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-prometheus-reward-scale-01
|
| 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-prometheus-reward-scale-01
|
| 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: 0.5012
|
| 21 |
+
- Rewards/chosen: -1.7553
|
| 22 |
+
- Rewards/rejected: -2.9981
|
| 23 |
+
- Rewards/accuracies: 0.7198
|
| 24 |
+
- Rewards/margins: 1.2428
|
| 25 |
+
- Logps/rejected: -548.0841
|
| 26 |
+
- Logps/chosen: -435.4877
|
| 27 |
+
- Logits/rejected: 3.0596
|
| 28 |
+
- Logits/chosen: 1.9658
|
| 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 |
+
| 0.6661 | 0.1143 | 50 | 0.6541 | -0.0354 | -0.1594 | 0.6552 | 0.1240 | -264.2141 | -263.4992 | -2.5732 | -2.6193 |
|
| 66 |
+
| 0.561 | 0.2286 | 100 | 0.5663 | -1.0515 | -1.8367 | 0.7069 | 0.7852 | -431.9447 | -365.1131 | -0.0076 | -0.4111 |
|
| 67 |
+
| 0.5324 | 0.3429 | 150 | 0.5437 | -1.4518 | -2.4385 | 0.6853 | 0.9868 | -492.1300 | -405.1368 | 2.0029 | 1.3258 |
|
| 68 |
+
| 0.5261 | 0.4571 | 200 | 0.5247 | -1.5625 | -2.5913 | 0.6853 | 1.0288 | -507.4055 | -416.2077 | 2.7389 | 1.7313 |
|
| 69 |
+
| 0.5274 | 0.5714 | 250 | 0.5148 | -1.6815 | -2.8054 | 0.7155 | 1.1239 | -528.8192 | -428.1107 | 2.1266 | 1.0144 |
|
| 70 |
+
| 0.5 | 0.6857 | 300 | 0.5078 | -1.6879 | -2.8754 | 0.7198 | 1.1875 | -535.8170 | -428.7552 | 2.7028 | 1.5160 |
|
| 71 |
+
| 0.4879 | 0.8 | 350 | 0.5050 | -1.8872 | -3.0745 | 0.7198 | 1.1873 | -555.7252 | -448.6785 | 3.2477 | 2.2065 |
|
| 72 |
+
| 0.5082 | 0.9143 | 400 | 0.5012 | -1.7553 | -2.9981 | 0.7198 | 1.2428 | -548.0841 | -435.4877 | 3.0596 | 1.9658 |
|
| 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": 0.9988571428571429,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5499054316252265,
|
| 5 |
+
"train_runtime": 11738.6737,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.77,
|
| 8 |
+
"train_steps_per_second": 0.037
|
| 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": 0.9988571428571429,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5499054316252265,
|
| 5 |
+
"train_runtime": 11738.6737,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.77,
|
| 8 |
+
"train_steps_per_second": 0.037
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,815 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9988571428571429,
|
| 5 |
+
"eval_steps": 50,
|
| 6 |
+
"global_step": 437,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.022857142857142857,
|
| 13 |
+
"grad_norm": 5.484237277918217,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.6541342735290527,
|
| 16 |
+
"logits/rejected": -2.6346869468688965,
|
| 17 |
+
"logps/chosen": -287.93011474609375,
|
| 18 |
+
"logps/rejected": -266.5004577636719,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.40625,
|
| 21 |
+
"rewards/chosen": -0.00015118246665224433,
|
| 22 |
+
"rewards/margins": -3.311875479994342e-05,
|
| 23 |
+
"rewards/rejected": -0.00011806372640421614,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045714285714285714,
|
| 28 |
+
"grad_norm": 5.141117524550972,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.6553006172180176,
|
| 31 |
+
"logits/rejected": -2.623598575592041,
|
| 32 |
+
"logps/chosen": -271.3798828125,
|
| 33 |
+
"logps/rejected": -260.2137145996094,
|
| 34 |
+
"loss": 0.6925,
|
| 35 |
+
"rewards/accuracies": 0.59375,
|
| 36 |
+
"rewards/chosen": 0.003302179742604494,
|
| 37 |
+
"rewards/margins": 0.0022685881704092026,
|
| 38 |
+
"rewards/rejected": 0.0010335915721952915,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06857142857142857,
|
| 43 |
+
"grad_norm": 4.867644605403788,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.6496236324310303,
|
| 46 |
+
"logits/rejected": -2.558840751647949,
|
| 47 |
+
"logps/chosen": -271.86370849609375,
|
| 48 |
+
"logps/rejected": -254.6623077392578,
|
| 49 |
+
"loss": 0.6896,
|
| 50 |
+
"rewards/accuracies": 0.71875,
|
| 51 |
+
"rewards/chosen": 0.020251339301466942,
|
| 52 |
+
"rewards/margins": 0.011756613850593567,
|
| 53 |
+
"rewards/rejected": 0.008494723588228226,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09142857142857143,
|
| 58 |
+
"grad_norm": 4.600987966838872,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.6127796173095703,
|
| 61 |
+
