Model save
Browse files- README.md +83 -0
- all_results.json +9 -0
- train_results.json +9 -0
- trainer_state.json +1731 -0
README.md
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: princeton-nlp/Llama-3-Base-8B-SFT
|
| 3 |
+
library_name: peft
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
model-index:
|
| 9 |
+
- name: llama3-wpo-lora
|
| 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 |
+
# llama3-wpo-lora
|
| 17 |
+
|
| 18 |
+
This model is a fine-tuned version of [princeton-nlp/Llama-3-Base-8B-SFT](https://huggingface.co/princeton-nlp/Llama-3-Base-8B-SFT) on the None dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.5134
|
| 21 |
+
- Rewards/chosen: -0.2023
|
| 22 |
+
- Rewards/rejected: -1.1119
|
| 23 |
+
- Rewards/accuracies: 0.7480
|
| 24 |
+
- Rewards/margins: 0.9095
|
| 25 |
+
- Logps/rejected: -287.7953
|
| 26 |
+
- Logps/chosen: -294.5704
|
| 27 |
+
- Logps/ref Response: -0.5364
|
| 28 |
+
- Logits/rejected: -0.1602
|
| 29 |
+
- Logits/chosen: -0.2100
|
| 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: 5e-06
|
| 49 |
+
- train_batch_size: 1
|
| 50 |
+
- eval_batch_size: 4
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 4
|
| 54 |
+
- gradient_accumulation_steps: 16
|
| 55 |
+
- total_train_batch_size: 64
|
| 56 |
+
- total_eval_batch_size: 16
|
| 57 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
+
- lr_scheduler_type: cosine
|
| 59 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
+
- num_epochs: 1
|
| 61 |
+
|
| 62 |
+
### Training results
|
| 63 |
+
|
| 64 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logps/ref Response | Logits/rejected | Logits/chosen |
|
| 65 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:------------------:|:---------------:|:-------------:|
|
| 66 |
+
| 0.6044 | 0.1047 | 100 | 0.5889 | 0.1186 | -0.2672 | 0.6840 | 0.3859 | -279.3490 | -291.3607 | -0.5364 | -0.5369 | -0.5447 |
|
| 67 |
+
| 0.5438 | 0.2094 | 200 | 0.5452 | 0.0540 | -0.6279 | 0.7180 | 0.6819 | -282.9556 | -292.0069 | -0.5364 | -0.4631 | -0.4851 |
|
| 68 |
+
| 0.5367 | 0.3141 | 300 | 0.5323 | -0.0871 | -0.8542 | 0.7240 | 0.7671 | -285.2182 | -293.4178 | -0.5364 | -0.3777 | -0.4077 |
|
| 69 |
+
| 0.5196 | 0.4187 | 400 | 0.5236 | -0.0378 | -0.8614 | 0.7320 | 0.8235 | -285.2903 | -292.9255 | -0.5364 | -0.2899 | -0.3281 |
|
| 70 |
+
| 0.509 | 0.5234 | 500 | 0.5185 | -0.2693 | -1.1302 | 0.7360 | 0.8610 | -287.9790 | -295.2397 | -0.5364 | -0.2296 | -0.2739 |
|
| 71 |
+
| 0.5012 | 0.6281 | 600 | 0.5152 | -0.3520 | -1.2471 | 0.7480 | 0.8951 | -289.1475 | -296.0675 | -0.5364 | -0.1926 | -0.2397 |
|
| 72 |
+
| 0.5168 | 0.7328 | 700 | 0.5139 | -0.2521 | -1.1562 | 0.7440 | 0.9041 | -288.2387 | -295.0681 | -0.5364 | -0.1665 | -0.2158 |
|
| 73 |
+
| 0.5156 | 0.8375 | 800 | 0.5135 | -0.2204 | -1.1304 | 0.7520 | 0.9099 | -287.9801 | -294.7516 | -0.5364 | -0.1603 | -0.2103 |
|
| 74 |
+
| 0.506 | 0.9422 | 900 | 0.5134 | -0.2023 | -1.1119 | 0.7480 | 0.9095 | -287.7953 | -294.5704 | -0.5364 | -0.1602 | -0.2100 |
|
| 75 |
+
|
| 76 |
+
|
| 77 |
+
### Framework versions
|
| 78 |
+
|
| 79 |
+
- PEFT 0.7.1
|
| 80 |
+
- Transformers 4.44.2
|
| 81 |
+
- Pytorch 2.2.1+cu121
|
| 82 |
+
- Datasets 2.14.6
|
| 83 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9997382884061764,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5342667900454936,
|
| 5 |
+
"train_runtime": 19113.1655,
|
| 6 |
+
"train_samples": 61135,
|
| 7 |
+
"train_samples_per_second": 3.199,
|
| 8 |
+
"train_steps_per_second": 0.05
|
| 9 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9997382884061764,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5342667900454936,
|
| 5 |
+
"train_runtime": 19113.1655,
|
| 6 |
+
"train_samples": 61135,
|
| 7 |
+
"train_samples_per_second": 3.199,
|
| 8 |
+
"train_steps_per_second": 0.05
