kennethge123
commited on
Commit
•
ab00479
1
Parent(s):
bc5d5c7
Training in progress, step 500
Browse files- config.json +68 -0
- model.safetensors +3 -0
- runs/Mar29_11-10-09_dsail2/events.out.tfevents.1711725010.dsail2.11956.2 +3 -0
- runs/Mar29_15-54-50_dsail2/events.out.tfevents.1711742092.dsail2.13672.0 +3 -0
- runs/Mar29_15-55-11_dsail2/events.out.tfevents.1711742113.dsail2.13857.0 +3 -0
- runs/Mar29_18-39-11_dsail2/events.out.tfevents.1711751953.dsail2.20777.0 +3 -0
- runs/Mar29_18-59-28_dsail2/events.out.tfevents.1711753170.dsail2.25439.0 +3 -0
- runs/Mar29_21-07-44_dsail2/events.out.tfevents.1711760866.dsail2.32250.0 +3 -0
- training_args.bin +3 -0
config.json
ADDED
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "t5-base",
|
3 |
+
"architectures": [
|
4 |
+
"T5ForSequenceClassification"
|
5 |
+
],
|
6 |
+
"classifier_dropout": 0.0,
|
7 |
+
"d_ff": 3072,
|
8 |
+
"d_kv": 64,
|
9 |
+
"d_model": 768,
|
10 |
+
"decoder_start_token_id": 0,
|
11 |
+
"dense_act_fn": "relu",
|
12 |
+
"dropout_rate": 0.1,
|
13 |
+
"eos_token_id": 1,
|
14 |
+
"feed_forward_proj": "relu",
|
15 |
+
"id2label": {
|
16 |
+
"0": "LABEL_0"
|
17 |
+
},
|
18 |
+
"initializer_factor": 1.0,
|
19 |
+
"is_encoder_decoder": true,
|
20 |
+
"is_gated_act": false,
|
21 |
+
"label2id": {
|
22 |
+
"LABEL_0": 0
|
23 |
+
},
|
24 |
+
"layer_norm_epsilon": 1e-06,
|
25 |
+
"model_type": "t5",
|
26 |
+
"n_positions": 512,
|
27 |
+
"num_decoder_layers": 12,
|
28 |
+
"num_heads": 12,
|
29 |
+
"num_layers": 12,
|
30 |
+
"output_past": true,
|
31 |
+
"pad_token_id": 0,
|
32 |
+
"problem_type": "regression",
|
33 |
+
"relative_attention_max_distance": 128,
|
34 |
+
"relative_attention_num_buckets": 32,
|
35 |
+
"task_specific_params": {
|
36 |
+
"summarization": {
|
37 |
+
"early_stopping": true,
|
38 |
+
"length_penalty": 2.0,
|
39 |
+
"max_length": 200,
|
40 |
+
"min_length": 30,
|
41 |
+
"no_repeat_ngram_size": 3,
|
42 |
+
"num_beams": 4,
|
43 |
+
"prefix": "summarize: "
|
44 |
+
},
|
45 |
+
"translation_en_to_de": {
|
46 |
+
"early_stopping": true,
|
47 |
+
"max_length": 300,
|
48 |
+
"num_beams": 4,
|
49 |
+
"prefix": "translate English to German: "
|
50 |
+
},
|
51 |
+
"translation_en_to_fr": {
|
52 |
+
"early_stopping": true,
|
53 |
+
"max_length": 300,
|
54 |
+
"num_beams": 4,
|
55 |
+
"prefix": "translate English to French: "
|
56 |
+
},
|
57 |
+
"translation_en_to_ro": {
|
58 |
+
"early_stopping": true,
|
59 |
+
"max_length": 300,
|
60 |
+
"num_beams": 4,
|
61 |
+
"prefix": "translate English to Romanian: "
|
62 |
+
}
|
63 |
+
},
|
64 |
+
"torch_dtype": "float32",
|
65 |
+
"transformers_version": "4.37.0",
|
66 |
+
"use_cache": true,
|
67 |
+
"vocab_size": 32128
|
68 |
+
}
|
model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0fad4d1bc4efef464104279bf096aee43002c6d6bd41f5b50777322f8d92c3bf
|
3 |
+
size 894013636
|
runs/Mar29_11-10-09_dsail2/events.out.tfevents.1711725010.dsail2.11956.2
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d87318d58fb2196945b72040e024bc07cc2d5e7bf1ac0321fabbf14ee93ddf2f
|
3 |
+
size 5237
|
runs/Mar29_15-54-50_dsail2/events.out.tfevents.1711742092.dsail2.13672.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f64c603e15d4225cb83f0972ad87411ee8a3414cb2c639aa985121d3cea7cbed
|
3 |
+
size 5237
|
runs/Mar29_15-55-11_dsail2/events.out.tfevents.1711742113.dsail2.13857.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d430c7603ca695a9489528645651698b564a6b52c1b7d36cb4a726c8e73d5cff
|
3 |
+
size 5237
|
runs/Mar29_18-39-11_dsail2/events.out.tfevents.1711751953.dsail2.20777.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8fdee0a2505799721d6da80a1895c85ed3e56849d2cdc0fa4106ff89a0804384
|
3 |
+
size 5237
|
runs/Mar29_18-59-28_dsail2/events.out.tfevents.1711753170.dsail2.25439.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3809b2c5d430d44a06d73e063a291455de129af5cc55588d69aa903da43aa5f1
|
3 |
+
size 5237
|
runs/Mar29_21-07-44_dsail2/events.out.tfevents.1711760866.dsail2.32250.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8a532262df1dca609d78cfe80987ffb8d3ab46a77c20fc60f5272536561b3801
|
3 |
+
size 5712
|
training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:db7bf198624b541e59367651a815dae002ea7145b93b77d28149d45ae36ca4ad
|
3 |
+
size 4347
|