"logits/rejected": -2.5982162952423096,
|
| 62 |
+
"logps/chosen": -261.87078857421875,
|
| 63 |
+
"logps/rejected": -255.64682006835938,
|
| 64 |
+
"loss": 0.6811,
|
| 65 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 66 |
+
"rewards/chosen": 0.039037298411130905,
|
| 67 |
+
"rewards/margins": 0.04095984250307083,
|
| 68 |
+
"rewards/rejected": -0.0019225452560931444,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11428571428571428,
|
| 73 |
+
"grad_norm": 6.01325732118956,
|
| 74 |
+
"learning_rate": 4.997124959943201e-07,
|
| 75 |
+
"logits/chosen": -2.59413743019104,
|
| 76 |
+
"logits/rejected": -2.525812864303589,
|
| 77 |
+
"logps/chosen": -288.6096496582031,
|
| 78 |
+
"logps/rejected": -264.43359375,
|
| 79 |
+
"loss": 0.6661,
|
| 80 |
+
"rewards/accuracies": 0.762499988079071,
|
| 81 |
+
"rewards/chosen": 0.008406409062445164,
|
| 82 |
+
"rewards/margins": 0.11031482368707657,
|
| 83 |
+
"rewards/rejected": -0.10190840810537338,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11428571428571428,
|
| 88 |
+
"eval_logits/chosen": -2.619272232055664,
|
| 89 |
+
"eval_logits/rejected": -2.573228359222412,
|
| 90 |
+
"eval_logps/chosen": -263.4992370605469,
|
| 91 |
+
"eval_logps/rejected": -264.2141418457031,
|
| 92 |
+
"eval_loss": 0.6540683507919312,
|
| 93 |
+
"eval_rewards/accuracies": 0.6551724076271057,
|
| 94 |
+
"eval_rewards/chosen": -0.0353790745139122,
|
| 95 |
+
"eval_rewards/margins": 0.12399081885814667,
|
| 96 |
+
"eval_rewards/rejected": -0.15936988592147827,
|
| 97 |
+
"eval_runtime": 93.0218,
|
| 98 |
+
"eval_samples_per_second": 19.684,
|
| 99 |
+
"eval_steps_per_second": 0.312,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13714285714285715,
|
| 104 |
+
"grad_norm": 8.643694975506126,
|
| 105 |
+
"learning_rate": 4.979579212164186e-07,
|
| 106 |
+
"logits/chosen": -2.6212775707244873,
|
| 107 |
+
"logits/rejected": -2.5676815509796143,
|
| 108 |
+
"logps/chosen": -305.444091796875,
|
| 109 |
+
"logps/rejected": -272.07867431640625,
|
| 110 |
+
"loss": 0.6429,
|
| 111 |
+
"rewards/accuracies": 0.71875,
|
| 112 |
+
"rewards/chosen": -0.05958025902509689,
|
| 113 |
+
"rewards/margins": 0.1966249793767929,
|
| 114 |
+
"rewards/rejected": -0.2562052309513092,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16,
|
| 119 |
+
"grad_norm": 12.272483900335105,
|
| 120 |
+
"learning_rate": 4.946196886175515e-07,
|
| 121 |
+
"logits/chosen": -2.59287166595459,
|
| 122 |
+
"logits/rejected": -2.5041251182556152,
|
| 123 |
+
"logps/chosen": -321.8868103027344,
|
| 124 |
+
"logps/rejected": -300.4162292480469,
|
| 125 |
+
"loss": 0.629,
|
| 126 |
+
"rewards/accuracies": 0.706250011920929,
|
| 127 |
+
"rewards/chosen": -0.18531829118728638,
|
| 128 |
+
"rewards/margins": 0.3314962685108185,
|
| 129 |
+
"rewards/rejected": -0.5168145895004272,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.18285714285714286,
|
| 134 |
+
"grad_norm": 13.506330348124033,
|
| 135 |
+
"learning_rate": 4.897191188239667e-07,
|
| 136 |
+
"logits/chosen": -2.069486141204834,
|
| 137 |
+
"logits/rejected": -1.9483461380004883,
|
| 138 |
+
"logps/chosen": -318.3644714355469,
|
| 139 |
+
"logps/rejected": -329.19659423828125,
|
| 140 |
+
"loss": 0.6107,
|
| 141 |
+
"rewards/accuracies": 0.637499988079071,
|
| 142 |
+
"rewards/chosen": -0.5116854906082153,
|
| 143 |
+
"rewards/margins": 0.31293898820877075,
|
| 144 |
+
"rewards/rejected": -0.8246244192123413,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.2057142857142857,
|
| 149 |
+
"grad_norm": 12.553950954592443,
|
| 150 |
+
"learning_rate": 4.832875107981763e-07,
|
| 151 |
+
"logits/chosen": -1.1866068840026855,
|
| 152 |
+
"logits/rejected": -0.9298005104064941,
|
| 153 |
+
"logps/chosen": -316.0557556152344,
|
| 154 |
+
"logps/rejected": -380.7778625488281,
|
| 155 |
+
"loss": 0.5915,
|
| 156 |
+
"rewards/accuracies": 0.706250011920929,
|
| 157 |
+
"rewards/chosen": -0.7446193695068359,
|
| 158 |
+
"rewards/margins": 0.5451987981796265,
|
| 159 |
+
"rewards/rejected": -1.2898180484771729,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22857142857142856,
|
| 164 |
+
"grad_norm": 30.03309840012713,
|
| 165 |
+
"learning_rate": 4.753659419387223e-07,
|
| 166 |
+
"logits/chosen": -0.9629250764846802,
