metadata
library_name: transformers
license: apache-2.0
base_model: tsavage68/IE_M2_1000steps_1e7rate_SFT
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: IE_M2_1000steps_1e8rate_05beta_cSFTDPO
results: []
IE_M2_1000steps_1e8rate_05beta_cSFTDPO
This model is a fine-tuned version of tsavage68/IE_M2_1000steps_1e7rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.5967
- Rewards/chosen: 0.0013
- Rewards/rejected: -0.2212
- Rewards/accuracies: 0.4600
- Rewards/margins: 0.2225
- Logps/rejected: -41.4643
- Logps/chosen: -42.2029
- Logits/rejected: -2.9153
- Logits/chosen: -2.8540
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: 1e-08
- train_batch_size: 2
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 100
- training_steps: 1000
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6945 | 0.4 | 50 | 0.6933 | 0.0121 | 0.0098 | 0.2450 | 0.0023 | -41.0022 | -42.1813 | -2.9159 | -2.8545 |
| 0.6936 | 0.8 | 100 | 0.6888 | 0.0052 | -0.0069 | 0.2150 | 0.0121 | -41.0356 | -42.1952 | -2.9158 | -2.8545 |
| 0.6628 | 1.2 | 150 | 0.6642 | 0.0025 | -0.0598 | 0.3650 | 0.0623 | -41.1414 | -42.2005 | -2.9158 | -2.8545 |
| 0.6553 | 1.6 | 200 | 0.6439 | -0.0046 | -0.1128 | 0.4350 | 0.1083 | -41.2475 | -42.2147 | -2.9156 | -2.8543 |
| 0.6399 | 2.0 | 250 | 0.6211 | -0.0017 | -0.1629 | 0.4600 | 0.1612 | -41.3475 | -42.2089 | -2.9153 | -2.8541 |
| 0.622 | 2.4 | 300 | 0.6110 | -0.0080 | -0.1940 | 0.4600 | 0.1859 | -41.4097 | -42.2216 | -2.9155 | -2.8542 |
| 0.6063 | 2.8 | 350 | 0.6052 | -0.0027 | -0.2028 | 0.4550 | 0.2001 | -41.4274 | -42.2109 | -2.9153 | -2.8540 |
| 0.6243 | 3.2 | 400 | 0.6005 | -0.0031 | -0.2152 | 0.4600 | 0.2121 | -41.4523 | -42.2118 | -2.9154 | -2.8541 |
| 0.6262 | 3.6 | 450 | 0.6019 | -0.0015 | -0.2107 | 0.4600 | 0.2092 | -41.4433 | -42.2085 | -2.9154 | -2.8541 |
| 0.6281 | 4.0 | 500 | 0.5955 | -0.0090 | -0.2341 | 0.4600 | 0.2251 | -41.4900 | -42.2236 | -2.9151 | -2.8538 |
| 0.5897 | 4.4 | 550 | 0.5966 | -0.0012 | -0.2244 | 0.4600 | 0.2232 | -41.4706 | -42.2079 | -2.9151 | -2.8538 |
| 0.5987 | 4.8 | 600 | 0.5991 | -0.0063 | -0.2220 | 0.4600 | 0.2157 | -41.4659 | -42.2182 | -2.9153 | -2.8541 |
| 0.6188 | 5.2 | 650 | 0.5972 | -0.0053 | -0.2271 | 0.4550 | 0.2218 | -41.4759 | -42.2160 | -2.9154 | -2.8541 |
| 0.6165 | 5.6 | 700 | 0.6060 | -0.0050 | -0.2031 | 0.4550 | 0.1981 | -41.4281 | -42.2156 | -2.9153 | -2.8540 |
| 0.5861 | 6.0 | 750 | 0.6007 | -0.0033 | -0.2152 | 0.4600 | 0.2119 | -41.4523 | -42.2121 | -2.9154 | -2.8540 |
| 0.5445 | 6.4 | 800 | 0.5984 | -0.0069 | -0.2252 | 0.4600 | 0.2183 | -41.4722 | -42.2193 | -2.9153 | -2.8539 |
| 0.6228 | 6.8 | 850 | 0.5987 | -0.0027 | -0.2205 | 0.4600 | 0.2178 | -41.4628 | -42.2110 | -2.9153 | -2.8539 |
| 0.5741 | 7.2 | 900 | 0.5967 | 0.0013 | -0.2212 | 0.4600 | 0.2225 | -41.4643 | -42.2029 | -2.9153 | -2.8540 |
| 0.5819 | 7.6 | 950 | 0.5967 | 0.0013 | -0.2212 | 0.4600 | 0.2225 | -41.4643 | -42.2029 | -2.9153 | -2.8540 |
| 0.607 | 8.0 | 1000 | 0.5967 | 0.0013 | -0.2212 | 0.4600 | 0.2225 | -41.4643 | -42.2029 | -2.9153 | -2.8540 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.0.0+cu117
- Datasets 3.0.0
- Tokenizers 0.19.1