Model save
Browse files- README.md +75 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +509 -0
README.md
ADDED
|
@@ -0,0 +1,75 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- dpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-7b-align-scan-0.0-0.9-polynomial-3
|
| 11 |
+
results: []
|
| 12 |
+
---
|
| 13 |
+
|
| 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 |
+
# zephyr-7b-align-scan-0.0-0.9-polynomial-3
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on an unknown dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.9404
|
| 22 |
+
- Rewards/chosen: 1.5621
|
| 23 |
+
- Rewards/rejected: 0.3036
|
| 24 |
+
- Rewards/accuracies: 0.3294
|
| 25 |
+
- Rewards/margins: 1.2585
|
| 26 |
+
- Logps/rejected: -80.8042
|
| 27 |
+
- Logps/chosen: -72.8236
|
| 28 |
+
- Logits/rejected: -2.5182
|
| 29 |
+
- Logits/chosen: -2.5344
|
| 30 |
+
|
| 31 |
+
## Model description
|
| 32 |
+
|
| 33 |
+
More information needed
|
| 34 |
+
|
| 35 |
+
## Intended uses & limitations
|
| 36 |
+
|
| 37 |
+
More information needed
|
| 38 |
+
|
| 39 |
+
## Training and evaluation data
|
| 40 |
+
|
| 41 |
+
More information needed
|
| 42 |
+
|
| 43 |
+
## Training procedure
|
| 44 |
+
|
| 45 |
+
### Training hyperparameters
|
| 46 |
+
|
| 47 |
+
The following hyperparameters were used during training:
|
| 48 |
+
- learning_rate: 4.135505422578228e-07
|
| 49 |
+
- train_batch_size: 8
|
| 50 |
+
- eval_batch_size: 8
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 4
|
| 54 |
+
- gradient_accumulation_steps: 2
|
| 55 |
+
- total_train_batch_size: 64
|
| 56 |
+
- total_eval_batch_size: 32
|
| 57 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
+
- lr_scheduler_type: polynomial
|
| 59 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
+
- num_epochs: 3
|
| 61 |
+
|
| 62 |
+
### Training results
|
| 63 |
+
|
| 64 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 65 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 66 |
+
| 0.5697 | 1.0417 | 100 | 0.7314 | 0.9086 | 0.2746 | 0.3433 | 0.6340 | -80.8352 | -73.5212 | -2.5080 | -2.5235 |
|
| 67 |
+
| 0.3568 | 2.0833 | 200 | 0.9404 | 1.5621 | 0.3036 | 0.3294 | 1.2585 | -80.8042 | -72.8236 | -2.5182 | -2.5344 |
|
| 68 |
+
|
| 69 |
+
|
| 70 |
+
### Framework versions
|
| 71 |
+
|
| 72 |
+
- Transformers 4.44.2
|
| 73 |
+
- Pytorch 2.4.0
|
| 74 |
+
- Datasets 2.21.0
|
| 75 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.479405563738611,
|
| 5 |
+
"train_runtime": 3137.8924,
|
| 6 |
+
"train_samples": 6113,
|
| 7 |
+
"train_samples_per_second": 5.844,
|
| 8 |
+
"train_steps_per_second": 0.092
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"transformers_version": "4.44.2"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.479405563738611,
|
| 5 |
+
"train_runtime": 3137.8924,
|
| 6 |
+
"train_samples": 6113,
|
| 7 |
+
"train_samples_per_second": 5.844,
|
| 8 |
+
"train_steps_per_second": 0.092
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,509 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 288,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.010416666666666666,
|
| 13 |
+
"grad_norm": 617.2018982452129,
|
| 14 |
+
"learning_rate": 1.426036352613182e-08,
|
| 15 |
+
"logits/chosen": -2.590585231781006,
