Downloads · 30 days
0
llama-7b-SFT-qlora-eli5-wiki_DPO_ds_RM_random_1024_r_64_alpha_16
dhmeltzer/llama-7b-SFT-qlora-eli5-wiki_DPO_ds_RM_random_1024_r_64_alpha_16
llama-7b-SFT-qlora-eli5-wiki_DPO_ds_RM_random_1024_r_64_alpha_16 is a machine learning model from dhmeltzer. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
should probably proofread and complete it, then remove this comment. --
Access
Public
Updated Aug 30, 2023
Repo size
20.5 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.bin640 MB · 100%
At a glance
- Access
- Public
- Created
- Aug 30, 2023
- Updated
- Aug 30, 2023
- SHA
- eb00ac37
Base models
Model card
From the Hugging Face model README
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
llama-7b-SFT-qlora-eli5-wiki_DPO_ds_RM_random_1024_r_64_alpha_16
This model is a fine-tuned version of dhmeltzer/llama-7b-SFT_eli5_wiki65k_1024_r_64_alpha_16_merged on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.6788
- Rewards/chosen: -0.0760
- Rewards/rejected: -0.1428
- Rewards/accuracies: 0.5781
- Rewards/margins: 0.0669
- Logps/rejected: -202.0682
- Logps/chosen: -199.2469
- Logits/rejected: 1.0323
- Logits/chosen: 1.0541
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0002
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.03
- num_epochs: 1
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6913 | 0.1 | 19 | 0.6845 | -0.4006 | -0.4672 | 0.5558 | 0.0665 | -205.3114 | -202.4936 | 1.0265 | 1.0467 |
| 0.6768 | 0.21 | 38 | 0.6796 | -0.3409 | -0.4196 | 0.5603 | 0.0787 | -204.8360 | -201.8965 | 1.0326 | 1.0538 |
| 0.6771 | 0.31 | 57 | 0.6788 | -0.0760 | -0.1428 | 0.5781 | 0.0669 | -202.0682 | -199.2469 | 1.0323 | 1.0541 |
| 0.6665 | 0.41 | 76 | 0.6826 | -0.1511 | -0.2355 | 0.5703 | 0.0843 | -202.9944 | -199.9986 | 1.0413 | 1.0635 |
| 0.6669 | 0.52 | 95 | 0.6830 | -0.1285 | -0.2165 | 0.5781 | 0.0880 | -202.8050 | -199.7720 | 1.0299 | 1.0522 |
| 0.669 | 0.62 | 114 | 0.6800 | -0.0932 | -0.1803 | 0.5725 | 0.0871 | -202.4429 | -199.4187 | 1.0126 | 1.0352 |
| 0.6559 | 0.72 | 133 | 0.6829 | -0.0011 | -0.1074 | 0.5759 | 0.1063 | -201.7135 | -198.4980 | 1.0015 | 1.0232 |
| 0.6698 | 0.83 | 152 | 0.6810 | -0.0519 | -0.1530 | 0.5781 | 0.1011 | -202.1696 | -199.0062 | 0.9974 | 1.0192 |
| 0.6643 | 0.93 | 171 | 0.6799 | -0.0579 | -0.1589 | 0.5658 | 0.1010 | -202.2284 | -199.0658 | 1.0002 | 1.0220 |
Framework versions
- Transformers 4.32.1
- Pytorch 2.0.1+cu118
- Datasets 2.14.4
- Tokenizers 0.13.3
- Created
- Aug 30, 2023
- Updated
- Aug 30, 2023