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-0.0-0.0-linear-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-0.0-0.0-linear-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: 0.6495
|
| 22 |
+
- Rewards/chosen: -0.1460
|
| 23 |
+
- Rewards/rejected: -0.2967
|
| 24 |
+
- Rewards/accuracies: 0.3294
|
| 25 |
+
- Rewards/margins: 0.1507
|
| 26 |
+
- Logps/rejected: -102.2009
|
| 27 |
+
- Logps/chosen: -84.8621
|
| 28 |
+
- Logits/rejected: -2.3195
|
| 29 |
+
- Logits/chosen: -2.3414
|
| 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: 2.8387779271947853e-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: linear
|
| 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 |
+
| 0.6466 | 1.0417 | 100 | 0.6659 | -0.0460 | -0.1148 | 0.3234 | 0.0688 | -89.2840 | -77.7589 | -2.4671 | -2.4857 |
|
| 67 |
+
| 0.5655 | 2.0833 | 200 | 0.6495 | -0.1460 | -0.2967 | 0.3294 | 0.1507 | -102.2009 | -84.8621 | -2.3195 | -2.3414 |
|
| 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": 0.6178496645556556,
|
| 5 |
+
"train_runtime": 3593.4638,
|
| 6 |
+
"train_samples": 6113,
|
| 7 |
+
"train_samples_per_second": 5.103,
|
| 8 |
+
"train_steps_per_second": 0.08
|
| 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": 0.6178496645556556,
|
| 5 |
+
"train_runtime": 3593.4638,
|
| 6 |
+
"train_samples": 6113,
|
| 7 |
+
"train_samples_per_second": 5.103,
|
| 8 |
+
"train_steps_per_second": 0.08
|
| 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": 9.275770073122036,
|
| 14 |
+
"learning_rate": 9.788889404119949e-09,
|
| 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": 8.275367876853776,
|
| 29 |
+
"learning_rate": 9.78888940411995e-08,
|
| 30 |
+
"logits/chosen": -2.556431531906128,
|
| 31 |
+
"logits/rejected": -2.538444995880127,
|
| 32 |
+
"logps/chosen": -87.91434478759766,
|
| 33 |
+
"logps/rejected": -81.0243148803711,
|
| 34 |
+
"loss": 0.6931,
|
| 35 |
+
"rewards/accuracies": 0.2152777761220932,
|
| 36 |
+
"rewards/chosen": 1.1403658390918281e-05,
|
| 37 |
+
"rewards/margins": -0.00018849593470804393,
|
| 38 |
+
"rewards/rejected": 0.00019989960128441453,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.20833333333333334,
|
| 43 |
+
"grad_norm": 7.402776854966623,
|
| 44 |
+
"learning_rate": 1.95777788082399e-07,
|
| 45 |
+
"logits/chosen": -2.6145739555358887,
|
| 46 |
+
"logits/rejected": -2.565951347351074,
|
| 47 |
+
"logps/chosen": -103.30118560791016,
|
| 48 |
+
"logps/rejected": -89.87110137939453,
|
| 49 |
+
"loss": 0.693,
|
| 50 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 51 |
+
"rewards/chosen": 0.0007008779793977737,
|
| 52 |
+
"rewards/margins": 0.0005141849396750331,
|
| 53 |
+
"rewards/rejected": 0.00018669303972274065,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.3125,
|
| 58 |
+
"grad_norm": 9.949219453281035,
|
| 59 |
+
"learning_rate": 2.8278173946573535e-07,
|
| 60 |
+
"logits/chosen": -2.5197548866271973,
|
| 61 |
+
"logits/rejected": -2.534052848815918,
|
| 62 |
+
"logps/chosen": -67.16825103759766,
|
| 63 |
+
"logps/rejected": -74.81047821044922,
|
| 64 |
+
"loss": 0.6922,
|
| 65 |
+
"rewards/accuracies": 0.3187499940395355,
|
| 66 |
+
"rewards/chosen": 0.003788034664466977,
|
| 67 |
+
"rewards/margins": 0.002163059776648879,
|
| 68 |
+
"rewards/rejected": 0.00162497500423342,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.4166666666666667,