|
| 16 |
+
"logits/rejected": -2.5664222240448,
|
| 17 |
+
"logps/chosen": -80.29847717285156,
|
| 18 |
+
"logps/rejected": -53.10200881958008,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.0,
|
| 21 |
+
"rewards/chosen": 0.0,
|
| 22 |
+
"rewards/margins": 0.0,
|
| 23 |
+
"rewards/rejected": 0.0,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.10416666666666667,
|
| 28 |
+
"grad_norm": 527.5551545265153,
|
| 29 |
+
"learning_rate": 1.426036352613182e-07,
|
| 30 |
+
"logits/chosen": -2.5563437938690186,
|
| 31 |
+
"logits/rejected": -2.5385162830352783,
|
| 32 |
+
"logps/chosen": -87.91403198242188,
|
| 33 |
+
"logps/rejected": -81.02542114257812,
|
| 34 |
+
"loss": 0.7003,
|
| 35 |
+
"rewards/accuracies": 0.1805555522441864,
|
| 36 |
+
"rewards/chosen": 0.0010530749568715692,
|
| 37 |
+
"rewards/margins": -0.01120604109019041,
|
| 38 |
+
"rewards/rejected": 0.012259114533662796,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.20833333333333334,
|
| 43 |
+
"grad_norm": 438.4524613822673,
|
| 44 |
+
"learning_rate": 2.852072705226364e-07,
|
| 45 |
+
"logits/chosen": -2.6143155097961426,
|
| 46 |
+
"logits/rejected": -2.5655746459960938,
|
| 47 |
+
"logps/chosen": -103.23042297363281,
|
| 48 |
+
"logps/rejected": -89.8201675415039,
|
| 49 |
+
"loss": 0.6727,
|
| 50 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 51 |
+
"rewards/chosen": 0.11292153596878052,
|
| 52 |
+
"rewards/margins": 0.0527835413813591,
|
| 53 |
+
"rewards/rejected": 0.06013798713684082,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.3125,
|
| 58 |
+
"grad_norm": 605.3404221140555,
|
| 59 |
+
"learning_rate": 4.123399224035455e-07,
|
| 60 |
+
"logits/chosen": -2.516937017440796,
|
| 61 |
+
"logits/rejected": -2.5314974784851074,
|
| 62 |
+
"logps/chosen": -67.30062103271484,
|
| 63 |
+
"logps/rejected": -75.01806640625,
|
| 64 |
+
"loss": 0.6755,
|
| 65 |
+
"rewards/accuracies": 0.33125001192092896,
|
| 66 |
+
"rewards/chosen": 0.12800100445747375,
|
| 67 |
+
"rewards/margins": 0.21436181664466858,
|
| 68 |
+
"rewards/rejected": -0.08636082708835602,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.4166666666666667,
|
| 73 |
+
"grad_norm": 444.59888459571647,
|
| 74 |
+
"learning_rate": 4.002337238607724e-07,
|
| 75 |
+
"logits/chosen": -2.58128023147583,
|
| 76 |
+
"logits/rejected": -2.5694420337677,
|
| 77 |
+
"logps/chosen": -71.9786376953125,
|
| 78 |
+
"logps/rejected": -70.76583099365234,
|
| 79 |
+
"loss": 0.707,
|
| 80 |
+
"rewards/accuracies": 0.25,
|
| 81 |
+
"rewards/chosen": 0.15727010369300842,
|
| 82 |
+
"rewards/margins": 0.20649325847625732,
|
| 83 |
+
"rewards/rejected": -0.049223143607378006,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.5208333333333334,
|
| 88 |
+
"grad_norm": 550.2437260335452,
|
| 89 |
+
"learning_rate": 3.881275253179993e-07,
|
| 90 |
+
"logits/chosen": -2.4956462383270264,
|
| 91 |
+
"logits/rejected": -2.5077638626098633,
|
| 92 |
+
"logps/chosen": -50.16924285888672,
|
| 93 |
+
"logps/rejected": -58.39775466918945,
|
| 94 |
+
"loss": 0.7248,
|
| 95 |
+
"rewards/accuracies": 0.24375000596046448,
|
| 96 |
+
"rewards/chosen": 0.685566782951355,
|
| 97 |
+
"rewards/margins": 0.20654968917369843,
|
| 98 |
