Model save
Browse files- README.md +80 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +815 -0
README.md
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
model-index:
|
| 9 |
+
- name: zephyr-7b-dpo-full-gpt_consistent-reward-scale-05
|
| 10 |
+
results: []
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 14 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 15 |
+
|
| 16 |
+
# zephyr-7b-dpo-full-gpt_consistent-reward-scale-05
|
| 17 |
+
|
| 18 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.4931
|
| 21 |
+
- Rewards/chosen: -1.6483
|
| 22 |
+
- Rewards/rejected: -2.9482
|
| 23 |
+
- Rewards/accuracies: 0.7414
|
| 24 |
+
- Rewards/margins: 1.2999
|
| 25 |
+
- Logps/rejected: -541.3424
|
| 26 |
+
- Logps/chosen: -449.9160
|
| 27 |
+
- Logits/rejected: 3.3584
|
| 28 |
+
- Logits/chosen: 2.0773
|
| 29 |
+
|
| 30 |
+
## Model description
|
| 31 |
+
|
| 32 |
+
More information needed
|
| 33 |
+
|
| 34 |
+
## Intended uses & limitations
|
| 35 |
+
|
| 36 |
+
More information needed
|
| 37 |
+
|
| 38 |
+
## Training and evaluation data
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Training procedure
|
| 43 |
+
|
| 44 |
+
### Training hyperparameters
|
| 45 |
+
|
| 46 |
+
The following hyperparameters were used during training:
|
| 47 |
+
- learning_rate: 5e-07
|
| 48 |
+
- train_batch_size: 8
|
| 49 |
+
- eval_batch_size: 8
|
| 50 |
+
- seed: 55
|
| 51 |
+
- distributed_type: multi-GPU
|
| 52 |
+
- num_devices: 8
|
| 53 |
+
- gradient_accumulation_steps: 2
|
| 54 |
+
- total_train_batch_size: 128
|
| 55 |
+
- total_eval_batch_size: 64
|
| 56 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 57 |
+
- lr_scheduler_type: cosine
|
| 58 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 59 |
+
- num_epochs: 1
|
| 60 |
+
|
| 61 |
+
### Training results
|
| 62 |
+
|
| 63 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 64 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 65 |
+
| 0.6653 | 0.1147 | 50 | 0.6513 | -0.0038 | -0.1422 | 0.6940 | 0.1384 | -260.7392 | -285.4682 | -2.4923 | -2.5731 |
|
| 66 |
+
| 0.5699 | 0.2294 | 100 | 0.5578 | -0.8876 | -1.6121 | 0.6853 | 0.7245 | -407.7342 | -373.8545 | 0.8771 | 0.3393 |
|
| 67 |
+
| 0.5416 | 0.3440 | 150 | 0.5320 | -1.1447 | -2.1157 | 0.7026 | 0.9709 | -458.0881 | -399.5647 | 2.1978 | 1.2442 |
|
| 68 |
+
| 0.5318 | 0.4587 | 200 | 0.5122 | -1.1906 | -2.2306 | 0.7284 | 1.0400 | -469.5803 | -404.1460 | 2.4483 | 1.3448 |
|
| 69 |
+
| 0.5178 | 0.5734 | 250 | 0.5029 | -1.4402 | -2.5615 | 0.7284 | 1.1212 | -502.6709 | -429.1149 | 2.5336 | 1.3215 |
|
| 70 |
+
| 0.519 | 0.6881 | 300 | 0.4985 | -1.4880 | -2.6823 | 0.7371 | 1.1943 | -514.7540 | -433.8906 | 2.6886 | 1.3895 |
|
| 71 |
+
| 0.5137 | 0.8028 | 350 | 0.4931 | -1.6128 | -2.8601 | 0.7328 | 1.2473 | -532.5296 | -446.3658 | 3.1580 | 1.8716 |
|
| 72 |
+
| 0.5033 | 0.9174 | 400 | 0.4931 | -1.6483 | -2.9482 | 0.7414 | 1.2999 | -541.3424 | -449.9160 | 3.3584 | 2.0773 |
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
### Framework versions
|
| 76 |
+
|
| 77 |
+
- Transformers 4.44.0.dev0
|
| 78 |
+
- Pytorch 2.1.2
|
| 79 |
+
- Datasets 2.20.0
|
| 80 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5464360101507344,
|
| 5 |
+
"train_runtime": 11333.4438,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.92,
|
| 8 |
+
"train_steps_per_second": 0.038
|
| 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": "4.44.0.dev0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5464360101507344,
|
| 5 |
+
"train_runtime": 11333.4438,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.92,
|
| 8 |
+
"train_steps_per_second": 0.038
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,815 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 50,
|
| 6 |
+
"global_step": 436,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.022935779816513763,
|
| 13 |
+
"grad_norm": 5.786441925419282,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.619342088699341,
|
| 16 |
+
"logits/rejected": -2.5526421070098877,
|
| 17 |
+
"logps/chosen": -265.45428466796875,
|
| 18 |
+
"logps/rejected": -236.1463165283203,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.41874998807907104,
|
| 21 |
+
"rewards/chosen": -8.460578828817233e-05,
|
| 22 |
+
"rewards/margins": 7.87067401688546e-05,
|
| 23 |
+
"rewards/rejected": -0.000163312564836815,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045871559633027525,
|
| 28 |
+
"grad_norm": 5.491351697550114,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.6576342582702637,
