Commit
Β·
39a714b
1
Parent(s):
44b1d94
Training in progress, step 7700
Browse files- adapter_model.bin +1 -1
- {checkpoint-7300 β checkpoint-7600/adapter_model}/README.md +0 -0
- {checkpoint-7300 β checkpoint-7600/adapter_model}/adapter_config.json +0 -0
- {checkpoint-7300 β checkpoint-7600/adapter_model}/adapter_model.bin +1 -1
- {checkpoint-7300/adapter_model β checkpoint-7700}/README.md +0 -0
- {checkpoint-7300/adapter_model β checkpoint-7700}/adapter_config.json +0 -0
- {checkpoint-7300/adapter_model β checkpoint-7700}/adapter_model.bin +1 -1
- {checkpoint-7300 β checkpoint-7700}/optimizer.pt +1 -1
- {checkpoint-7300 β checkpoint-7700}/rng_state_0.pth +1 -1
- {checkpoint-7300 β checkpoint-7700}/rng_state_1.pth +1 -1
- {checkpoint-7300 β checkpoint-7700}/rng_state_10.pth +1 -1
- {checkpoint-7300 β checkpoint-7700}/rng_state_11.pth +1 -1
- {checkpoint-7300 β checkpoint-7700}/rng_state_12.pth +1 -1
- {checkpoint-7300 β checkpoint-7700}/rng_state_13.pth +1 -1
- {checkpoint-7300 β checkpoint-7700}/rng_state_2.pth +1 -1
- {checkpoint-7300 β checkpoint-7700}/rng_state_3.pth +1 -1
- {checkpoint-7300 β checkpoint-7700}/rng_state_4.pth +1 -1
- {checkpoint-7300 β checkpoint-7700}/rng_state_5.pth +1 -1
- {checkpoint-7300 β checkpoint-7700}/rng_state_6.pth +1 -1
- {checkpoint-7300 β checkpoint-7700}/rng_state_7.pth +1 -1
- {checkpoint-7300 β checkpoint-7700}/rng_state_8.pth +1 -1
- {checkpoint-7300 β checkpoint-7700}/rng_state_9.pth +1 -1
- {checkpoint-7300 β checkpoint-7700}/scheduler.pt +1 -1
- {checkpoint-7300 β checkpoint-7700}/trainer_state.json +99 -3
- {checkpoint-7300 β checkpoint-7700}/training_args.bin +0 -0
adapter_model.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 500897101
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:988146f64d3c7cf9b370459002ad232cfc0b56592fe380ed0c11fd12cc50f240
|
3 |
size 500897101
|
{checkpoint-7300 β checkpoint-7600/adapter_model}/README.md
RENAMED
File without changes
|
{checkpoint-7300 β checkpoint-7600/adapter_model}/adapter_config.json
RENAMED
File without changes
|
{checkpoint-7300 β checkpoint-7600/adapter_model}/adapter_model.bin
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 500897101
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a5ab034d117a0225da8be2f42841fdd56781cdd412727c42b29659f8fe5e1326
|
3 |
size 500897101
|
{checkpoint-7300/adapter_model β checkpoint-7700}/README.md
RENAMED
File without changes
|
{checkpoint-7300/adapter_model β checkpoint-7700}/adapter_config.json
RENAMED
File without changes
|
{checkpoint-7300/adapter_model β checkpoint-7700}/adapter_model.bin
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 500897101
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:988146f64d3c7cf9b370459002ad232cfc0b56592fe380ed0c11fd12cc50f240
|
3 |
size 500897101
|
{checkpoint-7300 β checkpoint-7700}/optimizer.pt
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 1001752701
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7036379c7af7f50ac5d0fd32c6632909552ae86308d031266bb1be1ee7f726a7
|
3 |
size 1001752701
|
{checkpoint-7300 β checkpoint-7700}/rng_state_0.pth
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 27772
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9d7a823cd19b1c1e55658e4089d37e90c3209a62a7b02b3363e7dd9530319361
|
3 |
size 27772
|
{checkpoint-7300 β checkpoint-7700}/rng_state_1.pth
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 27772
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9f75000ee4ae374bf6186ef3da73abad807d802ad546e9f81652298dae7be4a0
|
3 |
size 27772
|
{checkpoint-7300 β checkpoint-7700}/rng_state_10.pth
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 27789
