Spaces:
Running
Running
Fix typos
Browse files- README.md +5 -17
- sample_text/en2es.m2m100_1.2B.json +123 -1
- sample_text/en2es.m2m100_418M.json +123 -1
README.md
CHANGED
@@ -13,20 +13,12 @@
|
|
13 |
<br>
|
14 |
</p>
|
15 |
|
16 |
-
Easy-translate is a script for translating large text files in your machine using
|
17 |
-
the [M2M100 models](https://arxiv.org/pdf/2010.11125.pdf) from Facebook/Meta AI.
|
18 |
-
We also privide a [script](#evaluate-translations) for Easy-Evaluation of your translations 🥳
|
19 |
|
20 |
|
21 |
-
M2M100 is a multilingual encoder-decoder (seq-to-seq) model trained for Many-to-Many multilingual translation.
|
22 |
-
It was introduced in this [paper](https://arxiv.org/abs/2010.11125) and first released in [this](https://github.com/pytorch/fairseq/tree/master/examples/m2m_100) repository.
|
23 |
-
The model that can directly translate between the 9,900 directions of 100 languages.
|
24 |
-
|
25 |
-
Easy-Translate is built on top of 🤗HuggingFace's
|
26 |
-
[Transformers](https://huggingface.co/docs/transformers/index) and
|
27 |
-
🤗HuggingFace's [Accelerate](https://huggingface.co/docs/accelerate/index) library.
|
28 |
-
We support:
|
29 |
|
|
|
30 |
* CPU / multi-CPU / GPU / multi-GPU / TPU acceleration
|
31 |
* BF16 / FP16 / FP32 precision.
|
32 |
* Automatic batch size finder: Forget CUDA OOM errors. Set an initial batch size, if it doesn't fit, we will automatically adjust it.
|
@@ -80,9 +72,7 @@ accelerate launch translate.py \
|
|
80 |
|
81 |
#### Multi-GPU:
|
82 |
See Accelerate documentation for more information (multi-node, TPU, Sharded model...): https://huggingface.co/docs/accelerate/index
|
83 |
-
You can use the Accelerate CLI to configure the Accelerate environment (Run
|
84 |
-
`accelerate config` in your terminal) instead of using the
|
85 |
-
`--multi_gpu and --num_processes` flags.
|
86 |
|
87 |
```bash
|
88 |
accelerate launch --multi_gpu --num_processes 2 --num_machines 1 translate.py \
|
@@ -94,9 +84,7 @@ accelerate launch --multi_gpu --num_processes 2 --num_machines 1 translate.py \
|
|
94 |
```
|
95 |
|
96 |
#### Automatic batch size finder:
|
97 |
-
We will automatically find a batch size that fits in your GPU memory.
|
98 |
-
The default initial batch size is 128 (You can set it with the `--starting_batch_size 128` flag).
|
99 |
-
If we find an Out Of Memory error, we will automatically decrease the batch size until we find a working one.
|
100 |
|
101 |
|
102 |
|
|
|
13 |
<br>
|
14 |
</p>
|
15 |
|
16 |
+
Easy-translate is a script for translating large text files in your machine using the [M2M100 models](https://arxiv.org/pdf/2010.11125.pdf) from Facebook/Meta AI. We also privide a [script](#evaluate-translations) for Easy-Evaluation of your translations 🥳
|
|
|
|
|
17 |
|
18 |
|
19 |
+
M2M100 is a multilingual encoder-decoder (seq-to-seq) model trained for Many-to-Many multilingual translation. It was introduced in this [paper](https://arxiv.org/abs/2010.11125) and first released in [this](https://github.com/pytorch/fairseq/tree/master/examples/m2m_100) repository. The model that can directly translate between the 9,900 directions of 100 languages.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
|
21 |
+
Easy-Translate is built on top of 🤗HuggingFace's [Transformers](https://huggingface.co/docs/transformers/index) and 🤗HuggingFace's [Accelerate](https://huggingface.co/docs/accelerate/index) library. We support:
|
22 |
* CPU / multi-CPU / GPU / multi-GPU / TPU acceleration
|
23 |
* BF16 / FP16 / FP32 precision.
|
24 |
* Automatic batch size finder: Forget CUDA OOM errors. Set an initial batch size, if it doesn't fit, we will automatically adjust it.
|
|
|
72 |
|
73 |
#### Multi-GPU:
|
74 |
See Accelerate documentation for more information (multi-node, TPU, Sharded model...): https://huggingface.co/docs/accelerate/index
|
75 |
+
You can use the Accelerate CLI to configure the Accelerate environment (Run `accelerate config` in your terminal) instead of using the `--multi_gpu and --num_processes` flags.
|
|
|
|
|
76 |
|
77 |
```bash
|
78 |
accelerate launch --multi_gpu --num_processes 2 --num_machines 1 translate.py \
|
|
|
84 |
```
|
85 |
|
86 |
#### Automatic batch size finder:
|
87 |
+
We will automatically find a batch size that fits in your GPU memory. The default initial batch size is 128 (You can set it with the `--starting_batch_size 128` flag). If we find an Out Of Memory error, we will automatically decrease the batch size until we find a working one.