|
| 167 |
+
"logits/rejected": -0.5570683479309082,
|
| 168 |
+
"logps/chosen": -331.9052429199219,
|
| 169 |
+
"logps/rejected": -379.1312561035156,
|
| 170 |
+
"loss": 0.561,
|
| 171 |
+
"rewards/accuracies": 0.71875,
|
| 172 |
+
"rewards/chosen": -0.7065539360046387,
|
| 173 |
+
"rewards/margins": 0.7784632444381714,
|
| 174 |
+
"rewards/rejected": -1.4850170612335205,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22857142857142856,
|
| 179 |
+
"eval_logits/chosen": -0.4111216068267822,
|
| 180 |
+
"eval_logits/rejected": -0.007617468014359474,
|
| 181 |
+
"eval_logps/chosen": -365.1130676269531,
|
| 182 |
+
"eval_logps/rejected": -431.9446716308594,
|
| 183 |
+
"eval_loss": 0.5662747621536255,
|
| 184 |
+
"eval_rewards/accuracies": 0.7068965435028076,
|
| 185 |
+
"eval_rewards/chosen": -1.0515177249908447,
|
| 186 |
+
"eval_rewards/margins": 0.7851568460464478,
|
| 187 |
+
"eval_rewards/rejected": -1.8366745710372925,
|
| 188 |
+
"eval_runtime": 93.7201,
|
| 189 |
+
"eval_samples_per_second": 19.537,
|
| 190 |
+
"eval_steps_per_second": 0.309,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25142857142857145,
|
| 195 |
+
"grad_norm": 27.312780453008052,
|
| 196 |
+
"learning_rate": 4.660050057270191e-07,
|
| 197 |
+
"logits/chosen": -0.45199713110923767,
|
| 198 |
+
"logits/rejected": 0.33052048087120056,
|
| 199 |
+
"logps/chosen": -373.90545654296875,
|
| 200 |
+
"logps/rejected": -399.00909423828125,
|
| 201 |
+
"loss": 0.5415,
|
| 202 |
+
"rewards/accuracies": 0.793749988079071,
|
| 203 |
+
"rewards/chosen": -0.6764419674873352,
|
| 204 |
+
"rewards/margins": 0.9693374633789062,
|
| 205 |
+
"rewards/rejected": -1.6457793712615967,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.2742857142857143,
|
| 210 |
+
"grad_norm": 29.282942520454633,
|
| 211 |
+
"learning_rate": 4.5526448859687144e-07,
|
| 212 |
+
"logits/chosen": 0.3397678732872009,
|
| 213 |
+
"logits/rejected": 0.8190152049064636,
|
| 214 |
+
"logps/chosen": -406.60565185546875,
|
| 215 |
+
"logps/rejected": -463.072021484375,
|
| 216 |
+
"loss": 0.5753,
|
| 217 |
+
"rewards/accuracies": 0.762499988079071,
|
| 218 |
+
"rewards/chosen": -1.3377411365509033,
|
| 219 |
+
"rewards/margins": 0.7959097623825073,
|
| 220 |
+
"rewards/rejected": -2.133650779724121,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.29714285714285715,
|
| 225 |
+
"grad_norm": 24.36403749878471,
|
| 226 |
+
"learning_rate": 4.432129880904388e-07,
|
| 227 |
+
"logits/chosen": 0.4417182505130768,
|
| 228 |
+
"logits/rejected": 1.1421695947647095,
|
| 229 |
+
"logps/chosen": -371.94744873046875,
|
| 230 |
+
"logps/rejected": -401.11962890625,
|
| 231 |
+
"loss": 0.5483,
|
| 232 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 233 |
+
"rewards/chosen": -0.9846571683883667,
|
| 234 |
+
"rewards/margins": 0.6376141309738159,
|
| 235 |
+
"rewards/rejected": -1.6222712993621826,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.32,
|
| 240 |
+
"grad_norm": 22.333886590434165,
|
| 241 |
+
"learning_rate": 4.299274747394055e-07,
|
| 242 |
+
"logits/chosen": 0.7880607843399048,
|
| 243 |
+
"logits/rejected": 1.3198670148849487,
|
| 244 |
+
"logps/chosen": -435.3951110839844,
|
| 245 |
+
"logps/rejected": -469.02911376953125,
|
| 246 |
+
"loss": 0.549,
|
| 247 |
+
"rewards/accuracies": 0.731249988079071,
|
| 248 |
+
"rewards/chosen": -1.175621747970581,
|
| 249 |
+
"rewards/margins": 0.9118834733963013,
|
| 250 |
+
"rewards/rejected": -2.0875051021575928,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.34285714285714286,
|
| 255 |
+
"grad_norm": 21.900935378631306,
|
| 256 |
+
"learning_rate": 4.1549280046953653e-07,
|
| 257 |
+
"logits/chosen": 1.040090560913086,
|
| 258 |
+
"logits/rejected": 1.5702952146530151,
|
| 259 |
+
"logps/chosen": -405.8429870605469,
|
| 260 |
+
"logps/rejected": -457.5497131347656,
|
| 261 |
+
"loss": 0.5324,
|
| 262 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 263 |
+
"rewards/chosen": -1.0978893041610718,
|
| 264 |
+
"rewards/margins": 0.8218156099319458,
|
| 265 |
+
"rewards/rejected": -1.9197051525115967,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.34285714285714286,
|
| 270 |
+
"eval_logits/chosen": 1.325810432434082,
|
| 271 |