|
| 31 |
+
"logits/rejected": -2.5759005546569824,
|
| 32 |
+
"logps/chosen": -298.7987060546875,
|
| 33 |
+
"logps/rejected": -274.3077392578125,
|
| 34 |
+
"loss": 0.6923,
|
| 35 |
+
"rewards/accuracies": 0.668749988079071,
|
| 36 |
+
"rewards/chosen": -0.00013454208965413272,
|
| 37 |
+
"rewards/margins": 0.0018433767836540937,
|
| 38 |
+
"rewards/rejected": -0.0019779191352427006,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06880733944954129,
|
| 43 |
+
"grad_norm": 5.277174601748908,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.6759681701660156,
|
| 46 |
+
"logits/rejected": -2.602142333984375,
|
| 47 |
+
"logps/chosen": -290.32366943359375,
|
| 48 |
+
"logps/rejected": -234.38919067382812,
|
| 49 |
+
"loss": 0.6891,
|
| 50 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 51 |
+
"rewards/chosen": 0.010090610012412071,
|
| 52 |
+
"rewards/margins": 0.014163595624268055,
|
| 53 |
+
"rewards/rejected": -0.004072986543178558,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09174311926605505,
|
| 58 |
+
"grad_norm": 5.753580388451781,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.660831928253174,
|
| 61 |
+
"logits/rejected": -2.61124324798584,
|
| 62 |
+
"logps/chosen": -280.9427490234375,
|
| 63 |
+
"logps/rejected": -267.7388000488281,
|
| 64 |
+
"loss": 0.68,
|
| 65 |
+
"rewards/accuracies": 0.6875,
|
| 66 |
+
"rewards/chosen": 0.042264439165592194,
|
| 67 |
+
"rewards/margins": 0.04172234237194061,
|
| 68 |
+
"rewards/rejected": 0.0005421031382866204,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11467889908256881,
|
| 73 |
+
"grad_norm": 6.220985202592042,
|
| 74 |
+
"learning_rate": 4.997110275491701e-07,
|
| 75 |
+
"logits/chosen": -2.61970591545105,
|
| 76 |
+
"logits/rejected": -2.6113502979278564,
|
| 77 |
+
"logps/chosen": -294.2338562011719,
|
| 78 |
+
"logps/rejected": -305.00823974609375,
|
| 79 |
+
"loss": 0.6653,
|
| 80 |
+
"rewards/accuracies": 0.706250011920929,
|
| 81 |
+
"rewards/chosen": 0.020173203200101852,
|
| 82 |
+
"rewards/margins": 0.07841379195451736,
|
| 83 |
+
"rewards/rejected": -0.058240581303834915,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11467889908256881,
|
| 88 |
+
"eval_logits/chosen": -2.5731215476989746,
|
| 89 |
+
"eval_logits/rejected": -2.4923324584960938,
|
| 90 |
+
"eval_logps/chosen": -285.4681701660156,
|
| 91 |
+
"eval_logps/rejected": -260.73919677734375,
|
| 92 |
+
"eval_loss": 0.6512896418571472,
|
| 93 |
+
"eval_rewards/accuracies": 0.693965494632721,
|
| 94 |
+
"eval_rewards/chosen": -0.003780907019972801,
|
| 95 |
+
"eval_rewards/margins": 0.13838669657707214,
|
| 96 |
+
"eval_rewards/rejected": -0.1421675980091095,
|
| 97 |
+
"eval_runtime": 91.4355,
|
| 98 |
+
"eval_samples_per_second": 19.883,
|
| 99 |
+
"eval_steps_per_second": 0.317,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13761467889908258,
|
| 104 |
+
"grad_norm": 8.996643288143726,
|
| 105 |
+
"learning_rate": 4.979475034558115e-07,
|
| 106 |
+
"logits/chosen": -2.5592479705810547,
|
| 107 |
+
"logits/rejected": -2.5012314319610596,
|
| 108 |
+
"logps/chosen": -296.82501220703125,
|
| 109 |
+
"logps/rejected": -279.3923645019531,
|
| 110 |
+
"loss": 0.6437,
|
| 111 |
+
"rewards/accuracies": 0.675000011920929,
|
| 112 |
+
"rewards/chosen": -0.07340322434902191,
|
| 113 |
+
"rewards/margins": 0.14495481550693512,
|
| 114 |
+
"rewards/rejected": -0.21835803985595703,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16055045871559634,
|
| 119 |
+
"grad_norm": 9.499312526523301,
|
| 120 |
+
"learning_rate": 4.945923025551788e-07,
|
| 121 |
+
"logits/chosen": -2.4377551078796387,
|
| 122 |
+
"logits/rejected": -2.392364740371704,
|
| 123 |
+
"logps/chosen": -343.14935302734375,
|
| 124 |
+
"logps/rejected": -305.904541015625,
|
| 125 |
+
"loss": 0.6225,
|
| 126 |
+
"rewards/accuracies": 0.71875,
|
| 127 |
+
"rewards/chosen": -0.15578363835811615,
|
| 128 |
+
"rewards/margins": 0.2891853451728821,
|
| 129 |
+
"rewards/rejected": -0.4449689984321594,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.1834862385321101,
|
| 134 |
+
"grad_norm": 13.383949209241152,
|
| 135 |
+
"learning_rate": 4.896669632591651e-07,
|
| 136 |
+
"logits/chosen": -2.324429512023926,
|
| 137 |
+
"logits/rejected": -2.2132680416107178,
|
| 138 |
+
"logps/chosen": -312.23431396484375,
|
| 139 |
+
"logps/rejected": -314.443115234375,
|
| 140 |
+
"loss": 0.6094,
|
| 141 |
+
"rewards/accuracies": 0.731249988079071,
|