|
| 73 |
+
"grad_norm": 6.4502947310265,
|
| 74 |
+
"learning_rate": 2.7182120692830374e-07,
|
| 75 |
+
"logits/chosen": -2.5752604007720947,
|
| 76 |
+
"logits/rejected": -2.562830924987793,
|
| 77 |
+
"logps/chosen": -71.4312973022461,
|
| 78 |
+
"logps/rejected": -70.39434814453125,
|
| 79 |
+
"loss": 0.6905,
|
| 80 |
+
"rewards/accuracies": 0.26875001192092896,
|
| 81 |
+
"rewards/chosen": 0.010071559809148312,
|
| 82 |
+
"rewards/margins": 0.005580292548984289,
|
| 83 |
+
"rewards/rejected": 0.004491268657147884,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.5208333333333334,
|
| 88 |
+
"grad_norm": 9.621064053414667,
|
| 89 |
+
"learning_rate": 2.6086067439087217e-07,
|
| 90 |
+
"logits/chosen": -2.480384111404419,
|
| 91 |
+
"logits/rejected": -2.492711067199707,
|
| 92 |
+
"logps/chosen": -49.86461639404297,
|
| 93 |
+
"logps/rejected": -58.42780685424805,
|
| 94 |
+
"loss": 0.6875,
|
| 95 |
+
"rewards/accuracies": 0.23125000298023224,
|
| 96 |
+
"rewards/chosen": 0.014595480635762215,
|
| 97 |
+
"rewards/margins": 0.00781795009970665,
|
| 98 |
+
"rewards/rejected": 0.006777531001716852,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.625,
|
| 103 |
+
"grad_norm": 8.026421482041735,
|
| 104 |
+
"learning_rate": 2.4990014185344056e-07,
|
| 105 |
+
"logits/chosen": -2.5557591915130615,
|
| 106 |
+
"logits/rejected": -2.5367629528045654,
|
| 107 |
+
"logps/chosen": -77.70902252197266,
|
| 108 |
+
"logps/rejected": -79.41876983642578,
|
| 109 |
+
"loss": 0.6842,
|
| 110 |
+
"rewards/accuracies": 0.3125,
|
| 111 |
+
"rewards/chosen": 0.019152989611029625,
|
| 112 |
+
"rewards/margins": 0.020297441631555557,
|
| 113 |
+
"rewards/rejected": -0.0011444514384493232,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.7291666666666666,
|
| 118 |
+
"grad_norm": 9.050000945016462,
|
| 119 |
+
"learning_rate": 2.3893960931600894e-07,
|
| 120 |
+
"logits/chosen": -2.519299030303955,
|
| 121 |
+
"logits/rejected": -2.5085341930389404,
|
| 122 |
+
"logps/chosen": -94.42185974121094,
|
| 123 |
+
"logps/rejected": -86.42499542236328,
|
| 124 |
+
"loss": 0.6786,
|
| 125 |
+
"rewards/accuracies": 0.34375,
|
| 126 |
+
"rewards/chosen": 0.02707870863378048,
|
| 127 |
+
"rewards/margins": 0.040806300938129425,
|
| 128 |
+
"rewards/rejected": -0.013727596029639244,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.8333333333333334,
|
| 133 |
+
"grad_norm": 12.300153236365405,
|
| 134 |
+
"learning_rate": 2.2797907677857735e-07,
|
| 135 |
+
"logits/chosen": -2.5100162029266357,
|
| 136 |
+
"logits/rejected": -2.4531941413879395,
|
| 137 |
+
"logps/chosen": -82.90911865234375,
|
| 138 |
+
"logps/rejected": -81.5802230834961,
|
| 139 |
+
"loss": 0.6702,
|
| 140 |
+
"rewards/accuracies": 0.375,
|
| 141 |
+
"rewards/chosen": 0.038527704775333405,
|
| 142 |
+
"rewards/margins": 0.06947566568851471,
|
| 143 |
+
"rewards/rejected": -0.030947959050536156,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.9375,
|
| 148 |
+
"grad_norm": 12.056245748843725,
|
| 149 |
+
"learning_rate": 2.1701854424114576e-07,
|
| 150 |
+
"logits/chosen": -2.449734687805176,
|
| 151 |
+
"logits/rejected": -2.444124698638916,
|
| 152 |
+
"logps/chosen": -56.5576057434082,
|
| 153 |
+
"logps/rejected": -69.25230407714844,
|
| 154 |
+
"loss": 0.6753,
|
