Andrey Kutuzov
commited on
Commit
•
fb38b14
1
Parent(s):
0067049
Camera ready
Browse files- README.md +29 -29
- all_results.json +14 -14
- config.json +2 -1
- eval_results.json +10 -10
- generation_config.json +1 -1
- pytorch_model-00001-of-00002.bin → pytorch_model-00001-of-00003.bin +2 -2
- pytorch_model-00002-of-00003.bin +3 -0
- pytorch_model-00002-of-00002.bin → pytorch_model-00003-of-00003.bin +1 -1
- pytorch_model.bin.index.json +0 -0
- special_tokens_map.json +21 -3
- tokenizer.json +2 -2
- tokenizer_config.json +27 -1
- train_results.json +5 -5
- trainer_state.json +230 -92
- training_args.bin +2 -2
- upload.py +0 -11
README.md
CHANGED
@@ -12,34 +12,46 @@ language:
|
|
12 |
widget:
|
13 |
- text: "Мы сели в тачку и поехали по ресторанам. Что такое тачка?"
|
14 |
example_title: "Definition generation"
|
|
|
15 |
---
|
16 |
|
17 |
-
#
|
18 |
|
19 |
-
This model is a version of [
|
|
|
20 |
|
21 |
-
It
|
22 |
-
|
23 |
-
- Rouge1: 0.2536
|
24 |
-
- Rouge2: 0.003
|
25 |
-
- Rougel: 0.2531
|
26 |
-
- Rougelsum: 0.2527
|
27 |
-
- Gen Len: 24.0693
|
28 |
|
29 |
## Model description
|
30 |
|
31 |
-
|
|
|
32 |
|
33 |
## Intended uses & limitations
|
34 |
|
35 |
-
|
|
|
36 |
|
37 |
## Training and evaluation data
|
38 |
|
39 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
|
41 |
## Training procedure
|
42 |
|
|
|
|
|
43 |
### Training hyperparameters
|
44 |
|
45 |
The following hyperparameters were used during training:
|
@@ -56,23 +68,11 @@ The following hyperparameters were used during training:
|
|
56 |
- lr_scheduler_type: linear
|
57 |
- num_epochs: 20.0
|
58 |
|
59 |
-
### Training results
|
60 |
-
|
61 |
-
| Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len |
|
62 |
-
|:-------------:|:-----:|:----:|:---------------:|:------:|:------:|:------:|:---------:|:-------:|
|
63 |
-
| 2.0449 | 1.0 | 512 | 1.6755 | 0.0817 | 0.0 | 0.0778 | 0.0817 | 16.7581 |
|
64 |
-
| 1.707 | 2.0 | 1025 | 1.6182 | 0.096 | 0.0 | 0.097 | 0.1 | 15.8621 |
|
65 |
-
| 1.5398 | 3.0 | 1537 | 1.6085 | 0.1394 | 0.0034 | 0.1401 | 0.1416 | 16.4765 |
|
66 |
-
| 1.4142 | 4.0 | 2050 | 1.6016 | 0.1132 | 0.0 | 0.1132 | 0.1098 | 16.2732 |
|
67 |
-
| 1.3102 | 5.0 | 2562 | 1.6241 | 0.2082 | 0.0034 | 0.2054 | 0.2061 | 16.2877 |
|
68 |
-
| 1.2162 | 6.0 | 3075 | 1.6281 | 0.1549 | 0.0 | 0.1549 | 0.1549 | 16.1581 |
|
69 |
-
| 1.1364 | 7.0 | 3587 | 1.6622 | 0.1583 | 0.0 | 0.1575 | 0.1589 | 15.9925 |
|
70 |
-
| 1.0649 | 8.0 | 4100 | 1.6812 | 0.2033 | 0.0137 | 0.2012 | 0.2027 | 16.5099 |
|
71 |
-
|
72 |
-
|
73 |
### Framework versions
|
74 |
|
75 |
-
- Transformers 4.
|
76 |
- Pytorch 1.13.1+rocm5.2
|
77 |
-
- Datasets 2.
|
78 |
-
- Tokenizers 0.
|
|
|
|
|
|
12 |
widget:
|
13 |
- text: "Мы сели в тачку и поехали по ресторанам. Что такое тачка?"
|
14 |
example_title: "Definition generation"
|
15 |
+
license: cc-by-sa-4.0
|
16 |
---
|
17 |
|
18 |
+
# mT0-Definition-Ru XL
|
19 |
|
20 |
+
This model is a version of [mT0 XL](https://huggingface.co/bigscience/mt0-xl) finetuned on the Russian part of [CodWoE](https://aclanthology.org/2022.semeval-1.1/),
|
21 |
+
a dataset of definitions and usage examples.
|
22 |
|
23 |
+
It generates definitions of Russian words in context.
|
24 |
+
Its input is the usage example and the instruction question "Что такое TARGET_WORD?"
