-
Notifications
You must be signed in to change notification settings - Fork 21
/
run.py
411 lines (336 loc) 路 13.2 KB
/
run.py
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
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
import gzip
import os
from collections import defaultdict
from typing import Any, Dict, List
import numpy as np
import pandas as pd
from numba import types
from numba.typed import Dict as TypedDict
from numba.typed import List as TypedList
from ..io import download, load_json, load_lz4, save_json, save_lz4
from .common import (
add_and_sort,
create_and_sort,
sort_dict_by_key,
sort_dict_of_dict_by_value,
to_typed_list,
)
from .generic import create_empty_results_dict
from .qrels import Qrels
class Run(object):
"""`Run` stores the relevance scores estimated by the model under evaluation.<\br>
The preferred way for creating a `Run` instance is converting a Python dictionary as follows:
```python
run_dict = {
"q_1": {
"d_1": 1.5,
"d_2": 2.6,
},
"q_2": {
"d_3": 2.8,
"d_2": 1.2,
"d_5": 3.1,
},
}
run = Run(run_dict, name="bm25")
run = Run() # Creates an empty Run with no name
```
"""
def __init__(self, run: Dict[str, Dict[str, float]] = None, name: str = None):
if run is None:
self.run = TypedDict.empty(
key_type=types.unicode_type,
value_type=types.DictType(types.unicode_type, types.float64),
)
self.sorted = False
else:
# Query IDs
q_ids = list(run.keys())
q_ids = TypedList(q_ids)
# Doc IDs
doc_ids = [list(doc.keys()) for doc in run.values()]
max_len = max(len(y) for x in doc_ids for y in x)
dtype = f"<U{max_len}"
doc_ids = TypedList([np.array(x, dtype=dtype) for x in doc_ids])
# Scores
scores = [list(doc.values()) for doc in run.values()]
scores = TypedList([np.array(x, dtype=float) for x in scores])
self.run = create_and_sort(q_ids, doc_ids, scores)
self.sorted = True
self.name = name
self.metadata = {}
self.scores = defaultdict(dict)
self.mean_scores = {}
self.std_scores = {}
def keys(self):
"""Returns query ids. Used internally."""
return self.run.keys()
def add_score(self, q_id: str, doc_id: str, score: int):
"""Add a (doc_id, score) pair to a query (or, change its value if it already exists).
Args:
q_id (str): Query ID
doc_id (str): Document ID
score (int): Relevance score
"""
if self.run.get(q_id) is None:
self.run[q_id] = TypedDict.empty(
key_type=types.unicode_type,
value_type=types.float64,
)
self.run[q_id][doc_id] = float(score)
self.sorted = False
def add(self, q_id: str, doc_ids: List[str], scores: List[float]):
"""Add a query and its relevant documents with the associated relevance score.
Args:
q_id (str): Query ID
doc_ids (List[str]): List of Document IDs
scores (List[int]): List of relevance scores
"""
self.add_multi([q_id], [doc_ids], [scores])
def add_multi(
self,
q_ids: List[str],
doc_ids: List[List[str]],
scores: List[List[float]],
):
"""Add multiple queries at once.
Args:
q_ids (List[str]): List of Query IDs
doc_ids (List[List[str]]): List of list of Document IDs
scores (List[List[int]]): List of list of relevance scores
"""
q_ids = TypedList(q_ids)
doc_ids = TypedList([TypedList(x) for x in doc_ids])
scores = TypedList([TypedList(map(float, x)) for x in scores])
self.run = add_and_sort(self.run, q_ids, doc_ids, scores)
self.sorted = True
def get_query_ids(self):
"""Returns query ids."""
return list(self.run.keys())
def get_doc_ids_and_scores(self):
"""Returns doc ids and relevance scores."""
return list(self.run.values())
# Sort in place
def sort(self):
"""Sort. Used internally."""
self.run = sort_dict_by_key(self.run)
self.run = sort_dict_of_dict_by_value(self.run)
self.sorted = True
def make_comparable(self, qrels: Qrels):
"""Adds empty results for queries missing from the run and removes those not appearing in qrels."""