| 155 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 156 |
+
"rewards/chosen": -0.03629917651414871,
|
| 157 |
+
"rewards/margins": 0.04515077546238899,
|
| 158 |
+
"rewards/rejected": -0.081449955701828,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 1.0416666666666667,
|
| 163 |
+
"grad_norm": 10.09831393288397,
|
| 164 |
+
"learning_rate": 2.0605801170371414e-07,
|
| 165 |
+
"logits/chosen": -2.4516613483428955,
|
| 166 |
+
"logits/rejected": -2.431792736053467,
|
| 167 |
+
"logps/chosen": -72.71832275390625,
|
| 168 |
+
"logps/rejected": -80.36412048339844,
|
| 169 |
+
"loss": 0.6466,
|
| 170 |
+
"rewards/accuracies": 0.3687500059604645,
|
| 171 |
+
"rewards/chosen": 0.011315432377159595,
|
| 172 |
+
"rewards/margins": 0.12874236702919006,
|
| 173 |
+
"rewards/rejected": -0.11742694675922394,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 1.0416666666666667,
|
| 178 |
+
"eval_logits/chosen": -2.4856722354888916,
|
| 179 |
+
"eval_logits/rejected": -2.4671130180358887,
|
| 180 |
+
"eval_logps/chosen": -77.75892639160156,
|
| 181 |
+
"eval_logps/rejected": -89.28395080566406,
|
| 182 |
+
"eval_loss": 0.6659004092216492,
|
| 183 |
+
"eval_rewards/accuracies": 0.32341268658638,
|
| 184 |
+
"eval_rewards/chosen": -0.04601436108350754,
|
| 185 |
+
"eval_rewards/margins": 0.06882937997579575,
|
| 186 |
+
"eval_rewards/rejected": -0.11484373360872269,
|
| 187 |
+
"eval_runtime": 113.747,
|
| 188 |
+
"eval_samples_per_second": 17.583,
|
| 189 |
+
"eval_steps_per_second": 0.554,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 1.1458333333333333,
|
| 194 |
+
"grad_norm": 12.658367804317662,
|
| 195 |
+
"learning_rate": 1.9509747916628253e-07,
|
| 196 |
+
"logits/chosen": -2.4512288570404053,
|
| 197 |
+
"logits/rejected": -2.4872188568115234,
|
| 198 |
+
"logps/chosen": -66.1578369140625,
|
| 199 |
+
"logps/rejected": -93.39311218261719,
|
| 200 |
+
"loss": 0.6174,
|
| 201 |
+
"rewards/accuracies": 0.375,
|
| 202 |
+
"rewards/chosen": -0.011113283224403858,
|
| 203 |
+
"rewards/margins": 0.15658384561538696,
|
| 204 |
+
"rewards/rejected": -0.16769713163375854,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 1.25,
|
| 209 |
+
"grad_norm": 16.005314541222447,
|
| 210 |
+
"learning_rate": 1.8413694662885094e-07,
|
| 211 |
+
"logits/chosen": -2.4917149543762207,
|
| 212 |
+
"logits/rejected": -2.4566116333007812,
|
| 213 |
+
"logps/chosen": -105.62614440917969,
|
| 214 |
+
"logps/rejected": -108.29341125488281,
|
| 215 |
+
"loss": 0.6204,
|
| 216 |
+
"rewards/accuracies": 0.45625001192092896,
|
| 217 |
+
"rewards/chosen": 0.033508118242025375,
|
| 218 |
+
"rewards/margins": 0.24412801861763,
|
| 219 |
+
"rewards/rejected": -0.21061992645263672,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 1.3541666666666667,
|
| 224 |
+
"grad_norm": 18.85056869781382,
|
| 225 |
+
"learning_rate": 1.7317641409141935e-07,
|
| 226 |
+
"logits/chosen": -2.3912250995635986,
|
| 227 |
+
"logits/rejected": -2.3987691402435303,
|
| 228 |
+
"logps/chosen": -87.25572204589844,
|
| 229 |
+
"logps/rejected": -103.68363189697266,
|
| 230 |
+
"loss": 0.6116,
|
| 231 |
+
"rewards/accuracies": 0.375,
|
| 232 |
+
"rewards/chosen": -0.00971260853111744,
|
| 233 |
+
"rewards/margins": 0.19231685996055603,
|
| 234 |
+
"rewards/rejected": -0.20202946662902832,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 1.4583333333333333,