| 142 |
+
"rewards/chosen": -0.2638443112373352,
|
| 143 |
+
"rewards/margins": 0.35402628779411316,
|
| 144 |
+
"rewards/rejected": -0.617870569229126,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.20642201834862386,
|
| 149 |
+
"grad_norm": 13.193135435969568,
|
| 150 |
+
"learning_rate": 4.832031033425662e-07,
|
| 151 |
+
"logits/chosen": -1.3427231311798096,
|
| 152 |
+
"logits/rejected": -1.2150843143463135,
|
| 153 |
+
"logps/chosen": -337.8877258300781,
|
| 154 |
+
"logps/rejected": -351.61114501953125,
|
| 155 |
+
"loss": 0.588,
|
| 156 |
+
"rewards/accuracies": 0.699999988079071,
|
| 157 |
+
"rewards/chosen": -0.5029428601264954,
|
| 158 |
+
"rewards/margins": 0.5451563596725464,
|
| 159 |
+
"rewards/rejected": -1.048099160194397,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22935779816513763,
|
| 164 |
+
"grad_norm": 18.776758250311076,
|
| 165 |
+
"learning_rate": 4.752422169756047e-07,
|
| 166 |
+
"logits/chosen": -0.1327299326658249,
|
| 167 |
+
"logits/rejected": 0.009161601774394512,
|
| 168 |
+
"logps/chosen": -333.9382019042969,
|
| 169 |
+
"logps/rejected": -386.19952392578125,
|
| 170 |
+
"loss": 0.5699,
|
| 171 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 172 |
+
"rewards/chosen": -0.6151348352432251,
|
| 173 |
+
"rewards/margins": 0.5275757312774658,
|
| 174 |
+
"rewards/rejected": -1.142710566520691,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22935779816513763,
|
| 179 |
+
"eval_logits/chosen": 0.3393348455429077,
|
| 180 |
+
"eval_logits/rejected": 0.8771037459373474,
|
| 181 |
+
"eval_logps/chosen": -373.8545227050781,
|
| 182 |
+
"eval_logps/rejected": -407.7341613769531,
|
| 183 |
+
"eval_loss": 0.5577893257141113,
|
| 184 |
+
"eval_rewards/accuracies": 0.6853448152542114,
|
| 185 |
+
"eval_rewards/chosen": -0.8876444697380066,
|
| 186 |
+
"eval_rewards/margins": 0.7244728207588196,
|
| 187 |
+
"eval_rewards/rejected": -1.6121174097061157,
|
| 188 |
+
"eval_runtime": 92.82,
|
| 189 |
+
"eval_samples_per_second": 19.586,
|
| 190 |
+
"eval_steps_per_second": 0.312,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25229357798165136,
|
| 195 |
+
"grad_norm": 16.828561137860333,
|
| 196 |
+
"learning_rate": 4.658354083558188e-07,
|
| 197 |
+
"logits/chosen": 0.10855080932378769,
|
| 198 |
+
"logits/rejected": 0.6684142351150513,
|
| 199 |
+
"logps/chosen": -363.28369140625,
|
| 200 |
+
"logps/rejected": -392.91302490234375,
|
| 201 |
+
"loss": 0.5591,
|
| 202 |
+
"rewards/accuracies": 0.71875,
|
| 203 |
+
"rewards/chosen": -0.9809554815292358,
|
| 204 |
+
"rewards/margins": 0.6108429431915283,
|
| 205 |
+
"rewards/rejected": -1.5917984247207642,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.27522935779816515,
|
| 210 |
+
"grad_norm": 19.631970765133257,
|
| 211 |
+
"learning_rate": 4.550430636492389e-07,
|
| 212 |
+
"logits/chosen": 0.46126747131347656,
|
| 213 |
+
"logits/rejected": 0.8139599561691284,
|
| 214 |
+
"logps/chosen": -382.4955139160156,
|
| 215 |
+
"logps/rejected": -427.80206298828125,
|
| 216 |
+
"loss": 0.5619,
|
| 217 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 218 |
+
"rewards/chosen": -1.0667617321014404,
|
| 219 |
+
"rewards/margins": 0.6966198682785034,
|
| 220 |
+
"rewards/rejected": -1.7633816003799438,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.2981651376146789,
|
| 225 |
+
"grad_norm": 23.999804178074818,
|
| 226 |
+
"learning_rate": 4.429344633468004e-07,
|
| 227 |
+
"logits/chosen": 1.1685855388641357,
|
| 228 |
+
"logits/rejected": 1.665592908859253,
|
| 229 |
+
"logps/chosen": -352.095458984375,
|
| 230 |
+
"logps/rejected": -419.9580078125,
|
| 231 |
+
"loss": 0.5499,
|
| 232 |
+
"rewards/accuracies": 0.699999988079071,
|
| 233 |
+
"rewards/chosen": -0.9591536521911621,
|
| 234 |
+
"rewards/margins": 0.8143006563186646,
|
| 235 |
+
"rewards/rejected": -1.7734540700912476,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.3211009174311927,
|
| 240 |
+
"grad_norm": 18.64421738880492,
|
| 241 |
+
"learning_rate": 4.2958733752443187e-07,
|
| 242 |
+
"logits/chosen": 1.145119547843933,
|
| 243 |
+
"logits/rejected": 1.878694772720337,
|
| 244 |
+
"logps/chosen": -350.3140563964844,
|
| 245 |
+
"logps/rejected": -365.4109802246094,
|
| 246 |
+
"loss": 0.5568,
|
| 247 |
+
"rewards/accuracies": 0.706250011920929,
|
| 248 |
+
"rewards/chosen": -0.7412891387939453,
|
| 249 |
+
"rewards/margins": 0.6698554754257202,
|
| 250 |
+
"rewards/rejected": -1.411144733428955,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.3440366972477064,