+
"eval_logits/rejected": 2.0029168128967285,
|
| 272 |
+
"eval_logps/chosen": -405.1368103027344,
|
| 273 |
+
"eval_logps/rejected": -492.1299743652344,
|
| 274 |
+
"eval_loss": 0.5436938405036926,
|
| 275 |
+
"eval_rewards/accuracies": 0.6853448152542114,
|
| 276 |
+
"eval_rewards/chosen": -1.451755166053772,
|
| 277 |
+
"eval_rewards/margins": 0.986772894859314,
|
| 278 |
+
"eval_rewards/rejected": -2.438527822494507,
|
| 279 |
+
"eval_runtime": 94.6665,
|
| 280 |
+
"eval_samples_per_second": 19.342,
|
| 281 |
+
"eval_steps_per_second": 0.306,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3657142857142857,
|
| 286 |
+
"grad_norm": 20.99267450356195,
|
| 287 |
+
"learning_rate": 4.000011566683401e-07,
|
| 288 |
+
"logits/chosen": 0.8327977061271667,
|
| 289 |
+
"logits/rejected": 1.764451265335083,
|
| 290 |
+
"logps/chosen": -447.289794921875,
|
| 291 |
+
"logps/rejected": -494.8617248535156,
|
| 292 |
+
"loss": 0.5254,
|
| 293 |
+
"rewards/accuracies": 0.699999988079071,
|
| 294 |
+
"rewards/chosen": -1.381363034248352,
|
| 295 |
+
"rewards/margins": 0.8699455261230469,
|
| 296 |
+
"rewards/rejected": -2.2513084411621094,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38857142857142857,
|
| 301 |
+
"grad_norm": 19.625600234650467,
|
| 302 |
+
"learning_rate": 3.8355148537705047e-07,
|
| 303 |
+
"logits/chosen": 0.17654064297676086,
|
| 304 |
+
"logits/rejected": 1.0507433414459229,
|
| 305 |
+
"logps/chosen": -395.59124755859375,
|
| 306 |
+
"logps/rejected": -480.58038330078125,
|
| 307 |
+
"loss": 0.5315,
|
| 308 |
+
"rewards/accuracies": 0.75,
|
| 309 |
+
"rewards/chosen": -1.0189447402954102,
|
| 310 |
+
"rewards/margins": 1.0480358600616455,
|
| 311 |
+
"rewards/rejected": -2.0669806003570557,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.4114285714285714,
|
| 316 |
+
"grad_norm": 22.131134389700858,
|
| 317 |
+
"learning_rate": 3.662488473675315e-07,
|
| 318 |
+
"logits/chosen": 1.4423513412475586,
|
| 319 |
+
"logits/rejected": 2.64449143409729,
|
| 320 |
+
"logps/chosen": -448.270263671875,
|
| 321 |
+
"logps/rejected": -503.065185546875,
|
| 322 |
+
"loss": 0.5355,
|
| 323 |
+
"rewards/accuracies": 0.75,
|
| 324 |
+
"rewards/chosen": -1.4582993984222412,
|
| 325 |
+
"rewards/margins": 1.1077758073806763,
|
| 326 |
+
"rewards/rejected": -2.566075563430786,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.4342857142857143,
|
| 331 |
+
"grad_norm": 22.169020143018557,
|
| 332 |
+
"learning_rate": 3.48203751140067e-07,
|
| 333 |
+
"logits/chosen": 0.9658209085464478,
|
| 334 |
+
"logits/rejected": 2.3707427978515625,
|
| 335 |
+
"logps/chosen": -405.305419921875,
|
| 336 |
+
"logps/rejected": -484.43499755859375,
|
| 337 |
+
"loss": 0.538,
|
| 338 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 339 |
+
"rewards/chosen": -1.4446977376937866,
|
| 340 |
+
"rewards/margins": 1.050331711769104,
|
| 341 |
+
"rewards/rejected": -2.4950294494628906,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45714285714285713,
|
| 346 |
+
"grad_norm": 18.01149891090628,
|
| 347 |
+
"learning_rate": 3.2953144712759537e-07,
|
| 348 |
+
"logits/chosen": 1.7644532918930054,
|
| 349 |
+
"logits/rejected": 2.6767263412475586,
|
| 350 |
+
"logps/chosen": -395.5130310058594,
|
| 351 |
+
"logps/rejected": -487.91192626953125,
|
| 352 |
+
"loss": 0.5261,
|
| 353 |
+
"rewards/accuracies": 0.737500011920929,
|
| 354 |
+
"rewards/chosen": -1.451765537261963,
|
| 355 |
+
"rewards/margins": 1.0395876169204712,
|
| 356 |
+
"rewards/rejected": -2.4913532733917236,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45714285714285713,
|
| 361 |
+
"eval_logits/chosen": 1.7313151359558105,
|
| 362 |
+
"eval_logits/rejected": 2.7388670444488525,
|
| 363 |
+
"eval_logps/chosen": -416.20770263671875,
|
| 364 |
+
"eval_logps/rejected": -507.4054870605469,
|
| 365 |
+
"eval_loss": 0.5247488021850586,
|
| 366 |
+
"eval_rewards/accuracies": 0.6853448152542114,
|
| 367 |
+
"eval_rewards/chosen": -1.5624639987945557,
|
| 368 |
+
"eval_rewards/margins": 1.0288186073303223,
|
| 369 |
+
"eval_rewards/rejected": -2.591282606124878,
|
| 370 |
+
"eval_runtime": 94.5093,
|
| 371 |
+
"eval_samples_per_second": 19.374,
|
| 372 |
+