|
| 239 |
+
"grad_norm": 22.031793806569805,
|
| 240 |
+
"learning_rate": 1.6221588155398773e-07,
|
| 241 |
+
"logits/chosen": -2.375826358795166,
|
| 242 |
+
"logits/rejected": -2.40364408493042,
|
| 243 |
+
"logps/chosen": -50.64940643310547,
|
| 244 |
+
"logps/rejected": -78.66454315185547,
|
| 245 |
+
"loss": 0.6198,
|
| 246 |
+
"rewards/accuracies": 0.36250001192092896,
|
| 247 |
+
"rewards/chosen": -0.019174393266439438,
|
| 248 |
+
"rewards/margins": 0.17153134942054749,
|
| 249 |
+
"rewards/rejected": -0.19070573151111603,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 1.5625,
|
| 254 |
+
"grad_norm": 16.53568031633389,
|
| 255 |
+
"learning_rate": 1.5125534901655614e-07,
|
| 256 |
+
"logits/chosen": -2.4368748664855957,
|
| 257 |
+
"logits/rejected": -2.4099836349487305,
|
| 258 |
+
"logps/chosen": -78.88619232177734,
|
| 259 |
+
"logps/rejected": -97.15766906738281,
|
| 260 |
+
"loss": 0.6116,
|
| 261 |
+
"rewards/accuracies": 0.4124999940395355,
|
| 262 |
+
"rewards/chosen": -0.09488637745380402,
|
| 263 |
+
"rewards/margins": 0.2371809184551239,
|
| 264 |
+
"rewards/rejected": -0.3320673108100891,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 1.6666666666666665,
|
| 269 |
+
"grad_norm": 18.221103188672174,
|
| 270 |
+
"learning_rate": 1.4029481647912452e-07,
|
| 271 |
+
"logits/chosen": -2.391329050064087,
|
| 272 |
+
"logits/rejected": -2.402101993560791,
|
| 273 |
+
"logps/chosen": -87.55098724365234,
|
| 274 |
+
"logps/rejected": -114.4619369506836,
|
| 275 |
+
"loss": 0.6026,
|
| 276 |
+
"rewards/accuracies": 0.45625001192092896,
|
| 277 |
+
"rewards/chosen": -0.0933021530508995,
|
| 278 |
+
"rewards/margins": 0.2802050709724426,
|
| 279 |
+
"rewards/rejected": -0.3735072910785675,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 1.7708333333333335,
|
| 284 |
+
"grad_norm": 17.1043150905885,
|
| 285 |
+
"learning_rate": 1.293342839416929e-07,
|
| 286 |
+
"logits/chosen": -2.3455810546875,
|
| 287 |
+
"logits/rejected": -2.3396453857421875,
|
| 288 |
+
"logps/chosen": -77.65857696533203,
|
| 289 |
+
"logps/rejected": -104.78330993652344,
|
| 290 |
+
"loss": 0.6034,
|
| 291 |
+
"rewards/accuracies": 0.4625000059604645,
|
| 292 |
+
"rewards/chosen": -0.034337688237428665,
|
| 293 |
+
"rewards/margins": 0.27259936928749084,
|
| 294 |
+
"rewards/rejected": -0.306937038898468,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 1.875,
|
| 299 |
+
"grad_norm": 21.686788778311172,
|
| 300 |
+
"learning_rate": 1.1837375140426132e-07,
|
| 301 |
+
"logits/chosen": -2.368490219116211,
|
| 302 |
+
"logits/rejected": -2.337435483932495,
|
| 303 |
+
"logps/chosen": -78.26847076416016,
|
| 304 |
+
"logps/rejected": -98.10568237304688,
|
| 305 |
+
"loss": 0.5938,
|
| 306 |
+
"rewards/accuracies": 0.40625,
|
| 307 |
+
"rewards/chosen": -0.09030274301767349,
|
| 308 |
+
"rewards/margins": 0.25460249185562134,
|
| 309 |
+
"rewards/rejected": -0.3449052572250366,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 1.9791666666666665,
|
| 314 |
+
"grad_norm": 18.213789337934347,
|
| 315 |
+
"learning_rate": 1.0741321886682972e-07,
|
| 316 |
+
"logits/chosen": -2.3731675148010254,
|
| 317 |
+
"logits/rejected": -2.3687660694122314,
|
| 318 |
+
"logps/chosen": -88.2901611328125,
|
| 319 |
+
"logps/rejected": -111.50384521484375,
|
| 320 |
+
"loss": 0.5951,
|