|
|
|
|
|
|
|
|
|
|
|
25 |
|
26 |
## Model description
|
27 |
|
28 |
+
See details in the paper `Enriching Word Usage Graphs with Cluster Definitions` (LREC-COLING'2024) by
|
29 |
+
Mariia Fedorova, Andrey Kutuzov, Nikolay Arefyev and Dominik Schlechtweg.
|
30 |
|
31 |
## Intended uses & limitations
|
32 |
|
33 |
+
The model is intended for research purposes, as a source of contextualized dictionary-like lexical definitions.
|
34 |
+
Generated definitions can contain all sorts of biases and stereotypes, stemming from the underlying language model.
|
35 |
|
36 |
## Training and evaluation data
|
37 |
|
38 |
+
Russian subset of *CodWoE* ([Mickus et al., SemEval 2022](https://aclanthology.org/2022.semeval-1.1/)).
|
39 |
+
|
40 |
+
## Training results
|
41 |
+
|
42 |
+
mT0-Definition-Ru XL achieves the following results on the CodWoE evaluation set:
|
43 |
+
|
44 |
+
- Loss: 1.7996
|
45 |
+
- Rouge1: 17.5576
|
46 |
+
- Rouge2: 8.7614
|
47 |
+
- Rougel: 17.2533
|
48 |
+
- Rougelsum: 17.3204
|
49 |
+
- Gen Len: 21.6774
|
50 |
|
51 |
## Training procedure
|
52 |
|
53 |
+
mT0-Definition-Ru XL was fine-tuned in a sequence-to-sequence mode on examples of contextualized dictionary definitions.
|
54 |
+
|
55 |
### Training hyperparameters
|
56 |
|
57 |
The following hyperparameters were used during training:
|
|
|
68 |
- lr_scheduler_type: linear
|
69 |
- num_epochs: 20.0
|
70 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
71 |
### Framework versions
|
72 |
|
73 |
+
- Transformers 4.37.1
|
74 |
- Pytorch 1.13.1+rocm5.2
|
75 |
+
- Datasets 2.16.1
|
76 |
+
- Tokenizers 0.15.1
|
77 |
+
|
78 |
+
## Citation
|
all_results.json
CHANGED
@@ -1,18 +1,18 @@
|
|
1 |
{
|
2 |
-
"epoch":
|
3 |
-
"eval_gen_len":
|
4 |
-
"eval_loss": 1.
|
5 |
-
"eval_rouge1":
|
6 |
-
"eval_rouge2":
|
7 |
-
"eval_rougeL":
|
8 |
-
"eval_rougeLsum":
|
9 |
-
"eval_runtime":
|
10 |
"eval_samples": 7288,
|
11 |
-
"eval_samples_per_second":
|
12 |
-
"eval_steps_per_second": 0.
|
13 |
-
"train_loss": 1.
|
14 |
-
"train_runtime":
|
15 |
"train_samples": 65584,
|
16 |
-
"train_samples_per_second":
|
17 |
-
"train_steps_per_second":
|
18 |
}
|
|
|
1 |
{
|
2 |
+
"epoch": 15.0,
|
3 |
+
"eval_gen_len": 21.67735745614035,
|
4 |
+
"eval_loss": 1.7995822429656982,
|
5 |
+
"eval_rouge1": 17.5576,
|
6 |
+
"eval_rouge2": 8.7614,
|
7 |
+
"eval_rougeL": 17.2533,
|
8 |
+
"eval_rougeLsum": 17.3204,
|
9 |
+
"eval_runtime": 423.3617,
|
10 |
"eval_samples": 7288,
|
11 |
+
"eval_samples_per_second": 17.215,
|
12 |
+
"eval_steps_per_second": 0.539,
|
13 |
+
"train_loss": 1.153788715837463,
|
14 |
+
"train_runtime": 20083.7121,
|
15 |
"train_samples": 65584,
|
16 |
+
"train_samples_per_second": 65.311,
|
17 |
+
"train_steps_per_second": 0.51
|
18 |
}
|
config.json
CHANGED
@@ -3,6 +3,7 @@
|
|
3 |
"architectures": [
|
4 |
"MT5ForConditionalGeneration"
|
5 |
],
|
|
|
6 |
"d_ff": 5120,
|
7 |
"d_kv": 64,
|
8 |
"d_model": 2048,
|
@@ -26,7 +27,7 @@
|
|
26 |
"tie_word_embeddings": false,
|
27 |
"tokenizer_class": "T5Tokenizer",
|
28 |
"torch_dtype": "float32",
|
29 |
-
"transformers_version": "4.