+
"rewards/rejected": 0.47901707887649536,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.625,
|
| 103 |
+
"grad_norm": 398.6501309271306,
|
| 104 |
+
"learning_rate": 3.7602132677522625e-07,
|
| 105 |
+
"logits/chosen": -2.5804848670959473,
|
| 106 |
+
"logits/rejected": -2.561837673187256,
|
| 107 |
+
"logps/chosen": -77.69390869140625,
|
| 108 |
+
"logps/rejected": -78.58424377441406,
|
| 109 |
+
"loss": 0.6856,
|
| 110 |
+
"rewards/accuracies": 0.34375,
|
| 111 |
+
"rewards/chosen": 1.2882606983184814,
|
| 112 |
+
"rewards/margins": 0.582649827003479,
|
| 113 |
+
"rewards/rejected": 0.7056108713150024,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.7291666666666666,
|
| 118 |
+
"grad_norm": 522.2478868366107,
|
| 119 |
+
"learning_rate": 3.6391512823245316e-07,
|
| 120 |
+
"logits/chosen": -2.5641486644744873,
|
| 121 |
+
"logits/rejected": -2.5553622245788574,
|
| 122 |
+
"logps/chosen": -94.89775085449219,
|
| 123 |
+
"logps/rejected": -84.7375259399414,
|
| 124 |
+
"loss": 0.6908,
|
| 125 |
+
"rewards/accuracies": 0.35624998807907104,
|
| 126 |
+
"rewards/chosen": 1.3555562496185303,
|
| 127 |
+
"rewards/margins": 0.6880232691764832,
|
| 128 |
+
"rewards/rejected": 0.6675330400466919,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.8333333333333334,
|
| 133 |
+
"grad_norm": 784.7617891372116,
|
| 134 |
+
"learning_rate": 3.5180892968968006e-07,
|
| 135 |
+
"logits/chosen": -2.563314199447632,
|
| 136 |
+
"logits/rejected": -2.5102317333221436,
|
| 137 |
+
"logps/chosen": -84.65424346923828,
|
| 138 |
+
"logps/rejected": -79.29447174072266,
|
| 139 |
+
"loss": 0.6788,
|
| 140 |
+
"rewards/accuracies": 0.40625,
|
| 141 |
+
"rewards/chosen": 0.9282310605049133,
|
| 142 |
+
"rewards/margins": 0.845805823802948,
|
| 143 |
+
"rewards/rejected": 0.08242519199848175,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.9375,
|
| 148 |
+
"grad_norm": 653.0219382894146,
|
| 149 |
+
"learning_rate": 3.39702731146907e-07,
|
| 150 |
+
"logits/chosen": -2.4886748790740967,
|
| 151 |
+
"logits/rejected": -2.4833056926727295,
|
| 152 |
+
"logps/chosen": -53.71883010864258,
|
| 153 |
+
"logps/rejected": -63.658111572265625,
|
| 154 |
+
"loss": 0.7149,
|
| 155 |
+
"rewards/accuracies": 0.2562499940395355,
|
| 156 |
+
"rewards/chosen": 0.24449601769447327,
|
| 157 |
+
"rewards/margins": 0.4224325716495514,
|
| 158 |
+
"rewards/rejected": -0.17793650925159454,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 1.0416666666666667,
|
| 163 |
+
"grad_norm": 211.1512305376009,
|
| 164 |
+
"learning_rate": 3.2759653260413393e-07,
|
| 165 |
+
"logits/chosen": -2.486382246017456,
|
| 166 |
+
"logits/rejected": -2.4644775390625,
|
| 167 |
+
"logps/chosen": -70.5580062866211,
|
| 168 |
+
"logps/rejected": -73.16108703613281,
|
| 169 |
+
"loss": 0.5697,
|
| 170 |
+
"rewards/accuracies": 0.39375001192092896,
|
| 171 |
+
"rewards/chosen": 2.7764008045196533,
|
| 172 |
+
"rewards/margins": 3.8405537605285645,
|
| 173 |
+
"rewards/rejected": -1.0641529560089111,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 1.0416666666666667,
|
| 178 |
+
"eval_logits/chosen": -2.5235462188720703,
|
| 179 |
+
"eval_logits/rejected": -2.507991313934326,
|
| 180 |
+
"eval_logps/chosen": -73.52123260498047,
