Training in progress, step 200
Browse files- adapter_config.json +6 -1
- adapter_model.safetensors +2 -2
- angle.config +11 -2
- runs/Apr06_09-11-54_instance-20240226-074643/events.out.tfevents.1712394722.instance-20240226-074643.349954.0 +2 -2
- runs/Apr06_09-20-44_instance-20240226-074643/events.out.tfevents.1712395249.instance-20240226-074643.362017.0 +3 -0
- runs/Apr06_09-27-57_instance-20240226-074643/events.out.tfevents.1712395682.instance-20240226-074643.363086.0 +3 -0
- runs/Apr06_09-35-51_instance-20240226-074643/events.out.tfevents.1712396158.instance-20240226-074643.364834.0 +3 -0
- training_args.bin +1 -1
adapter_config.json
CHANGED
@@ -19,8 +19,13 @@
|
|
19 |
"rank_pattern": {},
|
20 |
"revision": null,
|
21 |
"target_modules": [
|
|
|
|
|
|
|
|
|
|
|
22 |
"q_proj",
|
23 |
-
"
|
24 |
],
|
25 |
"task_type": "CAUSAL_LM",
|
26 |
"use_rslora": false
|
|
|
19 |
"rank_pattern": {},
|
20 |
"revision": null,
|
21 |
"target_modules": [
|
22 |
+
"v_proj",
|
23 |
+
"o_proj",
|
24 |
+
"up_proj",
|
25 |
+
"gate_proj",
|
26 |
+
"down_proj",
|
27 |
"q_proj",
|
28 |
+
"k_proj"
|
29 |
],
|
30 |
"task_type": "CAUSAL_LM",
|
31 |
"use_rslora": false
|
adapter_model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:44e813bc3106b1f510e4957533ab10fe2446ad24829ff7b0cfbff3ac00c99d77
|
3 |
+
size 639691872
|
angle.config
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
{
|
2 |
"model_name_or_path": "NousResearch/Llama-2-7b-hf",
|
3 |
-
"max_length":
|
4 |
"model_kwargs": {},
|
5 |
"pooling_strategy": "cls",
|
6 |
"lora_config_kwargs": {
|
@@ -8,7 +8,16 @@
|
|
8 |
"r": 64,
|
9 |
"lora_alpha": 8,
|
10 |
"lora_dropout": 0.05,
|
11 |
-
"bias": "none"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
},
|
13 |
"apply_lora": 1
|
14 |
}
|
|
|
1 |
{
|
2 |
"model_name_or_path": "NousResearch/Llama-2-7b-hf",
|
3 |
+
"max_length": 75,
|
4 |
"model_kwargs": {},
|
5 |
"pooling_strategy": "cls",
|
6 |
"lora_config_kwargs": {
|
|
|
8 |
"r": 64,
|
9 |
"lora_alpha": 8,
|
10 |
"lora_dropout": 0.05,
|
11 |
+
"bias": "none",
|
12 |
+
"target_modules": [
|
13 |
+
"gate_proj",
|
14 |
+
"up_proj",
|
15 |
+
"v_proj",
|
16 |
+
"o_proj",
|
17 |
+
"down_proj",
|
18 |
+
"q_proj",
|
19 |
+
"k_proj"
|
20 |
+
]
|
21 |
},
|
22 |
"apply_lora": 1
|
23 |
}
|
runs/Apr06_09-11-54_instance-20240226-074643/events.out.tfevents.1712394722.instance-20240226-074643.349954.0
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:101987e82b8feb1362adc07bc733dbc3b25099164517810a435a572c3425282e
|
3 |
+
size 7926
|
runs/Apr06_09-20-44_instance-20240226-074643/events.out.tfevents.1712395249.instance-20240226-074643.362017.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:60fdf3d2ce9f95a20c498672ff2d308b1de60e13cdda3df4ae0dac0adbbc0d75
|
3 |
+
size 5207
|
runs/Apr06_09-27-57_instance-20240226-074643/events.out.tfevents.1712395682.instance-20240226-074643.363086.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8b5bff77647e91bcd32191cc3315aa7e01fde6fee4c2339bead6ccdcee847a8e
|
3 |
+
size 5207
|
runs/Apr06_09-35-51_instance-20240226-074643/events.out.tfevents.1712396158.instance-20240226-074643.364834.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:09efe21af810556e97e7b141c7bb6ee1228bc9494fed81eb2c12518ae9935031
|
3 |
+
size 5625
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 5048
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:527de58fca9187fcb166590e417572931c936bc44819563193463795a8aa5c28
|
3 |
size 5048
|