| 321 |
+
"rewards/accuracies": 0.4124999940395355,
|
| 322 |
+
"rewards/chosen": 0.00241573597304523,
|
| 323 |
+
"rewards/margins": 0.29647231101989746,
|
| 324 |
+
"rewards/rejected": -0.29405659437179565,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 2.0833333333333335,
|
| 329 |
+
"grad_norm": 17.365297005169634,
|
| 330 |
+
"learning_rate": 9.645268632939811e-08,
|
| 331 |
+
"logits/chosen": -2.3061447143554688,
|
| 332 |
+
"logits/rejected": -2.3294944763183594,
|
| 333 |
+
"logps/chosen": -89.75189208984375,
|
| 334 |
+
"logps/rejected": -133.4348907470703,
|
| 335 |
+
"loss": 0.5655,
|
| 336 |
+
"rewards/accuracies": 0.5062500238418579,
|
| 337 |
+
"rewards/chosen": -0.01052729319781065,
|
| 338 |
+
"rewards/margins": 0.43865451216697693,
|
| 339 |
+
"rewards/rejected": -0.44918179512023926,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 2.0833333333333335,
|
| 344 |
+
"eval_logits/chosen": -2.3414034843444824,
|
| 345 |
+
"eval_logits/rejected": -2.319497585296631,
|
| 346 |
+
"eval_logps/chosen": -84.86211395263672,
|
| 347 |
+
"eval_logps/rejected": -102.20085906982422,
|
| 348 |
+
"eval_loss": 0.6495013236999512,
|
| 349 |
+
"eval_rewards/accuracies": 0.329365074634552,
|
| 350 |
+
"eval_rewards/chosen": -0.14603844285011292,
|
| 351 |
+
"eval_rewards/margins": 0.1506960541009903,
|
| 352 |
+
"eval_rewards/rejected": -0.296734482049942,
|
| 353 |
+
"eval_runtime": 113.581,
|
| 354 |
+
"eval_samples_per_second": 17.609,
|
| 355 |
+
"eval_steps_per_second": 0.555,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 2.1875,
|
| 360 |
+
"grad_norm": 36.763502445523656,
|
| 361 |
+
"learning_rate": 8.549215379196651e-08,
|
| 362 |
+
"logits/chosen": -2.348222017288208,
|
| 363 |
+
"logits/rejected": -2.3373069763183594,
|
| 364 |
+
"logps/chosen": -64.14533996582031,
|
| 365 |
+
"logps/rejected": -96.91500854492188,
|
| 366 |
+
"loss": 0.5582,
|
| 367 |
+
"rewards/accuracies": 0.40625,
|
| 368 |
+
"rewards/chosen": -0.0077634840272367,
|
| 369 |
+
"rewards/margins": 0.3896782100200653,
|
| 370 |
+
"rewards/rejected": -0.39744165539741516,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 2.2916666666666665,
|
| 375 |
+
"grad_norm": 20.09517432602137,
|
| 376 |
+
"learning_rate": 7.453162125453489e-08,
|
| 377 |
+
"logits/chosen": -2.359616756439209,
|
| 378 |
+
"logits/rejected": -2.295492172241211,
|
| 379 |
+
"logps/chosen": -106.1574478149414,
|
| 380 |
+
"logps/rejected": -127.97090911865234,
|
| 381 |
+
"loss": 0.546,
|
| 382 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 383 |
+
"rewards/chosen": -0.057844799011945724,
|
| 384 |
+
"rewards/margins": 0.44004330039024353,
|
| 385 |
+
"rewards/rejected": -0.4978880286216736,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 2.3958333333333335,
|
| 390 |
+
"grad_norm": 14.900227896678382,
|
| 391 |
+
"learning_rate": 6.35710887171033e-08,
|
| 392 |
+
"logits/chosen": -2.3140273094177246,
|
| 393 |
+
"logits/rejected": -2.2706756591796875,
|
| 394 |
+
"logps/chosen": -60.497161865234375,
|
| 395 |
+
"logps/rejected": -80.89750671386719,
|
| 396 |
+
"loss": 0.5638,
|
| 397 |
+
"rewards/accuracies": 0.39375001192092896,
|
| 398 |
+
"rewards/chosen": -0.009366447106003761,
|
| 399 |
+
"rewards/margins": 0.3419914245605469,
|
| 400 |
+
"rewards/rejected": -0.3513578772544861,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 2.5,