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f9e3c371c42035aaf3fe5bb37e45f81f546540cdffab238c5b3421ca2ee0b1e6
|
3 |
size 27789
|
{checkpoint-7300 β checkpoint-7700}/rng_state_11.pth
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 27789
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a68bb70ded324b93ee37388305ab67d5051182eee9352729fda4ea57a23f78ce
|
3 |
size 27789
|
{checkpoint-7300 β checkpoint-7700}/rng_state_12.pth
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 27789
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a259f42dc77f61f2e6dc8611ebed7094184a6edd571aea753d57e6ac4562f87d
|
3 |
size 27789
|
{checkpoint-7300 β checkpoint-7700}/rng_state_13.pth
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 27789
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3967f7cebc9724687864fa3cf7b2ed95832c68b1fd3d7294d6d097397ab84eb5
|
3 |
size 27789
|
{checkpoint-7300 β checkpoint-7700}/rng_state_2.pth
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 27772
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3216b7571ea37b63ff94e5607a054dcbfd8c9415eb063e591cdf997477202a87
|
3 |
size 27772
|
{checkpoint-7300 β checkpoint-7700}/rng_state_3.pth
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 27772
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:60d51d6a2f52e1b9da0902663ad1dc37111cd5e4a88cfc61419222968d15dda7
|
3 |
size 27772
|
{checkpoint-7300 β checkpoint-7700}/rng_state_4.pth
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 27772
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3d76790e35bd628e63a3bd8ccb102583b70f62c643a2ccab6fea069ecd54ee87
|
3 |
size 27772
|
{checkpoint-7300 β checkpoint-7700}/rng_state_5.pth
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 27772
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e2cfb9a2b181fd4df9dc171415bed936d48362c3541e157302fe59c1ca024fbb
|
3 |
size 27772
|
{checkpoint-7300 β checkpoint-7700}/rng_state_6.pth
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 27772
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a111f33c6edd5882a15a51b0f4335da5136c22fc1777a4370c074cc971dc02e5
|
3 |
size 27772
|
{checkpoint-7300 β checkpoint-7700}/rng_state_7.pth
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 27772
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:12ee96adce3b6144ad1b7c1915e19f8c8026b8d87640cf36c38622c32be68962
|
3 |
size 27772
|
{checkpoint-7300 β checkpoint-7700}/rng_state_8.pth
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 27772
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:500263745456537babfd93f2cfcd87a347e0671a2cb22bb48b6bfce91af74942
|
3 |
size 27772
|
{checkpoint-7300 β checkpoint-7700}/rng_state_9.pth
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 27772
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b41ee1a2e9e4bac7897f987b0682669b950f4f4a0c93824363caa035231da564
|
3 |
size 27772
|
{checkpoint-7300 β checkpoint-7700}/scheduler.pt
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 627
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a0876d0cf49ff63c2b677b32345ac6adfbd02210c5fb233bc789a7b7abacd6fb
|
3 |
size 627
|
{checkpoint-7300 β checkpoint-7700}/trainer_state.json
RENAMED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"best_metric": null,
|
3 |
"best_model_checkpoint": null,
|
4 |
-
"epoch": 2.
|
5 |
-
"global_step":
|
6 |
"is_hyper_param_search": false,
|
7 |
"is_local_process_zero": true,
|
8 |
"is_world_process_zero": true,
|
@@ -1790,11 +1790,107 @@
|
|
1790 |
"learning_rate": 1.5742294436213223e-06,
|
1791 |
"loss": 0.7703,
|
1792 |
"step": 7300
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1793 |
}
|
1794 |
],
|
1795 |
"max_steps": 7737,
|
1796 |
"num_train_epochs": 3,
|
1797 |
-
"total_flos": 3.