|
30 |
"use_cache": true,
|
31 |
"vocab_size": 250112
|
32 |
}
|
|
|
3 |
"architectures": [
|
4 |
"MT5ForConditionalGeneration"
|
5 |
],
|
6 |
+
"classifier_dropout": 0.0,
|
7 |
"d_ff": 5120,
|
8 |
"d_kv": 64,
|
9 |
"d_model": 2048,
|
|
|
27 |
"tie_word_embeddings": false,
|
28 |
"tokenizer_class": "T5Tokenizer",
|
29 |
"torch_dtype": "float32",
|
30 |
+
"transformers_version": "4.37.1",
|
31 |
"use_cache": true,
|
32 |
"vocab_size": 250112
|
33 |
}
|
eval_results.json
CHANGED
@@ -1,13 +1,13 @@
|
|
1 |
{
|
2 |
-
"epoch":
|
3 |
-
"eval_gen_len":
|
4 |
-
"eval_loss": 1.
|
5 |
-
"eval_rouge1":
|
6 |
-
"eval_rouge2":
|
7 |
-
"eval_rougeL":
|
8 |
-
"eval_rougeLsum":
|
9 |
-
"eval_runtime":
|
10 |
"eval_samples": 7288,
|
11 |
-
"eval_samples_per_second":
|
12 |
-
"eval_steps_per_second": 0.
|
13 |
}
|
|
|
1 |
{
|
2 |
+
"epoch": 15.0,
|
3 |
+
"eval_gen_len": 21.67735745614035,
|
4 |
+
"eval_loss": 1.7995822429656982,
|
5 |
+
"eval_rouge1": 17.5576,
|
6 |
+
"eval_rouge2": 8.7614,
|
7 |
+
"eval_rougeL": 17.2533,
|
8 |
+
"eval_rougeLsum": 17.3204,
|
9 |
+
"eval_runtime": 423.3617,
|
10 |
"eval_samples": 7288,
|
11 |
+
"eval_samples_per_second": 17.215,
|
12 |
+
"eval_steps_per_second": 0.539
|
13 |
}
|
generation_config.json
CHANGED
@@ -2,5 +2,5 @@
|
|
2 |
"decoder_start_token_id": 0,
|
3 |
"eos_token_id": 1,
|
4 |
"pad_token_id": 0,
|
5 |
-
"transformers_version": "4.
|
6 |
}
|
|
|
2 |
"decoder_start_token_id": 0,
|
3 |
"eos_token_id": 1,
|
4 |
"pad_token_id": 0,
|
5 |
+
"transformers_version": "4.37.1"
|
6 |
}
|
pytorch_model-00001-of-00002.bin → pytorch_model-00001-of-00003.bin
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:883fe03c74638936701aa6db8ea888ebcdfcecaf9a5e896ed5d0db65dbefa436
|
3 |
+
size 4993619647
|
pytorch_model-00002-of-00003.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:06c1a331261c210d7adb2977052855975e876fcf786ac4df398d3a638c6af82c
|
3 |
+
size 4983398004
|
pytorch_model-00002-of-00002.bin → pytorch_model-00003-of-00003.bin
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4993663292
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:653556b3b21b03d7198cb94d6a4405d2e3cb6b329a06cc4333c710e9f43a1e45
|
3 |
size 4993663292
|
pytorch_model.bin.index.json
CHANGED
The diff for this file is too large to render.
See raw diff
|
|
special_tokens_map.json
CHANGED
@@ -1,5 +1,23 @@
|
|
1 |
{
|
2 |
-
"eos_token":
|
3 |
-
|
4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
}
|
|
|
1 |
{
|
2 |
+
"eos_token": {
|
3 |
+
"content": "</s>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"pad_token": {
|
10 |
+
"content": "<pad>",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"unk_token": {
|
17 |
+
"content": "<unk>",
|
18 |
+
"lstrip": false,
|
19 |
+
"normalized": false,
|
20 |
+
"rstrip": false,
|
21 |
+
"single_word": false
|
22 |
+
}
|
23 |
}
|
tokenizer.json
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:54e5c72a5ea09da48b2f316760b8bc5a445683ab9a5bc6b68db5d8db624ecceb
|
3 |
+
size 16315213
|
tokenizer_config.json
CHANGED
@@ -1,5 +1,31 @@
|
|
1 |
{
|
2 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
"clean_up_tokenization_spaces": true,
|
4 |
"eos_token": "</s>",
|
5 |
"extra_ids": 0,
|
|
|
1 |
{
|
2 |
+
"added_tokens_decoder": {
|
3 |
+
"0": {
|
4 |
+
"content": "<pad>",
|
5 |
+
"lstrip": false,
|
6 |
+
"normalized": false,
|
7 |
+
"rstrip": false,
|
8 |
+
"single_word": false,
|
9 |
+
"special": true
|
10 |
+
},
|
11 |
+
"1": {
|
12 |
+
"content": "</s>",
|
13 |
+
"lstrip": false,
|
14 |
+
"normalized": false,
|
15 |
+
"rstrip": false,
|
16 |
+
"single_word": false,
|
17 |
+
"special": true
|
18 |
+
},
|
19 |
+
"2": {
|
20 |
+
"content": "<unk>",
|
21 |
+
"lstrip": false,
|
22 |
+
"normalized": false,
|
23 |
+
"rstrip": false,
|
24 |
+
"single_word": false,
|
25 |
+
"special": true
|
26 |
+
}
|
27 |
+
},
|
28 |
+
"additional_special_tokens": [],
|
29 |
"clean_up_tokenization_spaces": true,
|
30 |
"eos_token": "</s>",
|
31 |
"extra_ids": 0,
|
train_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
-
"epoch":
|
3 |
-
"train_loss": 1.