"eval_steps_per_second": 0.307,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.48,
|
| 377 |
+
"grad_norm": 19.57404608223085,
|
| 378 |
+
"learning_rate": 3.103511916141658e-07,
|
| 379 |
+
"logits/chosen": 1.504326581954956,
|
| 380 |
+
"logits/rejected": 2.792241334915161,
|
| 381 |
+
"logps/chosen": -433.87969970703125,
|
| 382 |
+
"logps/rejected": -526.0897216796875,
|
| 383 |
+
"loss": 0.517,
|
| 384 |
+
"rewards/accuracies": 0.737500011920929,
|
| 385 |
+
"rewards/chosen": -1.4803601503372192,
|
| 386 |
+
"rewards/margins": 1.1599363088607788,
|
| 387 |
+
"rewards/rejected": -2.640296459197998,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5028571428571429,
|
| 392 |
+
"grad_norm": 21.692320362176854,
|
| 393 |
+
"learning_rate": 2.9078548506882117e-07,
|
| 394 |
+
"logits/chosen": 1.4797990322113037,
|
| 395 |
+
"logits/rejected": 2.450709581375122,
|
| 396 |
+
"logps/chosen": -449.7752990722656,
|
| 397 |
+
"logps/rejected": -549.132080078125,
|
| 398 |
+
"loss": 0.5296,
|
| 399 |
+
"rewards/accuracies": 0.731249988079071,
|
| 400 |
+
"rewards/chosen": -1.4539282321929932,
|
| 401 |
+
"rewards/margins": 1.1360087394714355,
|
| 402 |
+
"rewards/rejected": -2.5899369716644287,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5257142857142857,
|
| 407 |
+
"grad_norm": 19.33385563287397,
|
| 408 |
+
"learning_rate": 2.709592897595191e-07,
|
| 409 |
+
"logits/chosen": 1.3153431415557861,
|
| 410 |
+
"logits/rejected": 2.2327983379364014,
|
| 411 |
+
"logps/chosen": -424.7271423339844,
|
| 412 |
+
"logps/rejected": -506.617919921875,
|
| 413 |
+
"loss": 0.5439,
|
| 414 |
+
"rewards/accuracies": 0.706250011920929,
|
| 415 |
+
"rewards/chosen": -1.3993257284164429,
|
| 416 |
+
"rewards/margins": 0.8849741816520691,
|
| 417 |
+
"rewards/rejected": -2.2843000888824463,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5485714285714286,
|
| 422 |
+
"grad_norm": 19.794275801154928,
|
| 423 |
+
"learning_rate": 2.509992316440332e-07,
|
| 424 |
+
"logits/chosen": 2.2576842308044434,
|
| 425 |
+
"logits/rejected": 3.4659297466278076,
|
| 426 |
+
"logps/chosen": -488.4747619628906,
|
| 427 |
+
"logps/rejected": -558.2715454101562,
|
| 428 |
+
"loss": 0.5332,
|
| 429 |
+
"rewards/accuracies": 0.793749988079071,
|
| 430 |
+
"rewards/chosen": -1.8078649044036865,
|
| 431 |
+
"rewards/margins": 1.1585441827774048,
|
| 432 |
+
"rewards/rejected": -2.966409206390381,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.5714285714285714,
|
| 437 |
+
"grad_norm": 19.209773850229293,
|
| 438 |
+
"learning_rate": 2.3103279163519918e-07,
|
| 439 |
+
"logits/chosen": 1.9105958938598633,
|
| 440 |
+
"logits/rejected": 2.3923182487487793,
|
| 441 |
+
"logps/chosen": -476.888671875,
|
| 442 |
+
"logps/rejected": -540.67724609375,
|
| 443 |
+
"loss": 0.5274,
|
| 444 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 445 |
+
"rewards/chosen": -1.9962562322616577,
|
| 446 |
+
"rewards/margins": 0.6005629301071167,
|
| 447 |
+
"rewards/rejected": -2.5968191623687744,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.5714285714285714,
|
| 452 |
+
"eval_logits/chosen": 1.0144059658050537,
|
| 453 |
+
"eval_logits/rejected": 2.126599073410034,
|
| 454 |
+
"eval_logps/chosen": -428.11065673828125,
|
| 455 |
+
"eval_logps/rejected": -528.8191528320312,
|
| 456 |
+
"eval_loss": 0.514776349067688,
|
| 457 |
+
"eval_rewards/accuracies": 0.7155172228813171,
|
| 458 |
+
"eval_rewards/chosen": -1.6814932823181152,
|
| 459 |
+
"eval_rewards/margins": 1.123926043510437,
|
| 460 |
+
"eval_rewards/rejected": -2.8054192066192627,
|
| 461 |
+
"eval_runtime": 95.0873,
|
| 462 |
+
"eval_samples_per_second": 19.256,
|
| 463 |
+
"eval_steps_per_second": 0.305,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5942857142857143,
|
| 468 |
+
"grad_norm": 21.990080244291757,
|
| 469 |
+
"learning_rate": 2.1118749140573358e-07,
|
| 470 |
+
"logits/chosen": 1.3521721363067627,
|
| 471 |
+
"logits/rejected": 2.6136393547058105,
|
| 472 |
+
"logps/chosen": -439.0127868652344,
|
| 473 |
+
"logps/rejected": -476.6556701660156,
|
| 474 |
+
"loss": 0.5238,
|
| 475 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 476 |
+
"rewards/chosen": -1.6318817138671875,
|
| 477 |
+
"rewards/margins": 0.9099860191345215,
|
| 478 |