|
1798 |
"trial_name": null,
|
1799 |
"trial_params": null
|
1800 |
}
|
|
|
1 |
{
|
2 |
"best_metric": null,
|
3 |
"best_model_checkpoint": null,
|
4 |
+
"epoch": 2.9856533540131833,
|
5 |
+
"global_step": 7700,
|
6 |
"is_hyper_param_search": false,
|
7 |
"is_local_process_zero": true,
|
8 |
"is_world_process_zero": true,
|
|
|
1790 |
"learning_rate": 1.5742294436213223e-06,
|
1791 |
"loss": 0.7703,
|
1792 |
"step": 7300
|
1793 |
+
},
|
1794 |
+
{
|
1795 |
+
"epoch": 2.84,
|
1796 |
+
"learning_rate": 1.3996731458185697e-06,
|
1797 |
+
"loss": 0.7618,
|
1798 |
+
"step": 7325
|
1799 |
+
},
|
1800 |
+
{
|
1801 |
+
"epoch": 2.85,
|
1802 |
+
"learning_rate": 1.2353035254395352e-06,
|
1803 |
+
"loss": 0.7767,
|
1804 |
+
"step": 7350
|
1805 |
+
},
|
1806 |
+
{
|
1807 |
+
"epoch": 2.86,
|
1808 |
+
"learning_rate": 1.0811375639725341e-06,
|
1809 |
+
"loss": 0.7659,
|
1810 |
+
"step": 7375
|
1811 |
+
},
|
1812 |
+
{
|
1813 |
+
"epoch": 2.87,
|
1814 |
+
"learning_rate": 9.371911887371965e-07,
|
1815 |
+
"loss": 0.774,
|
1816 |
+
"step": 7400
|
1817 |
+
},
|
1818 |
+
{
|
1819 |
+
"epoch": 2.88,
|
1820 |
+
"learning_rate": 8.034792712388828e-07,
|
1821 |
+
"loss": 0.7658,
|
1822 |
+
"step": 7425
|
1823 |
+
},
|
1824 |
+
{
|
1825 |
+
"epoch": 2.89,
|
1826 |
+
"learning_rate": 6.800156256323243e-07,
|
1827 |
+
"loss": 0.7626,
|
1828 |
+
"step": 7450
|
1829 |
+
},
|
1830 |
+
{
|
1831 |
+
"epoch": 2.9,
|
1832 |
+
"learning_rate": 5.6681300729442e-07,
|
1833 |
+
"loss": 0.7744,
|
1834 |
+
"step": 7475
|
1835 |
+
},
|
1836 |
+
{
|
1837 |
+
"epoch": 2.91,
|
1838 |
+
"learning_rate": 4.6388311150644683e-07,
|
1839 |
+
"loss": 0.7768,
|
1840 |
+
"step": 7500
|
1841 |
+
},
|
1842 |
+
{
|
1843 |
+
"epoch": 2.92,
|
1844 |
+
"learning_rate": 3.712365722457922e-07,
|
1845 |
+
"loss": 0.7653,
|
1846 |
+
"step": 7525
|
1847 |
+
},
|
1848 |
+
{
|
1849 |
+
"epoch": 2.93,
|
1850 |
+
"learning_rate": 2.888829610873112e-07,
|
1851 |
+
"loss": 0.7675,
|
1852 |
+
"step": 7550
|
1853 |
+
},
|
1854 |
+
{
|
1855 |
+
"epoch": 2.94,
|
1856 |
+
"learning_rate": 2.168307862144725e-07,
|
1857 |
+
"loss": 0.7727,
|
1858 |
+
"step": 7575
|
1859 |
+
},
|
1860 |
+
{
|
1861 |
+
"epoch": 2.95,
|
1862 |
+
"learning_rate": 1.5508749154037327e-07,
|
1863 |
+
"loss": 0.7712,
|
1864 |
+
"step": 7600
|
1865 |
+
},
|
1866 |
+
{
|
1867 |
+
"epoch": 2.96,
|
1868 |
+
"learning_rate": 1.0365945593864279e-07,
|
1869 |
+
"loss": 0.7695,
|
1870 |
+
"step": 7625
|
1871 |
+
},
|
1872 |
+
{
|
1873 |
+
"epoch": 2.97,
|
1874 |
+
"learning_rate": 6.255199258446975e-08,
|
1875 |
+
"loss": 0.7699,
|
1876 |
+
"step": 7650
|
1877 |
+
},
|
1878 |
+
{
|
1879 |
+
"epoch": 2.98,
|
1880 |
+
"learning_rate": 3.1769348405652486e-08,
|
1881 |
+
"loss": 0.7624,
|
1882 |
+
"step": 7675
|
1883 |
+
},
|
1884 |
+
{
|
1885 |
+
"epoch": 2.99,
|
1886 |
+
"learning_rate": 1.1314703643827695e-08,
|
1887 |
+
"loss": 0.7703,
|
1888 |
+
"step": 7700
|
1889 |
}
|
1890 |
],
|
1891 |
"max_steps": 7737,
|
1892 |
"num_train_epochs": 3,
|
1893 |
+
"total_flos": 3.316121644828721e+19,
|
1894 |
"trial_name": null,
|
1895 |
"trial_params": null
|
1896 |
}
|
{checkpoint-7300 β checkpoint-7700}/training_args.bin
RENAMED
File without changes
|