|
| 405 |
+
"grad_norm": 19.49103632925891,
|
| 406 |
+
"learning_rate": 5.2610556179671694e-08,
|
| 407 |
+
"logits/chosen": -2.2797024250030518,
|
| 408 |
+
"logits/rejected": -2.210596799850464,
|
| 409 |
+
"logps/chosen": -90.98292541503906,
|
| 410 |
+
"logps/rejected": -107.18631744384766,
|
| 411 |
+
"loss": 0.5714,
|
| 412 |
+
"rewards/accuracies": 0.4312500059604645,
|
| 413 |
+
"rewards/chosen": -0.04347435384988785,
|
| 414 |
+
"rewards/margins": 0.3474506735801697,
|
| 415 |
+
"rewards/rejected": -0.39092501997947693,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 2.6041666666666665,
|
| 420 |
+
"grad_norm": 25.396423499917123,
|
| 421 |
+
"learning_rate": 4.16500236422401e-08,
|
| 422 |
+
"logits/chosen": -2.221781015396118,
|
| 423 |
+
"logits/rejected": -2.205749034881592,
|
| 424 |
+
"logps/chosen": -81.74434661865234,
|
| 425 |
+
"logps/rejected": -102.39741516113281,
|
| 426 |
+
"loss": 0.5669,
|
| 427 |
+
"rewards/accuracies": 0.42500001192092896,
|
| 428 |
+
"rewards/chosen": -0.1277952641248703,
|
| 429 |
+
"rewards/margins": 0.32224446535110474,
|
| 430 |
+
"rewards/rejected": -0.45003968477249146,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 2.7083333333333335,
|
| 435 |
+
"grad_norm": 29.7738794213807,
|
| 436 |
+
"learning_rate": 3.0689491104808494e-08,
|
| 437 |
+
"logits/chosen": -2.2012267112731934,
|
| 438 |
+
"logits/rejected": -2.2187840938568115,
|
| 439 |
+
"logps/chosen": -59.17329025268555,
|
| 440 |
+
"logps/rejected": -96.04655456542969,
|
| 441 |
+
"loss": 0.5657,
|
| 442 |
+
"rewards/accuracies": 0.4000000059604645,
|
| 443 |
+
"rewards/chosen": -0.014679071493446827,
|
| 444 |
+
"rewards/margins": 0.31036660075187683,
|
| 445 |
+
"rewards/rejected": -0.325045645236969,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 2.8125,
|
| 450 |
+
"grad_norm": 19.448135836458437,
|
| 451 |
+
"learning_rate": 1.9728958567376888e-08,
|
| 452 |
+
"logits/chosen": -2.205172061920166,
|
| 453 |
+
"logits/rejected": -2.253237009048462,
|
| 454 |
+
"logps/chosen": -76.06713104248047,
|
| 455 |
+
"logps/rejected": -105.37796783447266,
|
| 456 |
+
"loss": 0.5524,
|
| 457 |
+
"rewards/accuracies": 0.4312500059604645,
|
| 458 |
+
"rewards/chosen": 0.008548146113753319,
|
| 459 |
+
"rewards/margins": 0.39001256227493286,
|
| 460 |
+
"rewards/rejected": -0.381464421749115,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 2.9166666666666665,
|
| 465 |
+
"grad_norm": 27.69086027593312,
|
| 466 |
+
"learning_rate": 8.768426029945283e-09,
|
| 467 |
+
"logits/chosen": -2.263256788253784,
|
| 468 |
+
"logits/rejected": -2.214367389678955,
|
| 469 |
+
"logps/chosen": -84.65340423583984,
|
| 470 |
+
"logps/rejected": -112.45893859863281,
|
| 471 |
+
"loss": 0.5571,
|
| 472 |
+
"rewards/accuracies": 0.4437499940395355,
|
| 473 |
+
"rewards/chosen": 0.009914604015648365,
|
| 474 |
+
"rewards/margins": 0.4335610270500183,
|
| 475 |
+
"rewards/rejected": -0.4236464500427246,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 3.0,
|
| 480 |
+
"step": 288,
|
| 481 |
+
"total_flos": 0.0,
|
| 482 |
+
"train_loss": 0.6178496645556556,
|
| 483 |
+
"train_runtime": 3593.4638,
|
| 484 |
+
"train_samples_per_second": 5.103,
|
| 485 |
+
"train_steps_per_second": 0.08
|
| 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 |
+
}
|