|
4 |
-
"train_runtime":
|
5 |
"train_samples": 65584,
|
6 |
-
"train_samples_per_second":
|
7 |
-
"train_steps_per_second":
|
8 |
}
|
|
|
1 |
{
|
2 |
+
"epoch": 15.0,
|
3 |
+
"train_loss": 1.153788715837463,
|
4 |
+
"train_runtime": 20083.7121,
|
5 |
"train_samples": 65584,
|
6 |
+
"train_samples_per_second": 65.311,
|
7 |
+
"train_steps_per_second": 0.51
|
8 |
}
|
trainer_state.json
CHANGED
@@ -1,8 +1,9 @@
|
|
1 |
{
|
2 |
-
"best_metric":
|
3 |
-
"best_model_checkpoint": "mt0-
|
4 |
-
"epoch":
|
5 |
-
"
|
|
|
6 |
"is_hyper_param_search": false,
|
7 |
"is_local_process_zero": true,
|
8 |
"is_world_process_zero": true,
|
@@ -10,168 +11,305 @@
|
|
10 |
{
|
11 |
"epoch": 1.0,
|
12 |
"learning_rate": 4.75e-05,
|
13 |
-
"loss": 2.
|
14 |
"step": 512
|
15 |
},
|
16 |
{
|
17 |
"epoch": 1.0,
|
18 |
-
"eval_gen_len": 16.
|
19 |
-
"eval_loss": 1.
|
20 |
-
"eval_rouge1":
|
21 |
-
"eval_rouge2":
|
22 |
-
"eval_rougeL":
|
23 |
-
"eval_rougeLsum":
|
24 |
-
"eval_runtime":
|
25 |
-
"eval_samples_per_second":
|
26 |
-
"eval_steps_per_second":
|
27 |
"step": 512
|
28 |
},
|
29 |
{
|
30 |
"epoch": 2.0,
|
31 |
"learning_rate": 4.4995117187500005e-05,
|
32 |
-
"loss": 1.
|
33 |
"step": 1025
|
34 |
},
|
35 |
{
|
36 |
"epoch": 2.0,
|
37 |
-
"eval_gen_len":
|
38 |
-
"eval_loss": 1.
|
39 |
-
"eval_rouge1":
|
40 |
-
"eval_rouge2":
|
41 |
-
"eval_rougeL":
|
42 |
-
"eval_rougeLsum":
|
43 |
-
"eval_runtime":
|
44 |
-
"eval_samples_per_second":
|
45 |
-
"eval_steps_per_second":
|
46 |
"step": 1025
|
47 |
},
|
48 |
{
|
49 |
"epoch": 3.0,
|
50 |
"learning_rate": 4.24951171875e-05,
|
51 |
-
"loss": 1.
|
52 |
"step": 1537
|
53 |
},
|
54 |
{
|
55 |
"epoch": 3.0,
|
56 |
-
"eval_gen_len": 16.
|
57 |
-
"eval_loss": 1.
|
58 |
-
"eval_rouge1":
|
59 |
-
"eval_rouge2":
|
60 |
-
"eval_rougeL":
|
61 |
-
"eval_rougeLsum":
|
62 |
-
"eval_runtime":
|
63 |
-
"eval_samples_per_second":
|
64 |
-
"eval_steps_per_second":
|
65 |
"step": 1537
|
66 |
},
|
67 |
{
|
68 |
"epoch": 4.0,
|
69 |
"learning_rate": 3.9990234375e-05,
|
70 |
-
"loss": 1.
|
71 |
"step": 2050
|
72 |
},
|
73 |
{
|
74 |
"epoch": 4.0,
|
75 |
-
"eval_gen_len": 16.
|
76 |
-
"eval_loss": 1.
|
77 |
-
"eval_rouge1":
|
78 |
-
"eval_rouge2":
|
79 |
-
"eval_rougeL":
|
80 |
-
"eval_rougeLsum":
|
81 |
-
"eval_runtime":
|
82 |
-
"eval_samples_per_second":
|
83 |
-
"eval_steps_per_second":
|
84 |
"step": 2050
|
85 |
},
|
86 |
{
|
87 |
"epoch": 5.0,
|
88 |
"learning_rate": 3.7490234375e-05,
|
89 |
-
"loss": 1.