|
| 181 |
+
"eval_logps/rejected": -80.83517456054688,
|
| 182 |
+
"eval_loss": 0.731433093547821,
|
| 183 |
+
"eval_rewards/accuracies": 0.3432539701461792,
|
| 184 |
+
"eval_rewards/chosen": 0.9086403846740723,
|
| 185 |
+
"eval_rewards/margins": 0.6340094804763794,
|
| 186 |
+
"eval_rewards/rejected": 0.27463099360466003,
|
| 187 |
+
"eval_runtime": 113.6041,
|
| 188 |
+
"eval_samples_per_second": 17.605,
|
| 189 |
+
"eval_steps_per_second": 0.555,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 1.1458333333333333,
|
| 194 |
+
"grad_norm": 53.69152144204709,
|
| 195 |
+
"learning_rate": 3.1549033406136084e-07,
|
| 196 |
+
"logits/chosen": -2.499805212020874,
|
| 197 |
+
"logits/rejected": -2.531052350997925,
|
| 198 |
+
"logps/chosen": -61.31892776489258,
|
| 199 |
+
"logps/rejected": -85.37156677246094,
|
| 200 |
+
"loss": 0.3654,
|
| 201 |
+
"rewards/accuracies": 0.48124998807907104,
|
| 202 |
+
"rewards/chosen": 3.7935471534729004,
|
| 203 |
+
"rewards/margins": 7.435075283050537,
|
| 204 |
+
"rewards/rejected": -3.6415278911590576,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 1.25,
|
| 209 |
+
"grad_norm": 20.68999344570945,
|
| 210 |
+
"learning_rate": 3.0338413551858774e-07,
|
| 211 |
+
"logits/chosen": -2.563140630722046,
|
| 212 |
+
"logits/rejected": -2.531294345855713,
|
| 213 |
+
"logps/chosen": -100.05778503417969,
|
| 214 |
+
"logps/rejected": -99.29850769042969,
|
| 215 |
+
"loss": 0.383,
|
| 216 |
+
"rewards/accuracies": 0.581250011920929,
|
| 217 |
+
"rewards/chosen": 7.445196628570557,
|
| 218 |
+
"rewards/margins": 13.030279159545898,
|
| 219 |
+
"rewards/rejected": -5.585082530975342,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 1.3541666666666667,
|
| 224 |
+
"grad_norm": 14.636781759163808,
|
| 225 |
+
"learning_rate": 2.9127793697581465e-07,
|
| 226 |
+
"logits/chosen": -2.4888994693756104,
|
| 227 |
+
"logits/rejected": -2.491908073425293,
|
| 228 |
+
"logps/chosen": -79.58487701416016,
|
| 229 |
+
"logps/rejected": -93.6611328125,
|
| 230 |
+
"loss": 0.3782,
|
| 231 |
+
"rewards/accuracies": 0.46875,
|
| 232 |
+
"rewards/chosen": 6.539540767669678,
|
| 233 |
+
"rewards/margins": 10.590567588806152,
|
| 234 |
+
"rewards/rejected": -4.051028251647949,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 1.4583333333333333,
|
| 239 |
+
"grad_norm": 244.07787693322098,
|
| 240 |
+
"learning_rate": 2.7917173843304155e-07,
|
| 241 |
+
"logits/chosen": -2.4806408882141113,
|
| 242 |
+
"logits/rejected": -2.495673179626465,
|
| 243 |
+
"logps/chosen": -43.72606658935547,
|
| 244 |
+
"logps/rejected": -66.98627471923828,
|
| 245 |
+
"loss": 0.3801,
|
| 246 |
+
"rewards/accuracies": 0.4375,
|
| 247 |
+
"rewards/chosen": 5.209879398345947,
|
| 248 |
+
"rewards/margins": 6.956570625305176,
|
| 249 |
+
"rewards/rejected": -1.746691346168518,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 1.5625,
|
| 254 |
+
"grad_norm": 6.49560423904862,
|
| 255 |
+
"learning_rate": 2.6706553989026857e-07,
|
| 256 |
+
"logits/chosen": -2.5112829208374023,
|
| 257 |
+
"logits/rejected": -2.486833333969116,
|
| 258 |
+
"logps/chosen": -65.41731262207031,
|
| 259 |
+
"logps/rejected": -75.73360443115234,
|
| 260 |
+
"loss": 0.396,