|
| 255 |
+
"grad_norm": 20.200927048362935,
|
| 256 |
+
"learning_rate": 4.150873668617898e-07,
|
| 257 |
+
"logits/chosen": 1.7342513799667358,
|
| 258 |
+
"logits/rejected": 2.4978702068328857,
|
| 259 |
+
"logps/chosen": -387.82183837890625,
|
| 260 |
+
"logps/rejected": -444.968505859375,
|
| 261 |
+
"loss": 0.5416,
|
| 262 |
+
"rewards/accuracies": 0.706250011920929,
|
| 263 |
+
"rewards/chosen": -1.2127825021743774,
|
| 264 |
+
"rewards/margins": 0.9050270915031433,
|
| 265 |
+
"rewards/rejected": -2.117809772491455,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.3440366972477064,
|
| 270 |
+
"eval_logits/chosen": 1.2442059516906738,
|
| 271 |
+
"eval_logits/rejected": 2.1977624893188477,
|
| 272 |
+
"eval_logps/chosen": -399.5646667480469,
|
| 273 |
+
"eval_logps/rejected": -458.08807373046875,
|
| 274 |
+
"eval_loss": 0.5319975018501282,
|
| 275 |
+
"eval_rewards/accuracies": 0.7025862336158752,
|
| 276 |
+
"eval_rewards/chosen": -1.1447453498840332,
|
| 277 |
+
"eval_rewards/margins": 0.9709104895591736,
|
| 278 |
+
"eval_rewards/rejected": -2.1156558990478516,
|
| 279 |
+
"eval_runtime": 91.9947,
|
| 280 |
+
"eval_samples_per_second": 19.762,
|
| 281 |
+
"eval_steps_per_second": 0.315,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3669724770642202,
|
| 286 |
+
"grad_norm": 21.063284567870923,
|
| 287 |
+
"learning_rate": 3.9952763262280397e-07,
|
| 288 |
+
"logits/chosen": 1.3700188398361206,
|
| 289 |
+
"logits/rejected": 1.9712657928466797,
|
| 290 |
+
"logps/chosen": -422.37750244140625,
|
| 291 |
+
"logps/rejected": -497.3843688964844,
|
| 292 |
+
"loss": 0.5292,
|
| 293 |
+
"rewards/accuracies": 0.706250011920929,
|
| 294 |
+
"rewards/chosen": -1.2582637071609497,
|
| 295 |
+
"rewards/margins": 0.8558940887451172,
|
| 296 |
+
"rewards/rejected": -2.1141579151153564,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38990825688073394,
|
| 301 |
+
"grad_norm": 20.8460851273499,
|
| 302 |
+
"learning_rate": 3.8300801912883414e-07,
|
| 303 |
+
"logits/chosen": 1.7427918910980225,
|
| 304 |
+
"logits/rejected": 2.409679412841797,
|
| 305 |
+
"logps/chosen": -401.26861572265625,
|
| 306 |
+
"logps/rejected": -495.7734375,
|
| 307 |
+
"loss": 0.5061,
|
| 308 |
+
"rewards/accuracies": 0.71875,
|
| 309 |
+
"rewards/chosen": -1.2966772317886353,
|
| 310 |
+
"rewards/margins": 0.9603677988052368,
|
| 311 |
+
"rewards/rejected": -2.257045269012451,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.41284403669724773,
|
| 316 |
+
"grad_norm": 17.79571920632354,
|
| 317 |
+
"learning_rate": 3.6563457256020884e-07,
|
| 318 |
+
"logits/chosen": 1.8351173400878906,
|
| 319 |
+
"logits/rejected": 2.8393807411193848,
|
| 320 |
+
"logps/chosen": -433.7701110839844,
|
| 321 |
+
"logps/rejected": -456.1665954589844,
|
| 322 |
+
"loss": 0.5391,
|
| 323 |
+
"rewards/accuracies": 0.731249988079071,
|
| 324 |
+
"rewards/chosen": -1.3041222095489502,
|
| 325 |
+
"rewards/margins": 0.8725945353507996,
|
| 326 |
+
"rewards/rejected": -2.1767165660858154,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.43577981651376146,
|
| 331 |
+
"grad_norm": 20.925215758529603,
|
| 332 |
+
"learning_rate": 3.475188202022617e-07,
|
| 333 |
+
"logits/chosen": 2.0596728324890137,
|
| 334 |
+
"logits/rejected": 2.587247371673584,
|
| 335 |
+
"logps/chosen": -375.49212646484375,
|
| 336 |
+
"logps/rejected": -489.371826171875,
|
| 337 |
+
"loss": 0.517,
|
| 338 |
+
"rewards/accuracies": 0.78125,
|
| 339 |
+
"rewards/chosen": -1.1737760305404663,
|
| 340 |
+
"rewards/margins": 1.0764976739883423,
|
| 341 |
+
"rewards/rejected": -2.2502734661102295,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45871559633027525,
|
| 346 |
+
"grad_norm": 21.33737350989529,
|
| 347 |
+
"learning_rate": 3.287770545059052e-07,
|
| 348 |
+
"logits/chosen": 2.203148365020752,
|
| 349 |
+
"logits/rejected": 2.9923298358917236,
|
| 350 |
+
"logps/chosen": -424.408203125,
|
| 351 |
+
"logps/rejected": -482.44879150390625,
|
| 352 |
+
"loss": 0.5318,
|
| 353 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 354 |
+
"rewards/chosen": -1.4612659215927124,
|
| 355 |
+
"rewards/margins": 0.9246621131896973,
|
| 356 |
+
"rewards/rejected": -2.38592791557312,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45871559633027525,
|
| 361 |
+
"eval_logits/chosen": 1.3448154926300049,
|
| 362 |
+
"eval_logits/rejected": 2.4483416080474854,
|
| 363 |
+
"eval_logps/chosen": -404.1459655761719,
|