+
"rewards/rejected": -2.54186749458313,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6171428571428571,
|
| 483 |
+
"grad_norm": 21.20271695558916,
|
| 484 |
+
"learning_rate": 1.9159007893272703e-07,
|
| 485 |
+
"logits/chosen": 1.4235649108886719,
|
| 486 |
+
"logits/rejected": 2.510680913925171,
|
| 487 |
+
"logps/chosen": -468.30975341796875,
|
| 488 |
+
"logps/rejected": -496.22406005859375,
|
| 489 |
+
"loss": 0.5192,
|
| 490 |
+
"rewards/accuracies": 0.65625,
|
| 491 |
+
"rewards/chosen": -1.8018286228179932,
|
| 492 |
+
"rewards/margins": 0.6999757885932922,
|
| 493 |
+
"rewards/rejected": -2.5018038749694824,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.64,
|
| 498 |
+
"grad_norm": 18.377021107490634,
|
| 499 |
+
"learning_rate": 1.7236571898357766e-07,
|
| 500 |
+
"logits/chosen": 1.396390676498413,
|
| 501 |
+
"logits/rejected": 2.4561963081359863,
|
| 502 |
+
"logps/chosen": -388.86273193359375,
|
| 503 |
+
"logps/rejected": -491.08612060546875,
|
| 504 |
+
"loss": 0.5273,
|
| 505 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 506 |
+
"rewards/chosen": -1.3913507461547852,
|
| 507 |
+
"rewards/margins": 1.1841250658035278,
|
| 508 |
+
"rewards/rejected": -2.5754759311676025,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6628571428571428,
|
| 513 |
+
"grad_norm": 23.869092599907706,
|
| 514 |
+
"learning_rate": 1.5363719371356882e-07,
|
| 515 |
+
"logits/chosen": 1.4888983964920044,
|
| 516 |
+
"logits/rejected": 2.697493076324463,
|
| 517 |
+
"logps/chosen": -438.34613037109375,
|
| 518 |
+
"logps/rejected": -505.64398193359375,
|
| 519 |
+
"loss": 0.5086,
|
| 520 |
+
"rewards/accuracies": 0.737500011920929,
|
| 521 |
+
"rewards/chosen": -1.5241912603378296,
|
| 522 |
+
"rewards/margins": 1.000713586807251,
|
| 523 |
+
"rewards/rejected": -2.524904727935791,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6857142857142857,
|
| 528 |
+
"grad_norm": 25.027316619015263,
|
| 529 |
+
"learning_rate": 1.3552411848071565e-07,
|
| 530 |
+
"logits/chosen": 1.6033051013946533,
|
| 531 |
+
"logits/rejected": 3.075709581375122,
|
| 532 |
+
"logps/chosen": -483.4491271972656,
|
| 533 |
+
"logps/rejected": -563.5081787109375,
|
| 534 |
+
"loss": 0.5,
|
| 535 |
+
"rewards/accuracies": 0.737500011920929,
|
| 536 |
+
"rewards/chosen": -1.700990080833435,
|
| 537 |
+
"rewards/margins": 1.2899569272994995,
|
| 538 |
+
"rewards/rejected": -2.9909472465515137,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6857142857142857,
|
| 543 |
+
"eval_logits/chosen": 1.5160114765167236,
|
| 544 |
+
"eval_logits/rejected": 2.702754259109497,
|
| 545 |
+
"eval_logps/chosen": -428.75518798828125,
|
| 546 |
+
"eval_logps/rejected": -535.8169555664062,
|
| 547 |
+
"eval_loss": 0.5077716112136841,
|
| 548 |
+
"eval_rewards/accuracies": 0.7198275923728943,
|
| 549 |
+
"eval_rewards/chosen": -1.6879388093948364,
|
| 550 |
+
"eval_rewards/margins": 1.187458872795105,
|
| 551 |
+
"eval_rewards/rejected": -2.8753974437713623,
|
| 552 |
+
"eval_runtime": 93.8182,
|
| 553 |
+
"eval_samples_per_second": 19.516,
|
| 554 |
+
"eval_steps_per_second": 0.309,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7085714285714285,
|
| 559 |
+
"grad_norm": 20.06935454117094,
|
| 560 |
+
"learning_rate": 1.1814217788631473e-07,
|
| 561 |
+
"logits/chosen": 1.097944974899292,
|
| 562 |
+
"logits/rejected": 2.714242458343506,
|
| 563 |
+
"logps/chosen": -464.17645263671875,
|
| 564 |
+
"logps/rejected": -550.5721435546875,
|
| 565 |
+
"loss": 0.512,
|
| 566 |
+
"rewards/accuracies": 0.762499988079071,
|
| 567 |
+
"rewards/chosen": -1.509181261062622,
|
| 568 |
+
"rewards/margins": 1.3204662799835205,
|
| 569 |
+
"rewards/rejected": -2.8296473026275635,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7314285714285714,
|
| 574 |
+
"grad_norm": 20.37966135449331,
|
| 575 |
+
"learning_rate": 1.0160238692045331e-07,
|
| 576 |
+
"logits/chosen": 1.1582145690917969,
|
| 577 |
+
"logits/rejected": 2.5812854766845703,
|
| 578 |
+
"logps/chosen": -450.8458557128906,
|
| 579 |
+
"logps/rejected": -533.5846557617188,
|
| 580 |
+
"loss": 0.5203,
|
| 581 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 582 |
+
"rewards/chosen": -1.5372800827026367,
|
| 583 |