# Adds empty results for missing queries
for q_id in qrels.qrels:
if q_id not in self.run:
self.run[q_id] = create_empty_results_dict()
# Remove results for additional queries
for q_id in self.run:
if q_id not in qrels.qrels:
del self.run[q_id]
self.sort()
return self
def to_typed_list(self):
"""Convert Run to Numba Typed List. Used internally."""
if not self.sorted:
self.sort()
return to_typed_list(self.run)
def to_dict(self):
"""Convert Run to Python dictionary.
Returns:
Dict[str, Dict[str, int]]: Run as Python dictionary
"""
d = defaultdict(dict)
for q_id in self.keys():
d[q_id] = dict(self[q_id])
return d
def to_dataframe(self) -> pd.DataFrame:
"""Convert Run to Pandas DataFrame with the following columns: `q_id`, `doc_id`, and `score`.
Returns:
pandas.DataFrame: Run as Pandas DataFrame.
"""
data = {"q_id": [], "doc_id": [], "score": []}
for q_id in self.run:
for doc_id in self.run[q_id]:
data["q_id"].append(q_id)
data["doc_id"].append(doc_id)
data["score"].append(self.run[q_id][doc_id])
return pd.DataFrame.from_dict(data)
def save(self, path: str = "run.json", kind: str = None):
"""Write `run` to `path` as JSON file, TREC run, LZ4 file, or Parquet file. File type is automatically inferred form the filename extension: ".json" -> "json", ".trec" -> "trec", ".txt" -> "trec", and ".lz4" -> "lz4", ".parq" -> "parquet", ".parquet" -> "parquet". Use the "kind" argument to override this behavior.
Args:
path (str, optional): Saving path. Defaults to "run.json".
kind (str, optional): Kind of file to save, must be either "json", "trec", or "ranxhub". If None, it will be automatically inferred from the filename extension.
"""
# Infer file extension -------------------------------------------------
kind = get_file_kind(path, kind)
# Save Run -------------------------------------------------------------
if not self.sorted:
self.sort()
if kind == "json":
save_json(self.to_dict(), path)
elif kind == "lz4":
save_lz4(self.to_dict(), path)
elif kind == "parquet":
self.to_dataframe().to_parquet(path, index=False)
else:
with open(path, "w") as f:
for i, q_id in enumerate(self.run.keys()):
for rank, doc_id in enumerate(self.run[q_id].keys()):
score = self.run[q_id][doc_id]
f.write(f"{q_id} Q0 {doc_id} {rank+1} {score} {self.name}")
if (
i != len(self.run.keys()) - 1
or rank != len(self.run[q_id].keys()) - 1
):
f.write("\n")
@staticmethod
def from_dict(d: Dict[str, Dict[str, float]], name: str = None):
"""Convert a Python dictionary in form of {q_id: {doc_id: score}} to ranx.Run.
Args:
d (Dict[str, Dict[str, int]]): Run as Python dictionary
name (str, optional): Run name. Defaults to None.
Returns:
Run: ranx.Run
"""
# Query IDs
q_ids = list(d.keys())
q_ids = TypedList(q_ids)
# Doc IDs
doc_ids = [list(doc.keys()) for doc in d.values()]
max_len = max(len(y) for x in doc_ids for y in x)
dtype = f"<U{max_len}"
doc_ids = TypedList([np.array(x, dtype=dtype) for x in doc_ids])
# Scores
scores = [list(doc.values()) for doc in d.values()]
scores = TypedList([np.array(x, dtype=float) for x in scores])
run = Run()
run.run = create_and_sort(q_ids, doc_ids, scores)
run.sorted = True
run.name = name
return run
@staticmethod
def from_file(path: str, kind: str = None, name: str = None):
"""Parse a run file into ranx.Run. Supported formats are JSON, TREC run, gzipped TREC run, and LZ4. Correct import behavior is inferred from the file extension: ".json" -> "json", ".trec" -> "trec", ".txt" -> "trec", ".gz" -> "gzipped trec", ".lz4" -> "lz4". Use the "kind" argument to override this behavior.