| 364 |
+
"eval_logps/rejected": -469.58026123046875,
|
| 365 |
+
"eval_loss": 0.5122287273406982,
|
| 366 |
+
"eval_rewards/accuracies": 0.7284482717514038,
|
| 367 |
+
"eval_rewards/chosen": -1.1905584335327148,
|
| 368 |
+
"eval_rewards/margins": 1.0400197505950928,
|
| 369 |
+
"eval_rewards/rejected": -2.2305781841278076,
|
| 370 |
+
"eval_runtime": 92.2657,
|
| 371 |
+
"eval_samples_per_second": 19.704,
|
| 372 |
+
"eval_steps_per_second": 0.314,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.481651376146789,
|
| 377 |
+
"grad_norm": 22.48419176653969,
|
| 378 |
+
"learning_rate": 3.0952958655864954e-07,
|
| 379 |
+
"logits/chosen": 1.9083038568496704,
|
| 380 |
+
"logits/rejected": 2.511636734008789,
|
| 381 |
+
"logps/chosen": -405.66839599609375,
|
| 382 |
+
"logps/rejected": -489.11663818359375,
|
| 383 |
+
"loss": 0.5124,
|
| 384 |
+
"rewards/accuracies": 0.71875,
|
| 385 |
+
"rewards/chosen": -1.3236396312713623,
|
| 386 |
+
"rewards/margins": 1.0241048336029053,
|
| 387 |
+
"rewards/rejected": -2.3477444648742676,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5045871559633027,
|
| 392 |
+
"grad_norm": 24.286637463373633,
|
| 393 |
+
"learning_rate": 2.898999737583448e-07,
|
| 394 |
+
"logits/chosen": 1.8827180862426758,
|
| 395 |
+
"logits/rejected": 2.9742465019226074,
|
| 396 |
+
"logps/chosen": -462.0406799316406,
|
| 397 |
+
"logps/rejected": -545.8687133789062,
|
| 398 |
+
"loss": 0.503,
|
| 399 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 400 |
+
"rewards/chosen": -1.3936035633087158,
|
| 401 |
+
"rewards/margins": 1.215381383895874,
|
| 402 |
+
"rewards/rejected": -2.6089847087860107,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5275229357798165,
|
| 407 |
+
"grad_norm": 19.94988074398488,
|
| 408 |
+
"learning_rate": 2.7001422664752333e-07,
|
| 409 |
+
"logits/chosen": 1.1973925828933716,
|
| 410 |
+
"logits/rejected": 1.7874739170074463,
|
| 411 |
+
"logps/chosen": -377.791748046875,
|
| 412 |
+
"logps/rejected": -472.1722717285156,
|
| 413 |
+
"loss": 0.5198,
|
| 414 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 415 |
+
"rewards/chosen": -1.0984822511672974,
|
| 416 |
+
"rewards/margins": 0.9670408368110657,
|
| 417 |
+
"rewards/rejected": -2.0655229091644287,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5504587155963303,
|
| 422 |
+
"grad_norm": 24.247680553138093,
|
| 423 |
+
"learning_rate": 2.5e-07,
|
| 424 |
+
"logits/chosen": 1.5466783046722412,
|
| 425 |
+
"logits/rejected": 2.6804802417755127,
|
| 426 |
+
"logps/chosen": -449.4345703125,
|
| 427 |
+
"logps/rejected": -522.0684814453125,
|
| 428 |
+
"loss": 0.5279,
|
| 429 |
+
"rewards/accuracies": 0.699999988079071,
|
| 430 |
+
"rewards/chosen": -1.6190277338027954,
|
| 431 |
+
"rewards/margins": 0.9881566762924194,
|
| 432 |
+
"rewards/rejected": -2.6071841716766357,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.573394495412844,
|
| 437 |
+
"grad_norm": 18.449169539289652,
|
| 438 |
+
"learning_rate": 2.2998577335247667e-07,
|
| 439 |
+
"logits/chosen": 1.7379165887832642,
|
| 440 |
+
"logits/rejected": 2.9272639751434326,
|
| 441 |
+
"logps/chosen": -468.9725646972656,
|
| 442 |
+
"logps/rejected": -532.5216674804688,
|
| 443 |
+
"loss": 0.5178,
|
| 444 |
+
"rewards/accuracies": 0.75,
|
| 445 |
+
"rewards/chosen": -1.6517877578735352,
|
| 446 |
+
"rewards/margins": 1.06899094581604,
|
| 447 |
+
"rewards/rejected": -2.720778226852417,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.573394495412844,
|
| 452 |
+
"eval_logits/chosen": 1.3214832544326782,
|
| 453 |
+
"eval_logits/rejected": 2.5336461067199707,
|
| 454 |
+
"eval_logps/chosen": -429.1148681640625,
|
| 455 |
+
"eval_logps/rejected": -502.6709289550781,
|
| 456 |
+
"eval_loss": 0.5028622150421143,
|
| 457 |
+
"eval_rewards/accuracies": 0.7284482717514038,
|
| 458 |
+
"eval_rewards/chosen": -1.440247893333435,
|
| 459 |
+
"eval_rewards/margins": 1.1212375164031982,
|
| 460 |
+
"eval_rewards/rejected": -2.561485528945923,
|
| 461 |
+
"eval_runtime": 92.4677,
|
| 462 |
+
"eval_samples_per_second": 19.661,
|
| 463 |
+
"eval_steps_per_second": 0.314,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5963302752293578,
|
| 468 |
+
"grad_norm": 18.218334579311854,
|
| 469 |
+
"learning_rate": 2.1010002624165524e-07,
|
| 470 |
+
"logits/chosen": 1.2481517791748047,
|
| 471 |
+
"logits/rejected": 2.1310501098632812,
|
| 472 |
+
"logps/chosen": -418.47222900390625,
|
| 473 |
+
"logps/rejected": -541.3790283203125,
