Hitisha commited on
Commit
3cdb100
·
verified ·
1 Parent(s): b21272b

Model save

Browse files
Files changed (1) hide show
  1. README.md +10 -10
README.md CHANGED
@@ -14,21 +14,21 @@ model-index:
14
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
15
  should probably proofread and complete it, then remove this comment. -->
16
 
17
- [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/algo-llm/huggingface/runs/8cgkk8bv)
18
  # orpo-phi3
19
 
20
  This model is a fine-tuned version of [microsoft/Phi-3-mini-4k-instruct](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct) on the None dataset.
21
  It achieves the following results on the evaluation set:
22
- - Loss: 3.9381
23
- - Rewards/chosen: -0.4248
24
- - Rewards/rejected: -0.4248
25
  - Rewards/accuracies: 0.0
26
  - Rewards/margins: 0.0
27
- - Logps/rejected: -4.2478
28
- - Logps/chosen: -4.2478
29
- - Logits/rejected: 18.4407
30
- - Logits/chosen: 18.4407
31
- - Nll Loss: 3.8688
32
  - Log Odds Ratio: -0.6931
33
  - Log Odds Chosen: 0.0
34
 
@@ -64,7 +64,7 @@ The following hyperparameters were used during training:
64
 
65
  | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | Nll Loss | Log Odds Ratio | Log Odds Chosen |
66
  |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:--------:|:--------------:|:---------------:|
67
- | 3.0778 | 1.0 | 1 | 3.9381 | -0.4248 | -0.4248 | 0.0 | 0.0 | -4.2478 | -4.2478 | 18.4407 | 18.4407 | 3.8688 | -0.6931 | 0.0 |
68
 
69
 
70
  ### Framework versions
 
14
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
15
  should probably proofread and complete it, then remove this comment. -->
16
 
17
+ [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/algo-llm/huggingface/runs/g44k7xr4)
18
  # orpo-phi3
19
 
20
  This model is a fine-tuned version of [microsoft/Phi-3-mini-4k-instruct](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct) on the None dataset.
21
  It achieves the following results on the evaluation set:
22
+ - Loss: 5.4687
23
+ - Rewards/chosen: -0.5200
24
+ - Rewards/rejected: -0.5200
25
  - Rewards/accuracies: 0.0
26
  - Rewards/margins: 0.0
27
+ - Logps/rejected: -5.2002
28
+ - Logps/chosen: -5.2002
29
+ - Logits/rejected: 23.0851
30
+ - Logits/chosen: 23.0851
31
+ - Nll Loss: 5.3994
32
  - Log Odds Ratio: -0.6931
33
  - Log Odds Chosen: 0.0
34
 
 
64
 
65
  | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | Nll Loss | Log Odds Ratio | Log Odds Chosen |
66
  |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:--------:|:--------------:|:---------------:|
67
+ | 2.9122 | 1.0 | 1 | 5.4687 | -0.5200 | -0.5200 | 0.0 | 0.0 | -5.2002 | -5.2002 | 23.0851 | 23.0851 | 5.3994 | -0.6931 | 0.0 |
68
 
69
 
70
  ### Framework versions