Training in progress, step 100
Browse files- .gitignore +1 -0
- config.json +31 -0
- pytorch_model.bin +3 -0
- runs/May29_11-53-00_6074c25d8e89/1685361191.4538476/events.out.tfevents.1685361191.6074c25d8e89.773.1 +3 -0
- runs/May29_11-53-00_6074c25d8e89/events.out.tfevents.1685361191.6074c25d8e89.773.0 +3 -0
- runs/May29_13-09-52_6074c25d8e89/1685365796.5664134/events.out.tfevents.1685365796.6074c25d8e89.773.3 +3 -0
- runs/May29_13-09-52_6074c25d8e89/1685365803.1838062/events.out.tfevents.1685365803.6074c25d8e89.773.4 +3 -0
- runs/May29_13-09-52_6074c25d8e89/events.out.tfevents.1685365796.6074c25d8e89.773.2 +3 -0
- runs/May29_13-10-32_6074c25d8e89/1685365837.1881313/events.out.tfevents.1685365837.6074c25d8e89.773.6 +3 -0
- runs/May29_13-10-32_6074c25d8e89/1685365843.169542/events.out.tfevents.1685365843.6074c25d8e89.773.7 +3 -0
- runs/May29_13-10-32_6074c25d8e89/events.out.tfevents.1685365837.6074c25d8e89.773.5 +3 -0
- runs/May29_13-14-30_6074c25d8e89/1685366074.6791286/events.out.tfevents.1685366074.6074c25d8e89.773.9 +3 -0
- runs/May29_13-14-30_6074c25d8e89/1685366080.548917/events.out.tfevents.1685366080.6074c25d8e89.773.10 +3 -0
- runs/May29_13-14-30_6074c25d8e89/events.out.tfevents.1685366074.6074c25d8e89.773.8 +3 -0
- runs/May29_13-31-38_6074c25d8e89/1685367103.6575532/events.out.tfevents.1685367103.6074c25d8e89.773.12 +3 -0
- runs/May29_13-31-38_6074c25d8e89/1685367109.5117726/events.out.tfevents.1685367109.6074c25d8e89.773.13 +3 -0
- runs/May29_13-31-38_6074c25d8e89/events.out.tfevents.1685367103.6074c25d8e89.773.11 +3 -0
- runs/May29_13-39-25_6074c25d8e89/1685367570.2322803/events.out.tfevents.1685367570.6074c25d8e89.773.15 +3 -0
- runs/May29_13-39-25_6074c25d8e89/1685367577.0492318/events.out.tfevents.1685367577.6074c25d8e89.773.16 +3 -0
- runs/May29_13-39-25_6074c25d8e89/events.out.tfevents.1685367570.6074c25d8e89.773.14 +3 -0
- runs/May29_13-40-25_6074c25d8e89/1685367630.035019/events.out.tfevents.1685367630.6074c25d8e89.773.18 +3 -0
- runs/May29_13-40-25_6074c25d8e89/1685367636.9695227/events.out.tfevents.1685367636.6074c25d8e89.773.19 +3 -0
- runs/May29_13-40-25_6074c25d8e89/events.out.tfevents.1685367630.6074c25d8e89.773.17 +3 -0
- runs/May29_14-09-36_6074c25d8e89/1685369381.8537831/events.out.tfevents.1685369381.6074c25d8e89.773.21 +3 -0
- runs/May29_14-09-36_6074c25d8e89/1685369388.7845886/events.out.tfevents.1685369388.6074c25d8e89.773.22 +3 -0
- runs/May29_14-09-36_6074c25d8e89/events.out.tfevents.1685369381.6074c25d8e89.773.20 +3 -0
- runs/May29_14-12-50_6074c25d8e89/1685369574.9372284/events.out.tfevents.1685369574.6074c25d8e89.773.24 +3 -0
- runs/May29_14-12-50_6074c25d8e89/1685369581.8439982/events.out.tfevents.1685369581.6074c25d8e89.773.25 +3 -0
- runs/May29_14-12-50_6074c25d8e89/events.out.tfevents.1685369574.6074c25d8e89.773.23 +3 -0
- training_args.bin +3 -0
.gitignore
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
checkpoint-*/
|
config.json
ADDED
|
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"activation_function": "gelu_new",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"GPT2LMHeadModel"
|
| 5 |
+
],
|
| 6 |
+
"attn_pdrop": 0.1,
|
| 7 |
+
"bos_token_id": 50256,
|
| 8 |
+
"embd_pdrop": 0.1,
|
| 9 |
+
"eos_token_id": 50256,
|
| 10 |
+
"initializer_range": 0.02,
|
| 11 |
+
"layer_norm_epsilon": 1e-05,
|
| 12 |
+
"model_type": "gpt2",
|
| 13 |
+
"n_embd": 512,
|
| 14 |
+
"n_head": 4,
|
| 15 |
+
"n_inner": 1024,
|
| 16 |
+
"n_layer": 4,
|
| 17 |
+
"n_positions": 514,