|
| 474 |
+
"loss": 0.5073,
|
| 475 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 476 |
+
"rewards/chosen": -1.4775768518447876,
|
| 477 |
+
"rewards/margins": 1.1922760009765625,
|
| 478 |
+
"rewards/rejected": -2.6698527336120605,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6192660550458715,
|
| 483 |
+
"grad_norm": 18.269676157352414,
|
| 484 |
+
"learning_rate": 1.9047041344135043e-07,
|
| 485 |
+
"logits/chosen": 1.8889141082763672,
|
| 486 |
+
"logits/rejected": 2.4701380729675293,
|
| 487 |
+
"logps/chosen": -426.7981872558594,
|
| 488 |
+
"logps/rejected": -536.7744750976562,
|
| 489 |
+
"loss": 0.504,
|
| 490 |
+
"rewards/accuracies": 0.75,
|
| 491 |
+
"rewards/chosen": -1.607358694076538,
|
| 492 |
+
"rewards/margins": 1.241914987564087,
|
| 493 |
+
"rewards/rejected": -2.849273920059204,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.6422018348623854,
|
| 498 |
+
"grad_norm": 21.937738574869034,
|
| 499 |
+
"learning_rate": 1.7122294549409482e-07,
|
| 500 |
+
"logits/chosen": 2.0935568809509277,
|
| 501 |
+
"logits/rejected": 3.0940475463867188,
|
| 502 |
+
"logps/chosen": -444.2198181152344,
|
| 503 |
+
"logps/rejected": -559.611328125,
|
| 504 |
+
"loss": 0.5276,
|
| 505 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 506 |
+
"rewards/chosen": -1.7122100591659546,
|
| 507 |
+
"rewards/margins": 1.2459454536437988,
|
| 508 |
+
"rewards/rejected": -2.958155393600464,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6651376146788991,
|
| 513 |
+
"grad_norm": 23.622077882281282,
|
| 514 |
+
"learning_rate": 1.524811797977383e-07,
|
| 515 |
+
"logits/chosen": 2.260361433029175,
|
| 516 |
+
"logits/rejected": 3.3843066692352295,
|
| 517 |
+
"logps/chosen": -451.55938720703125,
|
| 518 |
+
"logps/rejected": -546.6029052734375,
|
| 519 |
+
"loss": 0.5071,
|
| 520 |
+
"rewards/accuracies": 0.768750011920929,
|
| 521 |
+
"rewards/chosen": -1.6382853984832764,
|
| 522 |
+
"rewards/margins": 1.288576364517212,
|
| 523 |
+
"rewards/rejected": -2.9268617630004883,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6880733944954128,
|
| 528 |
+
"grad_norm": 20.079829542188833,
|
| 529 |
+
"learning_rate": 1.3436542743979125e-07,
|
| 530 |
+
"logits/chosen": 1.4699538946151733,
|
| 531 |
+
"logits/rejected": 2.1643927097320557,
|
| 532 |
+
"logps/chosen": -453.72039794921875,
|
| 533 |
+
"logps/rejected": -500.624267578125,
|
| 534 |
+
"loss": 0.519,
|
| 535 |
+
"rewards/accuracies": 0.762499988079071,
|
| 536 |
+
"rewards/chosen": -1.4777113199234009,
|
| 537 |
+
"rewards/margins": 0.939932644367218,
|
| 538 |
+
"rewards/rejected": -2.4176440238952637,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6880733944954128,
|
| 543 |
+
"eval_logits/chosen": 1.3894954919815063,
|
| 544 |
+
"eval_logits/rejected": 2.688563346862793,
|
| 545 |
+
"eval_logps/chosen": -433.8905944824219,
|
| 546 |
+
"eval_logps/rejected": -514.7540283203125,
|
| 547 |
+
"eval_loss": 0.4985021948814392,
|
| 548 |
+
"eval_rewards/accuracies": 0.7370689511299133,
|
| 549 |
+
"eval_rewards/chosen": -1.488004446029663,
|
| 550 |
+
"eval_rewards/margins": 1.1943109035491943,
|
| 551 |
+
"eval_rewards/rejected": -2.6823153495788574,
|
| 552 |
+
"eval_runtime": 91.7036,
|
| 553 |
+
"eval_samples_per_second": 19.825,
|
| 554 |
+
"eval_steps_per_second": 0.316,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7110091743119266,
|
| 559 |
+
"grad_norm": 20.41756706068905,
|
| 560 |
+
"learning_rate": 1.1699198087116588e-07,
|
| 561 |
+
"logits/chosen": 1.4253332614898682,
|
| 562 |
+
"logits/rejected": 2.419015407562256,
|
| 563 |
+
"logps/chosen": -438.324462890625,
|
| 564 |
+
"logps/rejected": -531.0131225585938,
|
| 565 |
+
"loss": 0.5304,
|
| 566 |
+
"rewards/accuracies": 0.731249988079071,
|
| 567 |
+
"rewards/chosen": -1.603592872619629,
|
| 568 |
+
"rewards/margins": 1.0420410633087158,
|
| 569 |
+
"rewards/rejected": -2.6456339359283447,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7339449541284404,
|
| 574 |
+
"grad_norm": 23.710557812034125,
|
| 575 |
+
"learning_rate": 1.00472367377196e-07,
|
| 576 |
+
"logits/chosen": 1.8713343143463135,
|
| 577 |
+
"logits/rejected": 3.118149757385254,
|
| 578 |
+
"logps/chosen": -430.84912109375,
|
| 579 |
+
"logps/rejected": -534.192626953125,
|
| 580 |
+
"loss": 0.5078,
|
| 581 |
+
"rewards/accuracies": 0.768750011920929,
|
| 582 |
+
"rewards/chosen": -1.5661717653274536,
|
| 583 |
+
"rewards/margins": 1.4308522939682007,
|
| 584 |
+