+
"rewards/margins": 1.0487934350967407,
|
| 584 |
+
"rewards/rejected": -2.586073637008667,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7542857142857143,
|
| 589 |
+
"grad_norm": 20.9552841041291,
|
| 590 |
+
"learning_rate": 8.601038193139438e-08,
|
| 591 |
+
"logits/chosen": 1.9143211841583252,
|
| 592 |
+
"logits/rejected": 2.735372304916382,
|
| 593 |
+
"logps/chosen": -400.63372802734375,
|
| 594 |
+
"logps/rejected": -489.5486755371094,
|
| 595 |
+
"loss": 0.52,
|
| 596 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 597 |
+
"rewards/chosen": -1.5494529008865356,
|
| 598 |
+
"rewards/margins": 0.9419981241226196,
|
| 599 |
+
"rewards/rejected": -2.491450786590576,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7771428571428571,
|
| 604 |
+
"grad_norm": 20.770737913391446,
|
| 605 |
+
"learning_rate": 7.146574594727572e-08,
|
| 606 |
+
"logits/chosen": 1.9478594064712524,
|
| 607 |
+
"logits/rejected": 3.201336622238159,
|
| 608 |
+
"logps/chosen": -425.11126708984375,
|
| 609 |
+
"logps/rejected": -508.321533203125,
|
| 610 |
+
"loss": 0.5043,
|
| 611 |
+
"rewards/accuracies": 0.762499988079071,
|
| 612 |
+
"rewards/chosen": -1.7910737991333008,
|
| 613 |
+
"rewards/margins": 1.1562799215316772,
|
| 614 |
+
"rewards/rejected": -2.947354316711426,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8,
|
| 619 |
+
"grad_norm": 23.43196809715842,
|
| 620 |
+
"learning_rate": 5.8061372659157306e-08,
|
| 621 |
+
"logits/chosen": 2.0667195320129395,
|
| 622 |
+
"logits/rejected": 3.518019914627075,
|
| 623 |
+
"logps/chosen": -500.10174560546875,
|
| 624 |
+
"logps/rejected": -575.167236328125,
|
| 625 |
+
"loss": 0.4879,
|
| 626 |
+
"rewards/accuracies": 0.737500011920929,
|
| 627 |
+
"rewards/chosen": -1.8992862701416016,
|
| 628 |
+
"rewards/margins": 1.168575406074524,
|
| 629 |
+
"rewards/rejected": -3.067861557006836,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8,
|
| 634 |
+
"eval_logits/chosen": 2.206516742706299,
|
| 635 |
+
"eval_logits/rejected": 3.2477219104766846,
|
| 636 |
+
"eval_logps/chosen": -448.67852783203125,
|
| 637 |
+
"eval_logps/rejected": -555.7252197265625,
|
| 638 |
+
"eval_loss": 0.5050091743469238,
|
| 639 |
+
"eval_rewards/accuracies": 0.7198275923728943,
|
| 640 |
+
"eval_rewards/chosen": -1.8871723413467407,
|
| 641 |
+
"eval_rewards/margins": 1.187308430671692,
|
| 642 |
+
"eval_rewards/rejected": -3.0744810104370117,
|
| 643 |
+
"eval_runtime": 93.1425,
|
| 644 |
+
"eval_samples_per_second": 19.658,
|
| 645 |
+
"eval_steps_per_second": 0.311,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8228571428571428,
|
| 650 |
+
"grad_norm": 21.572101434555105,
|
| 651 |
+
"learning_rate": 4.5882873127531614e-08,
|
| 652 |
+
"logits/chosen": 2.518174648284912,
|
| 653 |
+
"logits/rejected": 3.5241615772247314,
|
| 654 |
+
"logps/chosen": -472.5306701660156,
|
| 655 |
+
"logps/rejected": -603.8782958984375,
|
| 656 |
+
"loss": 0.5056,
|
| 657 |
+
"rewards/accuracies": 0.8125,
|
| 658 |
+
"rewards/chosen": -1.8001983165740967,
|
| 659 |
+
"rewards/margins": 1.5649583339691162,
|
| 660 |
+
"rewards/rejected": -3.365156650543213,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8457142857142858,
|
| 665 |
+
"grad_norm": 24.32119336663692,
|
| 666 |
+
"learning_rate": 3.500802900154412e-08,
|
| 667 |
+
"logits/chosen": 2.2133302688598633,
|
| 668 |
+
"logits/rejected": 2.970715284347534,
|
| 669 |
+
"logps/chosen": -420.18707275390625,
|
| 670 |
+
"logps/rejected": -547.6619873046875,
|
| 671 |
+
"loss": 0.5034,
|
| 672 |
+
"rewards/accuracies": 0.762499988079071,
|
| 673 |
+
"rewards/chosen": -1.6418088674545288,
|
| 674 |
+
"rewards/margins": 1.213688611984253,
|
| 675 |
+
"rewards/rejected": -2.855497360229492,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8685714285714285,
|
| 680 |
+
"grad_norm": 25.458784716444757,
|
| 681 |
+
"learning_rate": 2.550629574310309e-08,
|
| 682 |
+
"logits/chosen": 2.7819252014160156,
|
| 683 |
+
"logits/rejected": 3.5777878761291504,
|
| 684 |
+
"logps/chosen": -435.19268798828125,
|
| 685 |
+
"logps/rejected": -533.6688842773438,
|
| 686 |
+
"loss": 0.4987,
|
| 687 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 688 |
+
"rewards/chosen": -1.7933107614517212,