|
| 18 |
+
"reorder_and_upcast_attn": false,
|
| 19 |
+
"resid_pdrop": 0.1,
|
| 20 |
+
"scale_attn_by_inverse_layer_idx": false,
|
| 21 |
+
"scale_attn_weights": true,
|
| 22 |
+
"summary_activation": null,
|
| 23 |
+
"summary_first_dropout": 0.1,
|
| 24 |
+
"summary_proj_to_labels": true,
|
| 25 |
+
"summary_type": "cls_index",
|
| 26 |
+
"summary_use_proj": true,
|
| 27 |
+
"torch_dtype": "float32",
|
| 28 |
+
"transformers_version": "4.29.2",
|
| 29 |
+
"use_cache": true,
|
| 30 |
+
"vocab_size": 50258
|
| 31 |
+
}
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:12551ca54723c5944350331d5ffa46aa7df5ce3e4cbaf7ce69eacb3c977a1a73
|
| 3 |
+
size 138705765
|
runs/May29_11-53-00_6074c25d8e89/1685361191.4538476/events.out.tfevents.1685361191.6074c25d8e89.773.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8b8daedb63b7df6c096b064e27e17959f38848ea59feab17151e9295039c217a
|
| 3 |
+
size 6266
|
runs/May29_11-53-00_6074c25d8e89/events.out.tfevents.1685361191.6074c25d8e89.773.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:37556b11bdbdf355b880318e69d6acf82e8f8ec4f7dace2a68dbae222546616a
|
| 3 |
+
size 5359
|
runs/May29_13-09-52_6074c25d8e89/1685365796.5664134/events.out.tfevents.1685365796.6074c25d8e89.773.3
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9e25ba7f8eb35f780f8acb5c42b7d4f85114b2fdcb64a7ea138b469859b0b035
|
| 3 |
+
size 6266
|
runs/May29_13-09-52_6074c25d8e89/1685365803.1838062/events.out.tfevents.1685365803.6074c25d8e89.773.4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fd6cd90a3463df1d73a8b73d3900d08ca127c3132f9824b6ea97393d1f2e10cf
|
| 3 |
+
size 6266
|
runs/May29_13-09-52_6074c25d8e89/events.out.tfevents.1685365796.6074c25d8e89.773.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d582026949dc4e0a799b619a2de4b0c8acf4f12b31c80f216623a68bff13d92b
|
| 3 |
+
size 8442
|
runs/May29_13-10-32_6074c25d8e89/1685365837.1881313/events.out.tfevents.1685365837.6074c25d8e89.773.6
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7855bb3d8fd1d2bfd08f58a5ac947019bbf7c9840e0ab9ee70bef14fe2bf2f0d
|
| 3 |
+
size 6266
|
runs/May29_13-10-32_6074c25d8e89/1685365843.169542/events.out.tfevents.1685365843.6074c25d8e89.773.7
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:49f522eb0e63f782fd3df561e218e49f403df995eeb8f4de6d8169a93e995574
|
| 3 |
+
size 6266
|
runs/May29_13-10-32_6074c25d8e89/events.out.tfevents.1685365837.6074c25d8e89.773.5
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2f848d6e411fac983670d2b4f0103df6798094349cae20ba85ee0361af9cdfe7
|
| 3 |
+
size 8442
|
runs/May29_13-14-30_6074c25d8e89/1685366074.6791286/events.out.tfevents.1685366074.6074c25d8e89.773.9
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f545c2a5e693b7888c895fabff86bca64e2e5aba24a1cd02addb0b51a0f0f41d
|
| 3 |
+
size 6266
|
runs/May29_13-14-30_6074c25d8e89/1685366080.548917/events.out.tfevents.1685366080.6074c25d8e89.773.10
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1ee4d8134367e547057d98431cc64580186c761ad3002d6b67ee01f5cdf9472e
|
| 3 |
+
size 6266
|
runs/May29_13-14-30_6074c25d8e89/events.out.tfevents.1685366074.6074c25d8e89.773.8
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c82ff1f67ca4f81b932f532b1eacf0410b43c3cdcd70473b76d9ab5d3de828a8
|
| 3 |
+
size 9063
|
runs/May29_13-31-38_6074c25d8e89/1685367103.6575532/events.out.tfevents.1685367103.6074c25d8e89.773.12
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:821606e805bfd66659a6215edc9696d3ff71a164d20c03a10605b2169435c9ed
|
| 3 |