"rewards/rejected": -2.9970240592956543,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7568807339449541,
|
| 589 |
+
"grad_norm": 22.190813187925453,
|
| 590 |
+
"learning_rate": 8.49126331382102e-08,
|
| 591 |
+
"logits/chosen": 2.05544376373291,
|
| 592 |
+
"logits/rejected": 2.8256397247314453,
|
| 593 |
+
"logps/chosen": -446.35931396484375,
|
| 594 |
+
"logps/rejected": -508.9071350097656,
|
| 595 |
+
"loss": 0.5035,
|
| 596 |
+
"rewards/accuracies": 0.675000011920929,
|
| 597 |
+
"rewards/chosen": -1.7183713912963867,
|
| 598 |
+
"rewards/margins": 0.8996985554695129,
|
| 599 |
+
"rewards/rejected": -2.618070125579834,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7798165137614679,
|
| 604 |
+
"grad_norm": 26.678335434261424,
|
| 605 |
+
"learning_rate": 7.041266247556812e-08,
|
| 606 |
+
"logits/chosen": 1.8211781978607178,
|
| 607 |
+
"logits/rejected": 2.5711045265197754,
|
| 608 |
+
"logps/chosen": -450.910400390625,
|
| 609 |
+
"logps/rejected": -517.4847412109375,
|
| 610 |
+
"loss": 0.501,
|
| 611 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 612 |
+
"rewards/chosen": -1.6291780471801758,
|
| 613 |
+
"rewards/margins": 0.9458298683166504,
|
| 614 |
+
"rewards/rejected": -2.575007677078247,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8027522935779816,
|
| 619 |
+
"grad_norm": 22.960680980618804,
|
| 620 |
+
"learning_rate": 5.706553665319955e-08,
|
| 621 |
+
"logits/chosen": 1.8633334636688232,
|
| 622 |
+
"logits/rejected": 3.0288431644439697,
|
| 623 |
+
"logps/chosen": -437.11260986328125,
|
| 624 |
+
"logps/rejected": -511.2594299316406,
|
| 625 |
+
"loss": 0.5137,
|
| 626 |
+
"rewards/accuracies": 0.75,
|
| 627 |
+
"rewards/chosen": -1.6262149810791016,
|
| 628 |
+
"rewards/margins": 1.1097644567489624,
|
| 629 |
+
"rewards/rejected": -2.7359795570373535,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8027522935779816,
|
| 634 |
+
"eval_logits/chosen": 1.8716328144073486,
|
| 635 |
+
"eval_logits/rejected": 3.1580371856689453,
|
| 636 |
+
"eval_logps/chosen": -446.3657531738281,
|
| 637 |
+
"eval_logps/rejected": -532.5296020507812,
|
| 638 |
+
"eval_loss": 0.49309831857681274,
|
| 639 |
+
"eval_rewards/accuracies": 0.732758641242981,
|
| 640 |
+
"eval_rewards/chosen": -1.612756371498108,
|
| 641 |
+
"eval_rewards/margins": 1.247314453125,
|
| 642 |
+
"eval_rewards/rejected": -2.8600711822509766,
|
| 643 |
+
"eval_runtime": 91.4764,
|
| 644 |
+
"eval_samples_per_second": 19.874,
|
| 645 |
+
"eval_steps_per_second": 0.317,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8256880733944955,
|
| 650 |
+
"grad_norm": 25.170197026630024,
|
| 651 |
+
"learning_rate": 4.4956936350761005e-08,
|
| 652 |
+
"logits/chosen": 1.9063358306884766,
|
| 653 |
+
"logits/rejected": 2.6921870708465576,
|
| 654 |
+
"logps/chosen": -398.64984130859375,
|
| 655 |
+
"logps/rejected": -524.7725830078125,
|
| 656 |
+
"loss": 0.4898,
|
| 657 |
+
"rewards/accuracies": 0.6875,
|
| 658 |
+
"rewards/chosen": -1.4632974863052368,
|
| 659 |
+
"rewards/margins": 1.1990567445755005,
|
| 660 |
+
"rewards/rejected": -2.662353992462158,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8486238532110092,
|
| 665 |
+
"grad_norm": 23.25403059611051,
|
| 666 |
+
"learning_rate": 3.416459164418123e-08,
|
| 667 |
+
"logits/chosen": 1.4586713314056396,
|
| 668 |
+
"logits/rejected": 2.6852428913116455,
|
| 669 |
+
"logps/chosen": -454.4266052246094,
|
| 670 |
+
"logps/rejected": -542.4071044921875,
|
| 671 |
+
"loss": 0.4926,
|
| 672 |
+
"rewards/accuracies": 0.75,
|
| 673 |
+
"rewards/chosen": -1.5484213829040527,
|
| 674 |
+
"rewards/margins": 1.2499048709869385,
|
| 675 |
+
"rewards/rejected": -2.798326015472412,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8715596330275229,
|
| 680 |
+
"grad_norm": 28.54168436834153,
|
| 681 |
+
"learning_rate": 2.475778302439524e-08,
|
| 682 |
+
"logits/chosen": 1.7404067516326904,
|
| 683 |
+
"logits/rejected": 3.0292108058929443,
|
| 684 |
+
"logps/chosen": -451.39849853515625,
|
| 685 |
+
"logps/rejected": -544.4757080078125,
|
| 686 |
+
"loss": 0.5001,
|
| 687 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 688 |
+
"rewards/chosen": -1.5502045154571533,
|
| 689 |
+
"rewards/margins": 1.3179900646209717,
|
| 690 |
+
"rewards/rejected": -2.868194818496704,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8944954128440367,
|
| 695 |
+
"grad_norm": 25.43427497983774,
|
| 696 |
+
"learning_rate": 1.6796896657433805e-08,
|
| 697 |
+
