End of training
Browse files- README.md +10 -9
- all_results.json +13 -0
- eval_results.json +11 -11
README.md
CHANGED
@@ -2,6 +2,7 @@
|
|
2 |
base_model: princeton-nlp/Llama-3-Base-8B-SFT
|
3 |
library_name: peft
|
4 |
tags:
|
|
|
5 |
- trl
|
6 |
- dpo
|
7 |
- generated_from_trainer
|
@@ -17,15 +18,15 @@ should probably proofread and complete it, then remove this comment. -->
|
|
17 |
|
18 |
This model is a fine-tuned version of [princeton-nlp/Llama-3-Base-8B-SFT](https://huggingface.co/princeton-nlp/Llama-3-Base-8B-SFT) on the None dataset.
|
19 |
It achieves the following results on the evaluation set:
|
20 |
-
- Loss: 0.
|
21 |
-
- Rewards/chosen: 0.
|
22 |
-
- Rewards/rejected: -0.
|
23 |
-
- Rewards/accuracies: 0.
|
24 |
-
- Rewards/margins: 0.
|
25 |
-
- Logps/rejected: -284.
|
26 |
-
- Logps/chosen: -292.
|
27 |
-
- Logits/rejected: -0.
|
28 |
-
- Logits/chosen: -0.
|
29 |
|
30 |
## Model description
|
31 |
|
|
|
2 |
base_model: princeton-nlp/Llama-3-Base-8B-SFT
|
3 |
library_name: peft
|
4 |
tags:
|
5 |
+
- alignment-handbook
|
6 |
- trl
|
7 |
- dpo
|
8 |
- generated_from_trainer
|
|
|
18 |
|
19 |
This model is a fine-tuned version of [princeton-nlp/Llama-3-Base-8B-SFT](https://huggingface.co/princeton-nlp/Llama-3-Base-8B-SFT) on the None dataset.
|
20 |
It achieves the following results on the evaluation set:
|
21 |
+
- Loss: 0.5193
|
22 |
+
- Rewards/chosen: 0.0154
|
23 |
+
- Rewards/rejected: -0.7979
|
24 |
+
- Rewards/accuracies: 0.7280
|
25 |
+
- Rewards/margins: 0.8133
|
26 |
+
- Logps/rejected: -284.6558
|
27 |
+
- Logps/chosen: -292.3936
|
28 |
+
- Logits/rejected: -0.3843
|
29 |
+
- Logits/chosen: -0.4157
|
30 |
|
31 |
## Model description
|
32 |
|
all_results.json
CHANGED
@@ -1,5 +1,18 @@
|
|
1 |
{
|
2 |
"epoch": 0.9997382884061764,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
"total_flos": 0.0,
|
4 |
"train_loss": 0.5431846207973221,
|
5 |
"train_runtime": 19104.9346,
|
|
|
1 |
{
|
2 |
"epoch": 0.9997382884061764,
|
3 |
+
"eval_logits/chosen": -0.41565001010894775,
|
4 |
+
"eval_logits/rejected": -0.3842959403991699,
|
5 |
+
"eval_logps/chosen": -292.3936462402344,
|
6 |
+
"eval_logps/rejected": -284.65582275390625,
|
7 |
+
"eval_loss": 0.5193021297454834,
|
8 |
+
"eval_rewards/accuracies": 0.7279999852180481,
|
9 |
+
"eval_rewards/chosen": 0.015352733433246613,
|
10 |
+
"eval_rewards/margins": 0.813278079032898,
|
11 |
+
"eval_rewards/rejected": -0.7979253530502319,
|
12 |
+
"eval_runtime": 347.9856,
|
13 |
+
"eval_samples": 2000,
|
14 |
+
"eval_samples_per_second": 5.747,
|
15 |
+
"eval_steps_per_second": 0.359,
|
16 |
"total_flos": 0.0,
|
17 |
"train_loss": 0.5431846207973221,
|
18 |
"train_runtime": 19104.9346,
|
eval_results.json
CHANGED
@@ -1,16 +1,16 @@
|
|
1 |
{
|
2 |
"epoch": 0.9997382884061764,
|
3 |
-
"eval_logits/chosen": -0.
|
4 |
-
"eval_logits/rejected": -0.
|
5 |
-
"eval_logps/chosen": -
|
6 |
-
"eval_logps/rejected": -
|
7 |
-
"eval_loss": 0.
|
8 |
-
"eval_rewards/accuracies": 0.
|
9 |
-
"eval_rewards/chosen":
|
10 |
-
"eval_rewards/margins": 0.
|
11 |
-
"eval_rewards/rejected": -0.
|
12 |
-
"eval_runtime":
|
13 |
"eval_samples": 2000,
|
14 |
-
"eval_samples_per_second": 5.
|
15 |
"eval_steps_per_second": 0.359
|
16 |
}
|
|
|
1 |
{
|
2 |
"epoch": 0.9997382884061764,
|
3 |
+
"eval_logits/chosen": -0.41565001010894775,
|
4 |
+
"eval_logits/rejected": -0.3842959403991699,
|
5 |
+
"eval_logps/chosen": -292.3936462402344,
|
6 |
+
"eval_logps/rejected": -284.65582275390625,
|
7 |
+
"eval_loss": 0.5193021297454834,
|
8 |
+
"eval_rewards/accuracies": 0.7279999852180481,
|
9 |
+
"eval_rewards/chosen": 0.015352733433246613,
|
10 |
+
"eval_rewards/margins": 0.813278079032898,
|
11 |
+
"eval_rewards/rejected": -0.7979253530502319,
|
12 |
+
"eval_runtime": 347.9856,
|
13 |
"eval_samples": 2000,
|
14 |
+
"eval_samples_per_second": 5.747,
|
15 |
"eval_steps_per_second": 0.359
|
16 |
}
|