|
|
|
|
|
88 |
|
89 |
|
90 |
|
sample_text/en2es.m2m100_1.2B.json
CHANGED
@@ -1 +1,123 @@
|
|
1 |
-
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"sacrebleu": {
|
3 |
+
"score": 32.101150640281695,
|
4 |
+
"counts": [
|
5 |
+
19160,
|
6 |
+
11392,
|
7 |
+
7558,
|
8 |
+
5186
|
9 |
+
],
|
10 |
+
"totals": [
|
11 |
+
31477,
|
12 |
+
30479,
|
13 |
+
29481,
|
14 |
+
28485
|
15 |
+
],
|
16 |
+
"precisions": [
|
17 |
+
60.86984147155066,
|
18 |
+
37.37655434889596,
|
19 |
+
25.636850853091822,
|
20 |
+
18.20607337195015
|
21 |
+
],
|
22 |
+
"bp": 1.0,
|
23 |
+
"sys_len": 31477,
|
24 |
+
"ref_len": 30102
|
25 |
+
},
|
26 |
+
"rouge": {
|
27 |
+
"rouge1": [
|
28 |
+
[
|
29 |
+
0.5852396804366098,
|
30 |
+
0.6089057437338691,
|
31 |
+
0.5919486437026797
|
32 |
+
],
|
33 |
+
[
|
34 |
+
0.5964621218261164,
|
35 |
+
0.6200342221830797,
|
36 |
+
0.6029705008756368
|
37 |
+
],
|
38 |
+
[
|
39 |
+
0.6068321807422377,
|
40 |
+
0.6311106822798185,
|
41 |
+
0.61324805661008
|
42 |
+
]
|
43 |
+
],
|
44 |
+
"rouge2": [
|
45 |
+
[
|
46 |
+
0.3710985389559613,
|
47 |
+
0.38708055355385995,
|
48 |
+
0.3761201217327784
|
49 |
+
],
|
50 |
+
[
|
51 |
+
0.3844850790869714,
|
52 |
+
0.40017782122170353,
|
53 |
+
0.38920434271970195
|
54 |
+
],
|
55 |
+
[
|
56 |
+
0.3968990790506025,
|
57 |
+
0.41382310483690327,
|
58 |
+
0.4022299418726329
|
59 |
+
]
|
60 |
+
],
|
61 |
+
"rougeL": [
|
62 |
+
[
|
63 |
+
0.5351505034410595,
|
64 |
+
0.5564838960633809,
|
65 |
+
0.5410602618870524
|
66 |
+
],
|
67 |
+
[
|
68 |
+
0.5457898501195475,
|
69 |
+
0.5677049056091881,
|
70 |
+
0.5519189480892548
|
71 |
+
],
|
72 |
+
[
|
73 |
+
0.5575497491149766,
|
74 |
+
0.5787856637940312,
|
75 |
+
0.5630101422167583
|
76 |
+
]
|
77 |
+
],
|
78 |
+
"rougeLsum": [
|
79 |
+
[
|
80 |
+
0.5352116089085267,
|
81 |
+
0.5570236521823667,
|
82 |
+
0.5415939934790461
|
83 |
+
],
|
84 |
+
[
|
85 |
+
0.5463246235983789,
|
86 |
+
0.5676427704754348,
|
87 |
+
0.5522237812823654
|
88 |
+
],
|
89 |
+
[
|
90 |
+
0.5581141358005033,
|
91 |
+
0.5796683147249665,
|
92 |
+
0.5630221371759908
|
93 |
+
]
|
94 |
+
]
|
95 |
+
},
|
96 |
+
"bleu": {
|
97 |
+
"bleu": 0.2842153038526809,
|
98 |
+
"precisions": [
|
99 |
+
0.5535070989616444,
|
100 |
+
0.33646946844340314,
|
101 |
+
0.22383069265549602,
|
102 |
+
0.15653135365661033
|
103 |
+
],
|
104 |
+
"brevity_penalty": 1.0,
|
105 |
+
"length_ratio": 1.0469217970049918,
|
106 |
+
"translation_length": 28314,
|
107 |
+
"reference_length": 27045
|
108 |
+
},
|
109 |
+
"meteor": {
|
110 |
+
"meteor": 0.4880039569987408
|
111 |
+
},
|
112 |
+
"ter": {
|
113 |
+
"score": 59.500831946755405,
|
114 |
+
"num_edits": 16092,
|
115 |
+
"ref_length": 27045.0
|
116 |
+
},
|
117 |
+
"bert_score": {
|
118 |
+
"precision": 0.8192488248944283,
|
119 |
+
"recall": 0.8262857750356197,
|
120 |
+
"f1": 0.8223461411595344,
|
121 |
+
"hashcode": "microsoft/deberta-xlarge-mnli_L40_no-idf_version=0.3.11(hug_trans=4.18.0)_fast-tokenizer"
|
122 |
+