|
| 689 |
+
"rewards/margins": 0.9658328890800476,
|
| 690 |
+
"rewards/rejected": -2.759143352508545,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8914285714285715,
|
| 695 |
+
"grad_norm": 20.416342139746472,
|
| 696 |
+
"learning_rate": 1.7438359028687983e-08,
|
| 697 |
+
"logits/chosen": 2.2803592681884766,
|
| 698 |
+
"logits/rejected": 3.287682056427002,
|
| 699 |
+
"logps/chosen": -436.3902282714844,
|
| 700 |
+
"logps/rejected": -546.454833984375,
|
| 701 |
+
"loss": 0.5096,
|
| 702 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 703 |
+
"rewards/chosen": -1.6981942653656006,
|
| 704 |
+
"rewards/margins": 1.2526156902313232,
|
| 705 |
+
"rewards/rejected": -2.950809955596924,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9142857142857143,
|
| 710 |
+
"grad_norm": 23.31604206083898,
|
| 711 |
+
"learning_rate": 1.0855747162029361e-08,
|
| 712 |
+
"logits/chosen": 1.8559446334838867,
|
| 713 |
+
"logits/rejected": 2.6898691654205322,
|
| 714 |
+
"logps/chosen": -443.5389709472656,
|
| 715 |
+
"logps/rejected": -552.5650634765625,
|
| 716 |
+
"loss": 0.5082,
|
| 717 |
+
"rewards/accuracies": 0.75,
|
| 718 |
+
"rewards/chosen": -1.6940586566925049,
|
| 719 |
+
"rewards/margins": 1.158569574356079,
|
| 720 |
+
"rewards/rejected": -2.852627992630005,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9142857142857143,
|
| 725 |
+
"eval_logits/chosen": 1.9658149480819702,
|
| 726 |
+
"eval_logits/rejected": 3.059574842453003,
|
| 727 |
+
"eval_logps/chosen": -435.48773193359375,
|
| 728 |
+
"eval_logps/rejected": -548.0841064453125,
|
| 729 |
+
"eval_loss": 0.5012447834014893,
|
| 730 |
+
"eval_rewards/accuracies": 0.7198275923728943,
|
| 731 |
+
"eval_rewards/chosen": -1.7552645206451416,
|
| 732 |
+
"eval_rewards/margins": 1.2428052425384521,
|
| 733 |
+
"eval_rewards/rejected": -2.9980695247650146,
|
| 734 |
+
"eval_runtime": 94.0643,
|
| 735 |
+
"eval_samples_per_second": 19.465,
|
| 736 |
+
"eval_steps_per_second": 0.308,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9371428571428572,
|
| 741 |
+
"grad_norm": 19.227803698254093,
|
| 742 |
+
"learning_rate": 5.8005019731033615e-09,
|
| 743 |
+
"logits/chosen": 2.1140646934509277,
|
| 744 |
+
"logits/rejected": 3.0444817543029785,
|
| 745 |
+
"logps/chosen": -424.0716247558594,
|
| 746 |
+
"logps/rejected": -529.0479736328125,
|
| 747 |
+
"loss": 0.4899,
|
| 748 |
+
"rewards/accuracies": 0.768750011920929,
|
| 749 |
+
"rewards/chosen": -1.6376755237579346,
|
| 750 |
+
"rewards/margins": 1.076155662536621,
|
| 751 |
+
"rewards/rejected": -2.7138311862945557,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.96,
|
| 756 |
+
"grad_norm": 19.178542640219938,
|
| 757 |
+
"learning_rate": 2.3049103053431886e-09,
|
| 758 |
+
"logits/chosen": 1.7973829507827759,
|
| 759 |
+
"logits/rejected": 2.994089365005493,
|
| 760 |
+
"logps/chosen": -480.82342529296875,
|
| 761 |
+
"logps/rejected": -606.6632690429688,
|
| 762 |
+
"loss": 0.5082,
|
| 763 |
+
"rewards/accuracies": 0.737500011920929,
|
| 764 |
+
"rewards/chosen": -1.7015644311904907,
|
| 765 |
+
"rewards/margins": 1.422523856163025,
|
| 766 |
+
"rewards/rejected": -3.1240882873535156,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9828571428571429,
|
| 771 |
+
"grad_norm": 23.036308488754496,
|
| 772 |
+
"learning_rate": 3.9129780600541397e-10,
|
| 773 |
+
"logits/chosen": 2.1120381355285645,
|
| 774 |
+
"logits/rejected": 3.065523862838745,
|
| 775 |
+
"logps/chosen": -451.97515869140625,
|
| 776 |
+
"logps/rejected": -554.6312866210938,
|
| 777 |
+
"loss": 0.4855,
|
| 778 |
+
"rewards/accuracies": 0.78125,
|
| 779 |
+
"rewards/chosen": -1.6638119220733643,
|
| 780 |
+
"rewards/margins": 1.303351640701294,
|
| 781 |
+
"rewards/rejected": -2.967163324356079,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 0.9988571428571429,
|
| 786 |
+
"step": 437,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.5499054316252265,
|
| 789 |
+
"train_runtime": 11738.6737,
|
| 790 |
+
"train_samples_per_second": 4.77,
|
| 791 |
+
"train_steps_per_second": 0.037
|
| 792 |
+
}
|
| 793 |
+
],
|
| 794 |
+
"logging_steps": 10,
|
| 795 |
+
"max_steps": 437,
|
| 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 |
+
}
|