|
| 261 |
+
"rewards/accuracies": 0.4625000059604645,
|
| 262 |
+
"rewards/chosen": 6.304823875427246,
|
| 263 |
+
"rewards/margins": 8.325952529907227,
|
| 264 |
+
"rewards/rejected": -2.021127939224243,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 1.6666666666666665,
|
| 269 |
+
"grad_norm": 6.063779849774731,
|
| 270 |
+
"learning_rate": 2.5495934134749547e-07,
|
| 271 |
+
"logits/chosen": -2.474635362625122,
|
| 272 |
+
"logits/rejected": -2.482422113418579,
|
| 273 |
+
"logps/chosen": -72.21180725097656,
|
| 274 |
+
"logps/rejected": -92.7371826171875,
|
| 275 |
+
"loss": 0.3721,
|
| 276 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 277 |
+
"rewards/chosen": 8.162205696105957,
|
| 278 |
+
"rewards/margins": 12.658352851867676,
|
| 279 |
+
"rewards/rejected": -4.496148109436035,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 1.7708333333333335,
|
| 284 |
+
"grad_norm": 85.93105910386429,
|
| 285 |
+
"learning_rate": 2.428531428047224e-07,
|
| 286 |
+
"logits/chosen": -2.4428138732910156,
|
| 287 |
+
"logits/rejected": -2.432603359222412,
|
| 288 |
+
"logps/chosen": -69.3802261352539,
|
| 289 |
+
"logps/rejected": -89.16145324707031,
|
| 290 |
+
"loss": 0.3955,
|
| 291 |
+
"rewards/accuracies": 0.53125,
|
| 292 |
+
"rewards/chosen": 5.470473289489746,
|
| 293 |
+
"rewards/margins": 11.255060195922852,
|
| 294 |
+
"rewards/rejected": -5.784585952758789,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 1.875,
|
| 299 |
+
"grad_norm": 98.1345548828152,
|
| 300 |
+
"learning_rate": 2.307469442619493e-07,
|
| 301 |
+
"logits/chosen": -2.473292827606201,
|
| 302 |
+
"logits/rejected": -2.4571070671081543,
|
| 303 |
+
"logps/chosen": -66.85240936279297,
|
| 304 |
+
"logps/rejected": -79.23042297363281,
|
| 305 |
+
"loss": 0.3619,
|
| 306 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 307 |
+
"rewards/chosen": 4.686768531799316,
|
| 308 |
+
"rewards/margins": 9.949495315551758,
|
| 309 |
+
"rewards/rejected": -5.262727737426758,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 1.9791666666666665,
|
| 314 |
+
"grad_norm": 12.502260702545252,
|
| 315 |
+
"learning_rate": 2.186407457191762e-07,
|
| 316 |
+
"logits/chosen": -2.5277750492095947,
|
| 317 |
+
"logits/rejected": -2.5254924297332764,
|
| 318 |
+
"logps/chosen": -81.32090759277344,
|
| 319 |
+
"logps/rejected": -96.72193908691406,
|
| 320 |
+
"loss": 0.3801,
|
| 321 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 322 |
+
"rewards/chosen": 6.689125061035156,
|
| 323 |
+
"rewards/margins": 12.403680801391602,
|
| 324 |
+
"rewards/rejected": -5.714554309844971,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 2.0833333333333335,
|
| 329 |
+
"grad_norm": 0.8365271794834654,
|
| 330 |
+
"learning_rate": 2.065345471764031e-07,
|
| 331 |
+
"logits/chosen": -2.48215651512146,
|
| 332 |
+
"logits/rejected": -2.5054686069488525,
|
| 333 |
+
"logps/chosen": -81.12793731689453,
|
| 334 |
+
"logps/rejected": -109.39041900634766,
|
| 335 |
+
"loss": 0.3568,
|
| 336 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 337 |
+
"rewards/chosen": 7.378162384033203,
|
| 338 |
+
"rewards/margins": 14.7354154586792,
|
| 339 |
+
"rewards/rejected": -7.3572540283203125,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 2.0833333333333335,