Args:
path (str): File path.
kind (str, optional): Kind of file to load, must be either "json" or "trec".
name (str, optional): Run name. Defaults to None.
Returns:
Run: ranx.Run
"""
# Infer file extension -------------------------------------------------
kind = get_file_kind(path, kind)
# Load Run -------------------------------------------------------------
if kind == "json":
run = load_json(path)
elif kind == "lz4":
run = load_lz4(path)
else:
run = defaultdict(dict)
with gzip.open(path, "rt") if kind == "gz" else open(path) as f:
for line in f:
q_id, _, doc_id, _, rel, run_name = line.split()
run[q_id][doc_id] = float(rel)
if name is None:
name = run_name
run = Run.from_dict(run, name)
return run
@staticmethod
def from_df(
df: pd.DataFrame,
q_id_col: str = "q_id",
doc_id_col: str = "doc_id",
score_col: str = "score",
name: str = None,
):
"""Convert a Pandas DataFrame to ranx.Run.
Args:
df (pd.DataFrame): Run as Pandas DataFrame
q_id_col (str, optional): Query IDs column. Defaults to "q_id".
doc_id_col (str, optional): Document IDs column. Defaults to "doc_id".
score_col (str, optional): Relevance scores column. Defaults to "score".
name (str, optional): Run name. Defaults to None.
Returns:
Run: ranx.Run
"""
assert (
df[q_id_col].dtype == "O"
), "DataFrame Query IDs column dtype must be `object` (string)"
assert (
df[doc_id_col].dtype == "O"
), "DataFrame Document IDs column dtype must be `object` (string)"
assert (
df[score_col].dtype == np.float64
), "DataFrame scores column dtype must be `float`"
run_py = (
df.groupby(q_id_col)[[doc_id_col, score_col]]
.apply(lambda g: {x[0]: x[1] for x in g.values.tolist()})
.to_dict()
)
return Run.from_dict(run_py, name)
@staticmethod
def from_parquet(
path: str,
q_id_col: str = "q_id",
doc_id_col: str = "doc_id",
score_col: str = "score",
pd_kwargs: Dict[str, Any] = None,
name: str = None,
):
"""Convert a Parquet file to ranx.Run.
Args:
path (str): File path.
q_id_col (str, optional): Query IDs column. Defaults to "q_id".
doc_id_col (str, optional): Document IDs column. Defaults to "doc_id".
score_col (str, optional): Relevance scores column. Defaults to "score".
pd_kwargs (Dict[str, Any], optional): Additional arguments to pass to `pandas.read_parquet` (see https://pandas.pydata.org/docs/reference/api/pandas.read_parquet.html). Defaults to None.
name (str, optional): Run name. Defaults to None.
Returns:
Run: ranx.Run
"""
pd_kwargs = {} if pd_kwargs is None else pd_kwargs
return Run.from_df(
df=pd.read_parquet(path, *pd_kwargs),
q_id_col=q_id_col,
doc_id_col=doc_id_col,
score_col=score_col,
name=name,
)
@staticmethod
def from_ranxhub(id: str):
"""Download and load a ranx.Run from ranxhub.
Args:
path (str): Run ID.
Returns:
Run: ranx.Run
"""
content = download(id)
run = Run.from_dict(content["run"])
run.name = content["metadata"]["run"]["name"]
run.metadata = content["metadata"]
return run
@property
def size(self):
return len(self.run)
def __getitem__(self, q_id):
return dict(self.run[q_id])
def __len__(self) -> int:
return len(self.run)
def __repr__(self):
return self.run.__repr__()
def __str__(self):
return self.run.__str__()
def get_file_kind(path: str = "run.json", kind: str = None) -> str:
# Infer file extension
if kind is None:
kind = os.path.splitext(path)[1][1:]
kind = "trec" if kind == "txt" else kind
kind = "parquet" if kind == "parq" else kind
# Sanity check
assert kind in {
"json",
"trec",
"lz4",
"gz",
"parquet",
}, "Error `kind` must be 'json', 'trec', 'lz4', 'gz', or 'parquet'"
return kind