}
|
123 |
+
}
|
sample_text/en2es.m2m100_418M.json
CHANGED
@@ -1 +1,123 @@
|
|
1 |
-
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"sacrebleu": {
|
3 |
+
"score": 29.035496917461597,
|
4 |
+
"counts": [
|
5 |
+
18582,
|
6 |
+
10514,
|
7 |
+
6681,
|
8 |
+
4387
|
9 |
+
],
|
10 |
+
"totals": [
|
11 |
+
31477,
|
12 |
+
30479,
|
13 |
+
29481,
|
14 |
+
28485
|
15 |
+
],
|
16 |
+
"precisions": [
|
17 |
+
59.033580074339994,
|
18 |
+
34.49588241084025,
|
19 |
+
22.662053525999795,
|
20 |
+
15.401088292083553
|
21 |
+
],
|
22 |
+
"bp": 1.0,
|
23 |
+
"sys_len": 31477,
|
24 |
+
"ref_len": 30388
|
25 |
+
},
|
26 |
+
"rouge": {
|
27 |
+
"rouge1": [
|
28 |
+
[
|
29 |
+
0.5661701202298134,
|
30 |
+
0.5806961045770566,
|
31 |
+
0.5693885562082325
|
32 |
+
],
|
33 |
+
[
|
34 |
+
0.5768745925790656,
|
35 |
+
0.5926959547911554,
|
36 |
+
0.5803693779677083
|
37 |
+
],
|
38 |
+
[
|
39 |
+
0.5871085218904836,
|
40 |
+
0.6035331460243276,
|
41 |
+
0.5900979805085623
|
42 |
+
]
|
43 |
+
],
|
44 |
+
"rouge2": [
|
45 |
+
[
|
46 |
+
0.34243414046469267,
|
47 |
+
0.35226400857606666,
|
48 |
+
0.34469210847048837
|
49 |
+
],
|
50 |
+
[
|
51 |
+
0.3545484183384055,
|
52 |
+
0.36470783370743065,
|
53 |
+
0.3569058648048812
|
54 |
+
],
|
55 |
+
[
|
56 |
+
0.36612813327517263,
|
57 |
+
0.37717476449671,
|
58 |
+
0.3689653665404565
|
59 |
+
]
|
60 |
+
],
|
61 |
+
"rougeL": [
|
62 |
+
[
|
63 |
+
0.5129704896656746,
|
64 |
+
0.526995889564155,
|
65 |
+
0.5162056185006965
|
66 |
+
],
|
67 |
+
[
|
68 |
+
0.523632841460358,
|
69 |
+
0.5375452284094455,
|
70 |
+
0.5267080806612512
|
71 |
+
],
|
72 |
+
[
|
73 |
+
0.5350158816319085,
|
74 |
+
0.5480980981777757,
|
75 |
+
0.5372302857012781
|
76 |
+
]
|
77 |
+
],
|
78 |
+
"rougeLsum": [
|
79 |
+
[
|
80 |
+
0.5126805856827783,
|
81 |
+
0.5265189554049317,
|
82 |
+
0.5155154093959223
|
83 |
+
],
|
84 |
+
[
|
85 |
+
0.5239559133309495,
|
86 |
+
0.5380410013947112,
|
87 |
+
0.5271022617246641
|
88 |
+
],
|
89 |
+
[
|
90 |
+
0.5351934954578494,
|
91 |
+
0.5491115103854219,
|
92 |
+
0.5381174565735956
|
93 |
+
]
|
94 |
+
]
|
95 |
+
},
|
96 |
+
"bleu": {
|
97 |
+
"bleu": 0.2546886610724999,
|
98 |
+
"precisions": [
|
99 |
+
0.5339761248852158,
|
100 |
+
0.30784155806120955,
|
101 |
+
0.19560013678331242,
|
102 |
+
0.1308640025272469
|
103 |
+
],
|
104 |
+
"brevity_penalty": 1.0,
|
105 |
+
"length_ratio": 1.0353982300884956,
|
106 |
+
"translation_length": 28314,
|
107 |
+
"reference_length": 27346
|
108 |
+
},
|
109 |
+
"meteor": {
|
110 |
+
"meteor": 0.4630996837124251
|
111 |
+
},
|
112 |
+
"ter": {
|
113 |
+
"score": 61.848167922182405,
|
114 |
+
"num_edits": 16913,
|
115 |
+
"ref_length": 27346.0
|
116 |
+
},
|
117 |
+
"bert_score": {
|
118 |
+
"precision": 0.8128397642374039,
|
119 |
+
"recall": 0.8185485603511333,
|
120 |
+
"f1": 0.8153312988877296,
|
121 |
+
"hashcode": "microsoft/deberta-xlarge-mnli_L40_no-idf_version=0.3.11(hug_trans=4.18.0)_fast-tokenizer"
|
122 |
+
}
|
123 |
+
}
|