|
90 |
"step": 2562
|
91 |
},
|
92 |
{
|
93 |
"epoch": 5.0,
|
94 |
-
"eval_gen_len":
|
95 |
-
"eval_loss": 1.
|
96 |
-
"eval_rouge1":
|
97 |
-
"eval_rouge2":
|
98 |
-
"eval_rougeL":
|
99 |
-
"eval_rougeLsum":
|
100 |
-
"eval_runtime":
|
101 |
-
"eval_samples_per_second": 42.
|
102 |
-
"eval_steps_per_second": 1.
|
103 |
"step": 2562
|
104 |
},
|
105 |
{
|
106 |
"epoch": 6.0,
|
107 |
"learning_rate": 3.49853515625e-05,
|
108 |
-
"loss": 1.
|
109 |
"step": 3075
|
110 |
},
|
111 |
{
|
112 |
"epoch": 6.0,
|
113 |
-
"eval_gen_len":
|
114 |
-
"eval_loss": 1.
|
115 |
-
"eval_rouge1":
|
116 |
-
"eval_rouge2":
|
117 |
-
"eval_rougeL":
|
118 |
-
"eval_rougeLsum":
|
119 |
-
"eval_runtime":
|
120 |
-
"eval_samples_per_second":
|
121 |
-
"eval_steps_per_second": 1.
|
122 |
"step": 3075
|
123 |
},
|
124 |
{
|
125 |
"epoch": 7.0,
|
126 |
"learning_rate": 3.2485351562499996e-05,
|
127 |
-
"loss": 1.
|
128 |
"step": 3587
|
129 |
},
|
130 |
{
|
131 |
"epoch": 7.0,
|
132 |
-
"eval_gen_len":
|
133 |
-
"eval_loss": 1.
|
134 |
-
"eval_rouge1":
|
135 |
-
"eval_rouge2":
|
136 |
-
"eval_rougeL":
|
137 |
-
"eval_rougeLsum":
|
138 |
-
"eval_runtime":
|
139 |
-
"eval_samples_per_second":
|
140 |
-
"eval_steps_per_second":
|
141 |
"step": 3587
|
142 |
},
|
143 |
{
|
144 |
"epoch": 8.0,
|
145 |
"learning_rate": 2.998046875e-05,
|
146 |
-
"loss": 1.
|
147 |
"step": 4100
|
148 |
},
|
149 |
{
|
150 |
"epoch": 8.0,
|
151 |
-
"eval_gen_len": 16.
|
152 |
-
"eval_loss": 1.
|
153 |
-
"eval_rouge1":
|
154 |
-
"eval_rouge2":
|
155 |
-
"eval_rougeL":
|
156 |
-
"eval_rougeLsum":
|
157 |
-
"eval_runtime":
|
158 |
-
"eval_samples_per_second":
|
159 |
-
"eval_steps_per_second":
|
160 |
"step": 4100
|
161 |
},
|
162 |
{
|
163 |
-
"epoch":
|
164 |
-
"
|
165 |
-
"
|
166 |
-
"
|
167 |
-
|
168 |
-
|
169 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
170 |
}
|
171 |
],
|
|
|
172 |
"max_steps": 10240,
|
|
|
173 |
"num_train_epochs": 20,
|
174 |
-
"
|
|
|
|
|
175 |
"trial_name": null,
|
176 |
"trial_params": null
|
177 |
}
|
|
|
1 |
{
|
2 |
+
"best_metric": 17.3273,
|
3 |
+
"best_model_checkpoint": "models/mt0-xl_russian_natprompt_adafactor_updated/checkpoint-6150",
|
4 |
+
"epoch": 14.999024390243903,
|
5 |
+
"eval_steps": 500,
|
6 |
+
"global_step": 7687,
|
7 |
"is_hyper_param_search": false,
|
8 |
"is_local_process_zero": true,
|
9 |
"is_world_process_zero": true,
|
|
|
11 |
{
|
12 |
"epoch": 1.0,
|
13 |
"learning_rate": 4.75e-05,
|
14 |
+
"loss": 2.0388,
|
15 |
"step": 512
|
16 |
},
|
17 |
{
|
18 |
"epoch": 1.0,
|
19 |
+
"eval_gen_len": 16.58484100877193,
|
20 |
+
"eval_loss": 1.6734575033187866,
|
21 |
+
"eval_rouge1": 14.1367,
|
22 |
+
"eval_rouge2": 7.0437,
|
23 |
+
"eval_rougeL": 14.0625,
|
24 |
+
"eval_rougeLsum": 14.0916,
|
25 |
+
"eval_runtime": 270.6111,
|
26 |
+
"eval_samples_per_second": 26.932,
|
27 |
+
"eval_steps_per_second": 0.843,
|
28 |
"step": 512
|
29 |
},
|
30 |
{
|
31 |
"epoch": 2.0,
|
32 |
