robingeibel
commited on
Commit
•
5c52976
1
Parent(s):
197b368
End of training
Browse files- .gitignore +1 -0
- config.json +34 -0
- pytorch_model.bin +3 -0
- runs/Jun28_12-52-49_586d5df76fdc/1656431618.4676623/events.out.tfevents.1656431618.586d5df76fdc.97.1 +3 -0
- runs/Jun28_12-52-49_586d5df76fdc/1656432473.6572824/events.out.tfevents.1656432473.586d5df76fdc.97.3 +3 -0
- runs/Jun28_12-52-49_586d5df76fdc/events.out.tfevents.1656431611.586d5df76fdc.97.0 +3 -0
- runs/Jun28_12-52-49_586d5df76fdc/events.out.tfevents.1656432473.586d5df76fdc.97.2 +3 -0
- training_args.bin +3 -0
.gitignore
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
checkpoint-*/
|
config.json
ADDED
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "google/bigbird-roberta-large",
|
3 |
+
"architectures": [
|
4 |
+
"BigBirdForMaskedLM"
|
5 |
+
],
|
6 |
+
"attention_probs_dropout_prob": 0.1,
|
7 |
+
"attention_type": "block_sparse",
|
8 |
+
"block_size": 64,
|
9 |
+
"bos_token_id": 1,
|
10 |
+
"classifier_dropout": null,
|
11 |
+
"eos_token_id": 2,
|
12 |
+
"gradient_checkpointing": false,
|
13 |
+
"hidden_act": "gelu_new",
|
14 |
+
"hidden_dropout_prob": 0.1,
|
15 |
+
"hidden_size": 1024,
|
16 |
+
"initializer_range": 0.02,
|
17 |
+
"intermediate_size": 4096,
|
18 |
+
"layer_norm_eps": 1e-12,
|
19 |
+
"max_position_embeddings": 4096,
|
20 |
+
"model_type": "big_bird",
|
21 |
+
"num_attention_heads": 16,
|
22 |
+
"num_hidden_layers": 24,
|
23 |
+
"num_random_blocks": 3,
|
24 |
+
"pad_token_id": 0,
|
25 |
+
"position_embedding_type": "absolute",
|
26 |
+
"rescale_embeddings": false,
|
27 |
+
"sep_token_id": 66,
|
28 |
+
"torch_dtype": "float32",
|
29 |
+
"transformers_version": "4.20.1",
|
30 |
+
"type_vocab_size": 2,
|
31 |
+
"use_bias": true,
|
32 |
+
"use_cache": true,
|
33 |
+
"vocab_size": 50358
|
34 |
+
}
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e50e641b6c1db942f952312fdff1a49f548e54f058f1c5bb96bf4260bcfb3155
|
3 |
+
size 1441069671
|
runs/Jun28_12-52-49_586d5df76fdc/1656431618.4676623/events.out.tfevents.1656431618.586d5df76fdc.97.1
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:78c11d01e3b7c944da6c178cf49f392abe970eb63aa13b49394d058906633e9f
|
3 |
+
size 5372
|
runs/Jun28_12-52-49_586d5df76fdc/1656432473.6572824/events.out.tfevents.1656432473.586d5df76fdc.97.3
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d14c302281ee0a388fa69cec0faaca02fa0f310750dc3978c684eb10f459da18
|
3 |
+
size 5372
|
runs/Jun28_12-52-49_586d5df76fdc/events.out.tfevents.1656431611.586d5df76fdc.97.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3a8e1549b0fb66369463a749f5bea843086db8019356f1d1dc43fabff83f69dd
|
3 |
+
size 4078
|
runs/Jun28_12-52-49_586d5df76fdc/events.out.tfevents.1656432473.586d5df76fdc.97.2
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:05d94529d24ba181a1af4f55c988f52d7cf1aa6db3040ff8cfeb1d5089846d6f
|
3 |
+
size 5139
|
training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:655b30667283dae54a9dce24c5f53ad9b3a8c385cfb05e6e32971d9720a3eba6
|
3 |
+
size 3375
|