Upload checkpoint at step 0
Browse files
000000000000/metadata.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"git_commit": "
|
| 3 |
"config": {
|
| 4 |
"checkpoint": {
|
| 5 |
"source": "aklein4/forte-init-square",
|
|
@@ -8,7 +8,7 @@
|
|
| 8 |
"save": true,
|
| 9 |
"interval": 50,
|
| 10 |
"save_trainer": false,
|
| 11 |
-
"dir": "local_data/runs/forte-square_2026-10-
|
| 12 |
"huggingface": {
|
| 13 |
"enabled": true,
|
| 14 |
"private": false,
|
|
@@ -19,7 +19,7 @@
|
|
| 19 |
"initialize": "auto",
|
| 20 |
"fsdp_axis_size": -1
|
| 21 |
},
|
| 22 |
-
"dir": "local_data/runs/forte-square_2026-10-
|
| 23 |
"seed": 42,
|
| 24 |
"global_batch_size": 128,
|
| 25 |
"max_steps": 1000000,
|
|
@@ -74,7 +74,7 @@
|
|
| 74 |
"attention_backend": "cudnn",
|
| 75 |
"fast_weight_size": 1024,
|
| 76 |
"update_rank": 512,
|
| 77 |
-
"base_lr": 0.
|
| 78 |
"pool_size": 512,
|
| 79 |
"grad_rms_eps": 1e-12
|
| 80 |
},
|
|
|
|
| 1 |
{
|
| 2 |
+
"git_commit": "67086da943ffa16504bf8885b0f382fdc62c0f9a",
|
| 3 |
"config": {
|
| 4 |
"checkpoint": {
|
| 5 |
"source": "aklein4/forte-init-square",
|
|
|
|
| 8 |
"save": true,
|
| 9 |
"interval": 50,
|
| 10 |
"save_trainer": false,
|
| 11 |
+
"dir": "local_data/runs/forte-square_2026-10-23_12-33-05/checkpoints",
|
| 12 |
"huggingface": {
|
| 13 |
"enabled": true,
|
| 14 |
"private": false,
|
|
|
|
| 19 |
"initialize": "auto",
|
| 20 |
"fsdp_axis_size": -1
|
| 21 |
},
|
| 22 |
+
"dir": "local_data/runs/forte-square_2026-10-23_12-33-05",
|
| 23 |
"seed": 42,
|
| 24 |
"global_batch_size": 128,
|
| 25 |
"max_steps": 1000000,
|
|
|
|
| 74 |
"attention_backend": "cudnn",
|
| 75 |
"fast_weight_size": 1024,
|
| 76 |
"update_rank": 512,
|
| 77 |
+
"base_lr": 0.0003,
|
| 78 |
"pool_size": 512,
|
| 79 |
"grad_rms_eps": 1e-12
|
| 80 |
},
|
000000000000/model/model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7034679552
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:53b19b77e811ff6692e4dd76331c4b792b3c95ba00c36b88524b626de6994f4b
|
| 3 |
size 7034679552
|