"logits/chosen": 2.222611904144287,
|
| 698 |
+
"logits/rejected": 3.231661558151245,
|
| 699 |
+
"logps/chosen": -413.8394470214844,
|
| 700 |
+
"logps/rejected": -501.4020080566406,
|
| 701 |
+
"loss": 0.515,
|
| 702 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 703 |
+
"rewards/chosen": -1.6188195943832397,
|
| 704 |
+
"rewards/margins": 1.1021382808685303,
|
| 705 |
+
"rewards/rejected": -2.7209575176239014,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9174311926605505,
|
| 710 |
+
"grad_norm": 24.959131634596684,
|
| 711 |
+
"learning_rate": 1.0333036740834855e-08,
|
| 712 |
+
"logits/chosen": 2.535111427307129,
|
| 713 |
+
"logits/rejected": 3.2772762775421143,
|
| 714 |
+
"logps/chosen": -382.0926208496094,
|
| 715 |
+
"logps/rejected": -499.1607971191406,
|
| 716 |
+
"loss": 0.5033,
|
| 717 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 718 |
+
"rewards/chosen": -1.5753300189971924,
|
| 719 |
+
"rewards/margins": 1.1241223812103271,
|
| 720 |
+
"rewards/rejected": -2.6994521617889404,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9174311926605505,
|
| 725 |
+
"eval_logits/chosen": 2.0773062705993652,
|
| 726 |
+
"eval_logits/rejected": 3.358436346054077,
|
| 727 |
+
"eval_logps/chosen": -449.9159851074219,
|
| 728 |
+
"eval_logps/rejected": -541.3424072265625,
|
| 729 |
+
"eval_loss": 0.49311476945877075,
|
| 730 |
+
"eval_rewards/accuracies": 0.7413793206214905,
|
| 731 |
+
"eval_rewards/chosen": -1.6482590436935425,
|
| 732 |
+
"eval_rewards/margins": 1.299940824508667,
|
| 733 |
+
"eval_rewards/rejected": -2.948199987411499,
|
| 734 |
+
"eval_runtime": 92.5987,
|
| 735 |
+
"eval_samples_per_second": 19.633,
|
| 736 |
+
"eval_steps_per_second": 0.313,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9403669724770642,
|
| 741 |
+
"grad_norm": 28.32448783876109,
|
| 742 |
+
"learning_rate": 5.4076974448211685e-09,
|
| 743 |
+
"logits/chosen": 2.5014076232910156,
|
| 744 |
+
"logits/rejected": 3.729045867919922,
|
| 745 |
+
"logps/chosen": -446.04974365234375,
|
| 746 |
+
"logps/rejected": -546.2440185546875,
|
| 747 |
+
"loss": 0.4985,
|
| 748 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 749 |
+
"rewards/chosen": -1.7509788274765015,
|
| 750 |
+
"rewards/margins": 1.3309944868087769,
|
| 751 |
+
"rewards/rejected": -3.0819735527038574,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.963302752293578,
|
| 756 |
+
"grad_norm": 31.377213722949094,
|
| 757 |
+
"learning_rate": 2.052496544188487e-09,
|
| 758 |
+
"logits/chosen": 2.54590106010437,
|
| 759 |
+
"logits/rejected": 3.5291149616241455,
|
| 760 |
+
"logps/chosen": -431.8235778808594,
|
| 761 |
+
"logps/rejected": -553.3231201171875,
|
| 762 |
+
"loss": 0.4966,
|
| 763 |
+
"rewards/accuracies": 0.762499988079071,
|
| 764 |
+
"rewards/chosen": -1.723846197128296,
|
| 765 |
+
"rewards/margins": 1.350610613822937,
|
| 766 |
+
"rewards/rejected": -3.0744569301605225,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9862385321100917,
|
| 771 |
+
"grad_norm": 25.27983521226701,
|
| 772 |
+
"learning_rate": 2.889724508297886e-10,
|
| 773 |
+
"logits/chosen": 2.104177713394165,
|
| 774 |
+
"logits/rejected": 3.586791515350342,
|
| 775 |
+
"logps/chosen": -467.50860595703125,
|
| 776 |
+
"logps/rejected": -525.298828125,
|
| 777 |
+
"loss": 0.4944,
|
| 778 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 779 |
+
"rewards/chosen": -1.64333176612854,
|
| 780 |
+
"rewards/margins": 1.1692912578582764,
|
| 781 |
+
"rewards/rejected": -2.8126227855682373,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 1.0,
|
| 786 |
+
"step": 436,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.5464360101507344,
|
| 789 |
+
"train_runtime": 11333.4438,
|
| 790 |
+
"train_samples_per_second": 4.92,
|
| 791 |
+
"train_steps_per_second": 0.038
|
| 792 |
+
}
|
| 793 |
+
],
|
| 794 |
+
"logging_steps": 10,
|
| 795 |
+
"max_steps": 436,
|
| 796 |
+
"num_input_tokens_seen": 0,
|
| 797 |
+
"num_train_epochs": 1,
|
| 798 |
+
"save_steps": 100,
|
| 799 |
+
"stateful_callbacks": {
|
| 800 |
+
"TrainerControl": {
|
| 801 |
+
"args": {
|
| 802 |
+
"should_epoch_stop": false,
|
| 803 |
+
"should_evaluate": false,
|
| 804 |
+
"should_log": false,
|
| 805 |
+
"should_save": true,
|
| 806 |
+
"should_training_stop": true
|
| 807 |
+
},
|
| 808 |
+
"attributes": {}
|
| 809 |
+
}
|
| 810 |
+
},
|
| 811 |
+
"total_flos": 0.0,
|
| 812 |
+
"train_batch_size": 8,
|
| 813 |
+
"trial_name": null,
|
| 814 |
+
"trial_params": null
|
| 815 |
+
}
|