hugodk-sch commited on
Commit
e33c89a
1 Parent(s): 40d6fc5

End of training

Browse files
Files changed (3) hide show
  1. README.md +13 -11
  2. all_results.json +13 -0
  3. eval_results.json +11 -11
README.md CHANGED
@@ -1,11 +1,13 @@
1
  ---
2
- license: apache-2.0
3
  library_name: peft
4
  tags:
 
5
  - trl
6
  - dpo
7
  - generated_from_trainer
8
  base_model: norallm/normistral-7b-warm
 
 
9
  model-index:
10
  - name: ap-normistral-7b-align-scan
11
  results: []
@@ -16,17 +18,17 @@ should probably proofread and complete it, then remove this comment. -->
16
 
17
  # ap-normistral-7b-align-scan
18
 
19
- This model is a fine-tuned version of [norallm/normistral-7b-warm](https://huggingface.co/norallm/normistral-7b-warm) on the None dataset.
20
  It achieves the following results on the evaluation set:
21
- - Loss: 0.7744
22
- - Rewards/chosen: -0.0855
23
- - Rewards/rejected: -0.2061
24
- - Rewards/accuracies: 0.5428
25
- - Rewards/margins: 0.1206
26
- - Logps/rejected: -36.2242
27
- - Logps/chosen: -32.5501
28
- - Logits/rejected: 98.7517
29
- - Logits/chosen: 98.7722
30
 
31
  ## Model description
32
 
 
1
  ---
 
2
  library_name: peft
3
  tags:
4
+ - alignment-handbook
5
  - trl
6
  - dpo
7
  - generated_from_trainer
8
  base_model: norallm/normistral-7b-warm
9
+ datasets:
10
+ - hugodk-sch/aftonposten_title_prefs
11
  model-index:
12
  - name: ap-normistral-7b-align-scan
13
  results: []
 
18
 
19
  # ap-normistral-7b-align-scan
20
 
21
+ This model is a fine-tuned version of [data/ap-normistral-7b-sft-qlora](https://huggingface.co/data/ap-normistral-7b-sft-qlora) on the hugodk-sch/aftonposten_title_prefs dataset.
22
  It achieves the following results on the evaluation set:
23
+ - Loss: 0.7822
24
+ - Rewards/chosen: -0.1005
25
+ - Rewards/rejected: -0.2047
26
+ - Rewards/accuracies: 0.5249
27
+ - Rewards/margins: 0.1042
28
+ - Logps/rejected: -36.2224
29
+ - Logps/chosen: -32.5688
30
+ - Logits/rejected: 98.7467
31
+ - Logits/chosen: 98.7645
32
 
33
  ## Model description
34
 
all_results.json CHANGED
@@ -1,5 +1,18 @@
1
  {
2
  "epoch": 1.0,
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  "train_loss": 0.6050039254225694,
4
  "train_runtime": 2556.17,
5
  "train_samples": 3079,
 
1
  {
2
  "epoch": 1.0,
3
+ "eval_logits/chosen": 98.7645034790039,
4
+ "eval_logits/rejected": 98.74665832519531,
5
+ "eval_logps/chosen": -32.56877517700195,
6
+ "eval_logps/rejected": -36.22236251831055,
7
+ "eval_loss": 0.7821906208992004,
8
+ "eval_rewards/accuracies": 0.52491694688797,
9
+ "eval_rewards/chosen": -0.10047715157270432,
10
+ "eval_rewards/margins": 0.1041894406080246,
11
+ "eval_rewards/rejected": -0.20466658473014832,
12
+ "eval_runtime": 103.6205,
13
+ "eval_samples": 343,
14
+ "eval_samples_per_second": 3.31,
15
+ "eval_steps_per_second": 0.415,
16
  "train_loss": 0.6050039254225694,
17
  "train_runtime": 2556.17,
18
  "train_samples": 3079,
eval_results.json CHANGED
@@ -1,16 +1,16 @@
1
  {
2
  "epoch": 1.0,
3
- "eval_logits/chosen": 97.9656753540039,
4
- "eval_logits/rejected": 97.93604278564453,
5
- "eval_logps/chosen": -33.003868103027344,
6
- "eval_logps/rejected": -36.95783615112305,
7
- "eval_loss": 0.6761856079101562,
8
- "eval_rewards/accuracies": 0.5888704061508179,
9
- "eval_rewards/chosen": -0.05606912076473236,
10
- "eval_rewards/margins": 0.04306147247552872,
11
- "eval_rewards/rejected": -0.09913058578968048,
12
- "eval_runtime": 103.6924,
13
  "eval_samples": 343,
14
- "eval_samples_per_second": 3.308,
15
  "eval_steps_per_second": 0.415
16
  }
 
1
  {
2
  "epoch": 1.0,
3
+ "eval_logits/chosen": 98.7645034790039,
4
+ "eval_logits/rejected": 98.74665832519531,
5
+ "eval_logps/chosen": -32.56877517700195,
6
+ "eval_logps/rejected": -36.22236251831055,
7
+ "eval_loss": 0.7821906208992004,
8
+ "eval_rewards/accuracies": 0.52491694688797,
9
+ "eval_rewards/chosen": -0.10047715157270432,
10
+ "eval_rewards/margins": 0.1041894406080246,
11
+ "eval_rewards/rejected": -0.20466658473014832,
12
+ "eval_runtime": 103.6205,
13
  "eval_samples": 343,
14
+ "eval_samples_per_second": 3.31,
15
  "eval_steps_per_second": 0.415
16
  }