Add files using upload-large-folder tool
Browse filesThis view is limited to 50 files because it contains too many changes. See raw diff
- 2b_l10/outputs_add_free/all/lang/187/config.json +12 -0
- 2b_l10/outputs_add_free/all/lang/387/config.json +12 -0
- 2b_l10/outputs_add_free/all/lang/441/config.json +12 -0
- 2b_l10/outputs_add_free/all/lang/450/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/0/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/10/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/100/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/104/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/105/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/106/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/109/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/110/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/112/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/113/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/114/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/119/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/12/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/123/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/126/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/130/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/133/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/134/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/135/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/136/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/138/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/140/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/143/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/150/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/153/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/156/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/159/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/162/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/163/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/166/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/167/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/168/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/171/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/172/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/173/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/177/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/178/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/179/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/180/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/181/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/182/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/184/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/186/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/188/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/189/config.json +12 -0
- 2b_l10/outputs_add_free/all/simpo/19/config.json +12 -0
2b_l10/outputs_add_free/all/lang/187/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/lang/387/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/lang/441/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/lang/450/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/0/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/10/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/100/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/104/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/105/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/106/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/109/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/110/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/112/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/113/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/114/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/119/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/12/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/123/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/126/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/130/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/133/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/134/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/135/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/136/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/138/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/140/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/143/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/150/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/153/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/156/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/159/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/162/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/163/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/166/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/167/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/168/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/171/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/172/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/173/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/177/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/178/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/179/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/180/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/181/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/182/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/184/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/186/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/188/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/189/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|
2b_l10/outputs_add_free/all/simpo/19/config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"layer": 10,
|
| 4 |
+
"embed_dim": 2304,
|
| 5 |
+
"low_rank_dim": 1,
|
| 6 |
+
"target_module": "model.layers.10",
|
| 7 |
+
"intervention_type": "AdditionFreeIntervention",
|
| 8 |
+
"factor_init_scale": 2.0,
|
| 9 |
+
"vector_init_scale": 1.0,
|
| 10 |
+
"alpha": 0.0
|
| 11 |
+
}
|
| 12 |
+
]
|