"learning_rate": 4.4995117187500005e-05,
|
33 |
+
"loss": 1.7098,
|
34 |
"step": 1025
|
35 |
},
|
36 |
{
|
37 |
"epoch": 2.0,
|
38 |
+
"eval_gen_len": 16.68050986842105,
|
39 |
+
"eval_loss": 1.6203718185424805,
|
40 |
+
"eval_rouge1": 15.2619,
|
41 |
+
"eval_rouge2": 7.8124,
|
42 |
+
"eval_rougeL": 15.159,
|
43 |
+
"eval_rougeLsum": 15.2078,
|
44 |
+
"eval_runtime": 276.6842,
|
45 |
+
"eval_samples_per_second": 26.341,
|
46 |
+
"eval_steps_per_second": 0.824,
|
47 |
"step": 1025
|
48 |
},
|
49 |
{
|
50 |
"epoch": 3.0,
|
51 |
"learning_rate": 4.24951171875e-05,
|
52 |
+
"loss": 1.539,
|
53 |
"step": 1537
|
54 |
},
|
55 |
{
|
56 |
"epoch": 3.0,
|
57 |
+
"eval_gen_len": 16.61417214912281,
|
58 |
+
"eval_loss": 1.6058766841888428,
|
59 |
+
"eval_rouge1": 15.9942,
|
60 |
+
"eval_rouge2": 8.1827,
|
61 |
+
"eval_rougeL": 15.872,
|
62 |
+
"eval_rougeLsum": 15.9105,
|
63 |
+
"eval_runtime": 263.8074,
|
64 |
+
"eval_samples_per_second": 27.626,
|
65 |
+
"eval_steps_per_second": 0.864,
|
66 |
"step": 1537
|
67 |
},
|
68 |
{
|
69 |
"epoch": 4.0,
|
70 |
"learning_rate": 3.9990234375e-05,
|
71 |
+
"loss": 1.403,
|
72 |
"step": 2050
|
73 |
},
|
74 |
{
|
75 |
"epoch": 4.0,
|
76 |
+
"eval_gen_len": 16.26343201754386,
|
77 |
+
"eval_loss": 1.6041721105575562,
|
78 |
+
"eval_rouge1": 16.6383,
|
79 |
+
"eval_rouge2": 8.4603,
|
80 |
+
"eval_rougeL": 16.5096,
|
81 |
+
"eval_rougeLsum": 16.5635,
|
82 |
+
"eval_runtime": 251.4581,
|
83 |
+
"eval_samples_per_second": 28.983,
|
84 |
+
"eval_steps_per_second": 0.907,
|
85 |
"step": 2050
|
86 |
},
|
87 |
{
|
88 |
"epoch": 5.0,
|
89 |
"learning_rate": 3.7490234375e-05,
|
90 |
+
"loss": 1.295,
|
91 |
"step": 2562
|
92 |
},
|
93 |
{
|
94 |
"epoch": 5.0,
|
95 |
+
"eval_gen_len": 15.741365131578947,
|
96 |
+
"eval_loss": 1.6226089000701904,
|
97 |
+
"eval_rouge1": 16.9189,
|
98 |
+
"eval_rouge2": 8.8384,
|
99 |
+
"eval_rougeL": 16.7799,
|
100 |
+
"eval_rougeLsum": 16.8258,
|
101 |
+
"eval_runtime": 169.6881,
|
102 |
+
"eval_samples_per_second": 42.949,
|
103 |
+
"eval_steps_per_second": 1.344,
|
104 |
"step": 2562
|
105 |
},
|
106 |
{
|
107 |
"epoch": 6.0,
|
108 |
"learning_rate": 3.49853515625e-05,
|
109 |
+
"loss": 1.1984,
|
110 |
"step": 3075
|
111 |
},
|
112 |
{
|
113 |
"epoch": 6.0,
|
114 |
+
"eval_gen_len": 15.888157894736842,
|
115 |
+
"eval_loss": 1.6289030313491821,
|
116 |
+
"eval_rouge1": 16.9788,
|
117 |
+
"eval_rouge2": 8.7272,
|
118 |
+
"eval_rougeL": 16.8238,
|
119 |
+
"eval_rougeLsum": 16.8765,
|
120 |
+
"eval_runtime": 175.0677,
|
121 |
+
"eval_samples_per_second": 41.63,
|
122 |
+
"eval_steps_per_second": 1.302,
|
123 |
"step": 3075
|
124 |
},
|
125 |
{
|
126 |
"epoch": 7.0,
|
127 |
"learning_rate": 3.2485351562499996e-05,
|
128 |
+
"loss": 1.1195,
|
129 |
"step": 3587
|
130 |
},
|
131 |
{
|
132 |
"epoch": 7.0,
|
133 |
+
"eval_gen_len": 16.23519736842105,
|
134 |
+
"eval_loss": 1.6697918176651,
|
135 |
+
"eval_rouge1": 17.0912,
|
136 |
+
"eval_rouge2": 8.7061,
|