|
| 344 |
+
"eval_logits/chosen": -2.534412384033203,
|
| 345 |
+
"eval_logits/rejected": -2.5182077884674072,
|
| 346 |
+
"eval_logps/chosen": -72.82362365722656,
|
| 347 |
+
"eval_logps/rejected": -80.80423736572266,
|
| 348 |
+
"eval_loss": 0.9403718709945679,
|
| 349 |
+
"eval_rewards/accuracies": 0.329365074634552,
|
| 350 |
+
"eval_rewards/chosen": 1.5621305704116821,
|
| 351 |
+
"eval_rewards/margins": 1.2585159540176392,
|
| 352 |
+
"eval_rewards/rejected": 0.3036145269870758,
|
| 353 |
+
"eval_runtime": 113.4863,
|
| 354 |
+
"eval_samples_per_second": 17.623,
|
| 355 |
+
"eval_steps_per_second": 0.555,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 2.1875,
|
| 360 |
+
"grad_norm": 0.17365138232640678,
|
| 361 |
+
"learning_rate": 1.9442834863363e-07,
|
| 362 |
+
"logits/chosen": -2.538646697998047,
|
| 363 |
+
"logits/rejected": -2.531092405319214,
|
| 364 |
+
"logps/chosen": -57.24335861206055,
|
| 365 |
+
"logps/rejected": -73.53451538085938,
|
| 366 |
+
"loss": 0.381,
|
| 367 |
+
"rewards/accuracies": 0.42500001192092896,
|
| 368 |
+
"rewards/chosen": 5.948960781097412,
|
| 369 |
+
"rewards/margins": 10.486295700073242,
|
| 370 |
+
"rewards/rejected": -4.537335395812988,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 2.2916666666666665,
|
| 375 |
+
"grad_norm": 2.449230292077233,
|
| 376 |
+
"learning_rate": 1.8232215009085696e-07,
|
| 377 |
+
"logits/chosen": -2.558448314666748,
|
| 378 |
+
"logits/rejected": -2.511837959289551,
|
| 379 |
+
"logps/chosen": -92.26642608642578,
|
| 380 |
+
"logps/rejected": -99.34980010986328,
|
| 381 |
+
"loss": 0.3282,
|
| 382 |
+
"rewards/accuracies": 0.574999988079071,
|
| 383 |
+
"rewards/chosen": 9.164371490478516,
|
| 384 |
+
"rewards/margins": 15.474553108215332,
|
| 385 |
+
"rewards/rejected": -6.310182094573975,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 2.3958333333333335,
|
| 390 |
+
"grad_norm": 0.8917188204265395,
|
| 391 |
+
"learning_rate": 1.7021595154808387e-07,
|
| 392 |
+
"logits/chosen": -2.5435986518859863,
|
| 393 |
+
"logits/rejected": -2.512115001678467,
|
| 394 |
+
"logps/chosen": -54.11479949951172,
|
| 395 |
+
"logps/rejected": -59.78227615356445,
|
| 396 |
+
"loss": 0.3796,
|
| 397 |
+
"rewards/accuracies": 0.4312500059604645,
|
| 398 |
+
"rewards/chosen": 5.355578422546387,
|
| 399 |
+
"rewards/margins": 8.94926643371582,
|
| 400 |
+
"rewards/rejected": -3.593688488006592,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 2.5,
|
| 405 |
+
"grad_norm": 164.121389909511,
|
| 406 |
+
"learning_rate": 1.5810975300531077e-07,
|
| 407 |
+
"logits/chosen": -2.5293872356414795,
|
| 408 |
+
"logits/rejected": -2.483024835586548,
|
| 409 |
+
"logps/chosen": -81.53751373291016,
|
| 410 |
+
"logps/rejected": -85.67427062988281,
|
| 411 |
+
"loss": 0.3814,
|
| 412 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 413 |
+
"rewards/chosen": 5.955917835235596,
|
| 414 |
+
"rewards/margins": 11.81001091003418,
|
| 415 |
+
"rewards/rejected": -5.8540940284729,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 2.6041666666666665,
|
| 420 |
+
"grad_norm": 0.25571279122418833,
|
| 421 |
+
"learning_rate": 1.4600355446253773e-07,
