value": 1.0},
dataset=eval_dict,
)
return DistillReturn(
distill_output=[round(r["value"], 6) for r in result["examples"]]
)
@distill
def chrf(df, ops):
eval_dict = df[[ops.output_column, "reference"]].to_dict("records")
for d in eval_dict:
d["references"] = [d.pop("reference")]
d["target"] = d.pop(ops.output_column)
result = client.evaluate(
metric="chrf",
config={},
dataset=eval_dict,
)
return DistillReturn(
distill_output=[round(r["value"], 6) for r in result["examples"]]
)
@distill
def length_ratio(df, ops):
eval_dict = df[[ops.output_column, "reference"]].to_dict("records")
for d in eval_dict:
d["references"] = [d.pop("reference")]
d["target"] = d.pop(ops.output_column)
result = client.evaluate(
metric="length_ratio",
config={},
dataset=eval_dict,
)
return DistillReturn(
distill_output=[round(r["value"], 6) for r in result["examples"]]
)
@metric
def avg_bert_score(df, ops: ZenoOptions):
if len(df) == 0:
return MetricReturn(metric=0)
return MetricReturn(metric=df[ops.distill_columns["bert_score"]].fillna(0).mean())
@metric
def avg_bleu(df, ops: ZenoOptions):
if len(df) == 0:
return MetricReturn(metric=0)
return MetricReturn(metric=df[ops.distill_columns["bleu"]].fillna(0).mean())
@metric
def avg_chrf(df, ops: ZenoOptions):
if len(df) == 0:
return MetricReturn(metric=0)
return MetricReturn(metric=df[ops.distill_columns["chrf"]].fillna(0).mean())
@metric
def avg_length_ratio(df, ops: ZenoOptions):
if len(df) == 0:
return MetricReturn(metric=0)
return MetricReturn(metric=df[ops.distill_columns["length_ratio"]].fillna(0).mean())
@distill
def length(df, ops):
return DistillReturn(distill_output=df[ops.data_column].str.len())
|