137 |
+
"eval_rougeL": 16.9084,
|
138 |
+
"eval_rougeLsum": 16.9633,
|
139 |
+
"eval_runtime": 171.9395,
|
140 |
+
"eval_samples_per_second": 42.387,
|
141 |
+
"eval_steps_per_second": 1.326,
|
142 |
"step": 3587
|
143 |
},
|
144 |
{
|
145 |
"epoch": 8.0,
|
146 |
"learning_rate": 2.998046875e-05,
|
147 |
+
"loss": 1.0463,
|
148 |
"step": 4100
|
149 |
},
|
150 |
{
|
151 |
"epoch": 8.0,
|
152 |
+
"eval_gen_len": 16.14761513157895,
|
153 |
+
"eval_loss": 1.6845269203186035,
|
154 |
+
"eval_rouge1": 17.201,
|
155 |
+
"eval_rouge2": 8.7395,
|
156 |
+
"eval_rougeL": 17.003,
|
157 |
+
"eval_rougeLsum": 17.052,
|
158 |
+
"eval_runtime": 252.7052,
|
159 |
+
"eval_samples_per_second": 28.84,
|
160 |
+
"eval_steps_per_second": 0.902,
|
161 |
"step": 4100
|
162 |
},
|
163 |
{
|
164 |
+
"epoch": 9.0,
|
165 |
+
"learning_rate": 2.748046875e-05,
|
166 |
+
"loss": 0.9866,
|
167 |
+
"step": 4612
|
168 |
+
},
|
169 |
+
{
|
170 |
+
"epoch": 9.0,
|
171 |
+
"eval_gen_len": 15.878837719298245,
|
172 |
+
"eval_loss": 1.726230502128601,
|
173 |
+
"eval_rouge1": 17.3223,
|
174 |
+
"eval_rouge2": 8.8289,
|
175 |
+
"eval_rougeL": 17.1413,
|
176 |
+
"eval_rougeLsum": 17.1756,
|
177 |
+
"eval_runtime": 182.5703,
|
178 |
+
"eval_samples_per_second": 39.919,
|
179 |
+
"eval_steps_per_second": 1.249,
|
180 |
+
"step": 4612
|
181 |
+
},
|
182 |
+
{
|
183 |
+
"epoch": 10.0,
|
184 |
+
"learning_rate": 2.49755859375e-05,
|
185 |
+
"loss": 0.9326,
|
186 |
+
"step": 5125
|
187 |
+
},
|
188 |
+
{
|
189 |
+
"epoch": 10.0,
|
190 |
+
"eval_gen_len": 15.797149122807017,
|
191 |
+
"eval_loss": 1.7532711029052734,
|
192 |
+
"eval_rouge1": 17.2655,
|
193 |
+
"eval_rouge2": 8.7512,
|
194 |
+
"eval_rougeL": 17.0508,
|
195 |
+
"eval_rougeLsum": 17.1055,
|
196 |
+
"eval_runtime": 168.7949,
|
197 |
+
"eval_samples_per_second": 43.177,
|
198 |
+
"eval_steps_per_second": 1.351,
|
199 |
+
"step": 5125
|
200 |
+
},
|
201 |
+
{
|
202 |
+
"epoch": 11.0,
|
203 |
+
"learning_rate": 2.24755859375e-05,
|
204 |
+
"loss": 0.8844,
|
205 |
+
"step": 5637
|
206 |
+
},
|
207 |
+
{
|
208 |
+
"epoch": 11.0,
|
209 |
+
"eval_gen_len": 16.32360197368421,
|
210 |
+
"eval_loss": 1.7794246673583984,
|
211 |
+
"eval_rouge1": 17.008,
|
212 |
+
"eval_rouge2": 8.5404,
|
213 |
+
"eval_rougeL": 16.8044,
|
214 |
+
"eval_rougeLsum": 16.848,
|
215 |
+
"eval_runtime": 168.6102,
|
216 |
+
"eval_samples_per_second": 43.224,
|
217 |
+
"eval_steps_per_second": 1.352,
|
218 |
+
"step": 5637
|
219 |
+
},
|
220 |
+
{
|
221 |
+
"epoch": 12.0,
|
222 |
+
"learning_rate": 1.9970703125e-05,
|
223 |
+
"loss": 0.8393,
|
224 |
+
"step": 6150
|
225 |
+
},
|
226 |
+
{
|
227 |
+
"epoch": 12.0,
|
228 |
+
"eval_gen_len": 16.143092105263158,
|
229 |
+
"eval_loss": 1.7995822429656982,
|
230 |
+
"eval_rouge1": 17.3273,
|
231 |
+
"eval_rouge2": 8.7829,
|
232 |
+
"eval_rougeL": 17.097,
|
233 |
+
"eval_rougeLsum": 17.1644,
|
234 |
+
"eval_runtime": 171.5723,
|
235 |
+
"eval_samples_per_second": 42.478,