|
| 422 |
+
"logits/chosen": -2.485600233078003,
|
| 423 |
+
"logits/rejected": -2.473313570022583,
|
| 424 |
+
"logps/chosen": -66.71253204345703,
|
| 425 |
+
"logps/rejected": -76.95586395263672,
|
| 426 |
+
"loss": 0.3582,
|
| 427 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 428 |
+
"rewards/chosen": 5.579711437225342,
|
| 429 |
+
"rewards/margins": 11.685328483581543,
|
| 430 |
+
"rewards/rejected": -6.105617523193359,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 2.7083333333333335,
|
| 435 |
+
"grad_norm": 20.873017456890867,
|
| 436 |
+
"learning_rate": 1.3389735591976464e-07,
|
| 437 |
+
"logits/chosen": -2.487583637237549,
|
| 438 |
+
"logits/rejected": -2.5043182373046875,
|
| 439 |
+
"logps/chosen": -51.980262756347656,
|
| 440 |
+
"logps/rejected": -80.04218292236328,
|
| 441 |
+
"loss": 0.3794,
|
| 442 |
+
"rewards/accuracies": 0.45625001192092896,
|
| 443 |
+
"rewards/chosen": 5.761561393737793,
|
| 444 |
+
"rewards/margins": 12.39246654510498,
|
| 445 |
+
"rewards/rejected": -6.6309051513671875,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 2.8125,
|
| 450 |
+
"grad_norm": 39.42095791773263,
|
| 451 |
+
"learning_rate": 1.2179115737699154e-07,
|
| 452 |
+
"logits/chosen": -2.4937992095947266,
|
| 453 |
+
"logits/rejected": -2.5243396759033203,
|
| 454 |
+
"logps/chosen": -71.54154205322266,
|
| 455 |
+
"logps/rejected": -85.11618041992188,
|
| 456 |
+
"loss": 0.352,
|
| 457 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 458 |
+
"rewards/chosen": 4.807980537414551,
|
| 459 |
+
"rewards/margins": 11.203904151916504,
|
| 460 |
+
"rewards/rejected": -6.395923614501953,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 2.9166666666666665,
|
| 465 |
+
"grad_norm": 0.222552372589121,
|
| 466 |
+
"learning_rate": 1.0968495883421846e-07,
|
| 467 |
+
"logits/chosen": -2.5492477416992188,
|
| 468 |
+
"logits/rejected": -2.515148639678955,
|
| 469 |
+
"logps/chosen": -78.6363525390625,
|
| 470 |
+
"logps/rejected": -90.13154602050781,
|
| 471 |
+
"loss": 0.3626,
|
| 472 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 473 |
+
"rewards/chosen": 6.29600191116333,
|
| 474 |
+
"rewards/margins": 13.563056945800781,
|
| 475 |
+
"rewards/rejected": -7.267056465148926,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 3.0,
|
| 480 |
+
"step": 288,
|
| 481 |
+
"total_flos": 0.0,
|
| 482 |
+
"train_loss": 0.479405563738611,
|
| 483 |
+
"train_runtime": 3137.8924,
|
| 484 |
+
"train_samples_per_second": 5.844,
|
| 485 |
+
"train_steps_per_second": 0.092
|
| 486 |
+
}
|
| 487 |
+
],
|
| 488 |
+
"logging_steps": 10,
|
| 489 |
+
"max_steps": 288,
|
| 490 |
+
"num_input_tokens_seen": 0,
|
| 491 |
+
"num_train_epochs": 3,
|
| 492 |
+
"save_steps": 100,
|
| 493 |
+
"stateful_callbacks": {
|
| 494 |
+
"TrainerControl": {
|
| 495 |
+
"args": {
|
| 496 |
+
"should_epoch_stop": false,
|
| 497 |
+
"should_evaluate": false,
|
| 498 |
+
"should_log": false,
|
| 499 |
+
"should_save": true,
|
| 500 |
+
"should_training_stop": true
|
| 501 |
+
},
|
| 502 |
+
"attributes": {}
|
| 503 |
+
}
|
| 504 |
+
},
|
| 505 |
+
"total_flos": 0.0,
|
| 506 |
+
"train_batch_size": 8,
|
| 507 |
+
"trial_name": null,
|
| 508 |
+
"trial_params": null
|
| 509 |
+
}
|