Spaces:
Running
Running
Merge pull request #3 from ikergarcia1996/multigpu-bug
Browse files- dataset.py +8 -7
- translate.py +4 -0
dataset.py
CHANGED
@@ -14,6 +14,8 @@ class DatasetReader(IterableDataset):
|
|
14 |
self.current_line = 0
|
15 |
self.total_lines = count_lines(filename)
|
16 |
print(f"{self.total_lines} lines in {filename}")
|
|
|
|
|
17 |
|
18 |
def preprocess(self, text: str):
|
19 |
self.current_line += 1
|
@@ -29,9 +31,7 @@ class DatasetReader(IterableDataset):
|
|
29 |
)
|
30 |
|
31 |
def __iter__(self):
|
32 |
-
|
33 |
-
mapped_itr = map(self.preprocess, file_itr)
|
34 |
-
return mapped_itr
|
35 |
|
36 |
def __len__(self):
|
37 |
return self.total_lines
|
@@ -50,6 +50,10 @@ class ParallelTextReader(IterableDataset):
|
|
50 |
self.num_sentences = gold_path_lines
|
51 |
self.current_line = 0
|
52 |
|
|
|
|
|
|
|
|
|
53 |
def preprocess(self, pred: str, gold: str):
|
54 |
self.current_line += 1
|
55 |
pred = pred.rstrip().strip()
|
@@ -61,10 +65,7 @@ class ParallelTextReader(IterableDataset):
|
|
61 |
return pred, [gold]
|
62 |
|
63 |
def __iter__(self):
|
64 |
-
|
65 |
-
gold_itr = open(self.gold_path, "r", encoding="utf8")
|
66 |
-
mapped_itr = map(self.preprocess, pred_itr, gold_itr)
|
67 |
-
return mapped_itr
|
68 |
|
69 |
def __len__(self):
|
70 |
return self.num_sentences
|
|
|
14 |
self.current_line = 0
|
15 |
self.total_lines = count_lines(filename)
|
16 |
print(f"{self.total_lines} lines in {filename}")
|
17 |
+
file_itr = open(self.filename, "r", encoding="utf8")
|
18 |
+
self.mapped_itr = map(self.preprocess, file_itr)
|
19 |
|
20 |
def preprocess(self, text: str):
|
21 |
self.current_line += 1
|
|
|
31 |
)
|
32 |
|
33 |
def __iter__(self):
|
34 |
+
return self.mapped_itr
|
|
|
|
|
35 |
|
36 |
def __len__(self):
|
37 |
return self.total_lines
|
|
|
50 |
self.num_sentences = gold_path_lines
|
51 |
self.current_line = 0
|
52 |
|
53 |
+
pred_itr = open(self.pred_path, "r", encoding="utf8")
|
54 |
+
gold_itr = open(self.gold_path, "r", encoding="utf8")
|
55 |
+
self.mapped_itr = map(self.preprocess, pred_itr, gold_itr)
|
56 |
+
|
57 |
def preprocess(self, pred: str, gold: str):
|
58 |
self.current_line += 1
|
59 |
pred = pred.rstrip().strip()
|
|
|
65 |
return pred, [gold]
|
66 |
|
67 |
def __iter__(self):
|
68 |
+
return self.mapped_itr
|
|
|
|
|
|
|
69 |
|
70 |
def __len__(self):
|
71 |
return self.num_sentences
|
translate.py
CHANGED
@@ -19,6 +19,10 @@ from dataset import DatasetReader, count_lines
|
|
19 |
|
20 |
from accelerate import Accelerator, DistributedType, find_executable_batch_size
|
21 |
|
|
|
|
|
|
|
|
|
22 |
|
23 |
def get_dataloader(
|
24 |
accelerator: Accelerator,
|
|
|
19 |
|
20 |
from accelerate import Accelerator, DistributedType, find_executable_batch_size
|
21 |
|
22 |
+
torch.multiprocessing.set_sharing_strategy(
|
23 |
+
"file_system"
|
24 |
+
) # FIXES RuntimeError: Too many open files.
|
25 |
+
|
26 |
|
27 |
def get_dataloader(
|
28 |
accelerator: Accelerator,
|