+
size 6266
|
runs/May29_13-31-38_6074c25d8e89/1685367109.5117726/events.out.tfevents.1685367109.6074c25d8e89.773.13
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:84520affc77dbce996b78cd1e0d63d3c3867f00e51f2654661354d72dc432cae
|
| 3 |
+
size 6266
|
runs/May29_13-31-38_6074c25d8e89/events.out.tfevents.1685367103.6074c25d8e89.773.11
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d0f425967e246d667b0bb04979a2712399ab1987ab2f5def8d1ffd1bbc0985b8
|
| 3 |
+
size 8438
|
runs/May29_13-39-25_6074c25d8e89/1685367570.2322803/events.out.tfevents.1685367570.6074c25d8e89.773.15
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:db5c79b331a0f9e6313067b24e79d5773ba0b0ffe94e1712d0dd83e09c2ca280
|
| 3 |
+
size 6266
|
runs/May29_13-39-25_6074c25d8e89/1685367577.0492318/events.out.tfevents.1685367577.6074c25d8e89.773.16
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:184e928eb503771f7f89df168459ad51250094eafd74d1c9658baca631229cda
|
| 3 |
+
size 6266
|
runs/May29_13-39-25_6074c25d8e89/events.out.tfevents.1685367570.6074c25d8e89.773.14
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:067a73cc27a0d30b82c8b417e3837f4b5010feacd6bf81bdd7a2bd7d2f8767f0
|
| 3 |
+
size 8438
|
runs/May29_13-40-25_6074c25d8e89/1685367630.035019/events.out.tfevents.1685367630.6074c25d8e89.773.18
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c2c17c946e4f309bf9c3db84a95e5dfd4824e108ea737bd47993e8df1b32bb83
|
| 3 |
+
size 6266
|
runs/May29_13-40-25_6074c25d8e89/1685367636.9695227/events.out.tfevents.1685367636.6074c25d8e89.773.19
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8b9c8b560961929d79aeefada7527774f694e06ea4928b94925d7f4bb26f1e68
|
| 3 |
+
size 6266
|
runs/May29_13-40-25_6074c25d8e89/events.out.tfevents.1685367630.6074c25d8e89.773.17
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:afd67b9108b22c8c4fe148bea42f8de3a4d10801b317abd31480c0d151e7d7a6
|
| 3 |
+
size 10005
|
runs/May29_14-09-36_6074c25d8e89/1685369381.8537831/events.out.tfevents.1685369381.6074c25d8e89.773.21
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cc9ea20f750799da80e72b857f2c8b18ef183a965a991ed92999e81e762823b5
|
| 3 |
+
size 6266
|
runs/May29_14-09-36_6074c25d8e89/1685369388.7845886/events.out.tfevents.1685369388.6074c25d8e89.773.22
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:12c85492df3e5c920c20c0156304138dad30247d170470ec478b1c938a3a7b35
|
| 3 |
+
size 6266
|
runs/May29_14-09-36_6074c25d8e89/events.out.tfevents.1685369381.6074c25d8e89.773.20
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b2d76f3dc665a5f42dbfe0b9188c9cc92b2f36b44a48669c85ae3aeaab91f785
|
| 3 |
+
size 8592
|
runs/May29_14-12-50_6074c25d8e89/1685369574.9372284/events.out.tfevents.1685369574.6074c25d8e89.773.24
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:670daccf40a6bcf537ee164f8b49c2c3594b387b051e8aa95b69e873486390ca
|
| 3 |
+
size 6266
|
runs/May29_14-12-50_6074c25d8e89/1685369581.8439982/events.out.tfevents.1685369581.6074c25d8e89.773.25
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9d84ddb87416b65e17f5ec5d76ad4ffaf91fb04ed920ab9bbb7048f5b5c8d4f6
|
| 3 |
+
size 6266
|
runs/May29_14-12-50_6074c25d8e89/events.out.tfevents.1685369574.6074c25d8e89.773.23
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9a5cae4f2a8eb92f658ec005e60b0889b573be23ba4715d9e87b300c8b8b2295
|
| 3 |
+
size 8588
|
training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9d2e5803c89d8541a68b80d9a52bee9d6ce8182e3a456dd5ab20fda5ed8e007f
|
| 3 |
+
size 4091
|