|
236 |
+
"eval_steps_per_second": 1.329,
|
237 |
+
"step": 6150
|
238 |
+
},
|
239 |
+
{
|
240 |
+
"epoch": 13.0,
|
241 |
+
"learning_rate": 1.7470703125000003e-05,
|
242 |
+
"loss": 0.8046,
|
243 |
+
"step": 6662
|
244 |
+
},
|
245 |
+
{
|
246 |
+
"epoch": 13.0,
|
247 |
+
"eval_gen_len": 16.090597587719298,
|
248 |
+
"eval_loss": 1.8266295194625854,
|
249 |
+
"eval_rouge1": 17.1859,
|
250 |
+
"eval_rouge2": 8.6524,
|
251 |
+
"eval_rougeL": 16.9605,
|
252 |
+
"eval_rougeLsum": 17.0118,
|
253 |
+
"eval_runtime": 259.1646,
|
254 |
+
"eval_samples_per_second": 28.121,
|
255 |
+
"eval_steps_per_second": 0.88,
|
256 |
+
"step": 6662
|
257 |
+
},
|
258 |
+
{
|
259 |
+
"epoch": 14.0,
|
260 |
+
"learning_rate": 1.49658203125e-05,
|
261 |
+
"loss": 0.7682,
|
262 |
+
"step": 7175
|
263 |
+
},
|
264 |
+
{
|
265 |
+
"epoch": 14.0,
|
266 |
+
"eval_gen_len": 16.11239035087719,
|
267 |
+
"eval_loss": 1.8624775409698486,
|
268 |
+
"eval_rouge1": 17.0184,
|
269 |
+
"eval_rouge2": 8.5314,
|
270 |
+
"eval_rougeL": 16.8019,
|
271 |
+
"eval_rougeLsum": 16.847,
|
272 |
+
"eval_runtime": 170.9938,
|
273 |
+
"eval_samples_per_second": 42.621,
|
274 |
+
"eval_steps_per_second": 1.333,
|
275 |
+
"step": 7175
|
276 |
+
},
|
277 |
+
{
|
278 |
+
"epoch": 15.0,
|
279 |
+
"learning_rate": 1.2465820312500002e-05,
|
280 |
+
"loss": 0.7419,
|
281 |
+
"step": 7687
|
282 |
+
},
|
283 |
+
{
|
284 |
+
"epoch": 15.0,
|
285 |
+
"eval_gen_len": 15.95751096491228,
|
286 |
+
"eval_loss": 1.8779526948928833,
|
287 |
+
"eval_rouge1": 17.2742,
|
288 |
+
"eval_rouge2": 8.6795,
|
289 |
+
"eval_rougeL": 17.0699,
|
290 |
+
"eval_rougeLsum": 17.1118,
|
291 |
+
"eval_runtime": 177.9916,
|
292 |
+
"eval_samples_per_second": 40.946,
|
293 |
+
"eval_steps_per_second": 1.281,
|
294 |
+
"step": 7687
|
295 |
+
},
|
296 |
+
{
|
297 |
+
"epoch": 15.0,
|
298 |
+
"step": 7687,
|
299 |
+
"total_flos": 1.7085595424946913e+18,
|
300 |
+
"train_loss": 1.153788715837463,
|
301 |
+
"train_runtime": 20083.7121,
|
302 |
+
"train_samples_per_second": 65.311,
|
303 |
+
"train_steps_per_second": 0.51
|
304 |
}
|
305 |
],
|
306 |
+
"logging_steps": 500,
|
307 |
"max_steps": 10240,
|
308 |
+
"num_input_tokens_seen": 0,
|
309 |
"num_train_epochs": 20,
|
310 |
+
"save_steps": 500,
|
311 |
+
"total_flos": 1.7085595424946913e+18,
|
312 |
+
"train_batch_size": 4,
|
313 |
"trial_name": null,
|
314 |
"trial_params": null
|
315 |
}
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b05c536c471b5be16fe45bc4130c51587d993402ce5ba1bd7ba28b30f5b50b5b
|
3 |
+
size 4411
|
upload.py
DELETED
@@ -1,11 +0,0 @@
|
|
1 |
-
#!/bin/env python3
|
2 |
-
|
3 |
-
import sys
|
4 |
-
from huggingface_hub import HfApi
|
5 |
-
from huggingface_hub import create_repo
|
6 |
-
|
7 |
-
create_repo(sys.argv[1])
|
8 |
-
api = HfApi()
|
9 |
-
|
10 |
-
api.upload_folder(folder_path=".", repo_id=sys.argv[1], repo_type="model")
|
11 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|