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1731 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9997382884061764,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 955,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0010468463752944255,
|
| 13 |
+
"grad_norm": 4.03125,
|
| 14 |
+
"learning_rate": 5.208333333333333e-08,
|
| 15 |
+
"logits/chosen": -0.3494967222213745,
|
| 16 |
+
"logits/rejected": -0.3728627860546112,
|
| 17 |
+
"logps/chosen": -285.8127136230469,
|
| 18 |
+
"logps/ref_response": -0.3494967222213745,
|
| 19 |
+
"logps/rejected": -212.7957000732422,
|
| 20 |
+
"loss": 0.6931,
|
| 21 |
+
"rewards/accuracies": 0.0,
|
| 22 |
+
"rewards/chosen": 0.0,
|
| 23 |
+
"rewards/margins": 0.0,
|
| 24 |
+
"rewards/rejected": 0.0,
|
| 25 |
+
"step": 1
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 0.010468463752944255,
|
| 29 |
+
"grad_norm": 3.8125,
|
| 30 |
+
"learning_rate": 5.208333333333334e-07,
|
| 31 |
+
"logits/chosen": -0.5401131510734558,
|
| 32 |
+
"logits/rejected": -0.5498467683792114,
|
| 33 |
+
"logps/chosen": -315.3433532714844,
|
| 34 |
+
"logps/ref_response": -0.5399107336997986,
|
| 35 |
+
"logps/rejected": -278.06756591796875,
|
| 36 |
+
"loss": 0.6924,
|
| 37 |
+
"rewards/accuracies": 0.4444444477558136,
|
| 38 |
+
"rewards/chosen": -0.0011721360497176647,
|
| 39 |
+
"rewards/margins": 0.004719285294413567,
|
| 40 |
+
"rewards/rejected": -0.005891421809792519,
|
| 41 |
+
"step": 10
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"epoch": 0.02093692750588851,
|
| 45 |
+
"grad_norm": 3.859375,
|
| 46 |
+
"learning_rate": 1.0416666666666667e-06,
|
| 47 |
+
"logits/chosen": -0.5040869116783142,
|
| 48 |
+
"logits/rejected": -0.5244153738021851,
|
| 49 |
+
"logps/chosen": -306.72930908203125,
|
| 50 |
+
"logps/ref_response": -0.5032420754432678,
|
| 51 |
+
"logps/rejected": -271.22784423828125,
|
| 52 |
+
"loss": 0.6921,
|
| 53 |
+
"rewards/accuracies": 0.512499988079071,
|
| 54 |
+
"rewards/chosen": 0.004430481232702732,
|
| 55 |
+
"rewards/margins": 0.005479422397911549,
|
| 56 |
+
"rewards/rejected": -0.0010489404667168856,
|
| 57 |
+
"step": 20
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"epoch": 0.031405391258832765,
|
| 61 |
+
"grad_norm": 3.875,
|
| 62 |
+
"learning_rate": 1.5625e-06,
|
| 63 |
+
"logits/chosen": -0.5105286240577698,
|
| 64 |
+
"logits/rejected": -0.5181563496589661,
|
| 65 |
+
"logps/chosen": -290.9847717285156,
|
| 66 |
+
"logps/ref_response": -0.5080639123916626,
|
| 67 |
+
"logps/rejected": -252.4471435546875,
|
| 68 |
+
"loss": 0.6875,
|
| 69 |
+
"rewards/accuracies": 0.59375,
|
| 70 |
+
"rewards/chosen": 0.018009770661592484,
|
| 71 |
+
"rewards/margins": 0.021275093778967857,
|
| 72 |
+
"rewards/rejected": -0.0032653254456818104,
|
| 73 |
+
"step": 30
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"epoch": 0.04187385501177702,
|
| 77 |
+
"grad_norm": 3.25,
|
| 78 |
+
"learning_rate": 2.0833333333333334e-06,
|
| 79 |
+
"logits/chosen": -0.48318833112716675,
|
| 80 |
+
"logits/rejected": -0.5184761881828308,
|
| 81 |
+
"logps/chosen": -305.87347412109375,
|
| 82 |
+
"logps/ref_response": -0.47757530212402344,
|
| 83 |
+
"logps/rejected": -244.558349609375,
|
| 84 |
+
"loss": 0.6771,
|
| 85 |
+
"rewards/accuracies": 0.612500011920929,
|
| 86 |
+
"rewards/chosen": 0.04270617291331291,
|
| 87 |
+
"rewards/margins": 0.042043447494506836,
|
| 88 |
+
"rewards/rejected": 0.0006627263501286507,
|
| 89 |
+
"step": 40
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"epoch": 0.05234231876472128,
|
| 93 |
+
"grad_norm": 2.09375,
|
| 94 |
+
"learning_rate": 2.604166666666667e-06,
|
| 95 |
+
"logits/chosen": -0.5472795963287354,
|
| 96 |
+
"logits/rejected": -0.575782060623169,
|
| 97 |
+
"logps/chosen": -304.7160339355469,
|
| 98 |
+
"logps/ref_response": -0.5367640256881714,
|
| 99 |
+
"logps/rejected": -282.7024841308594,
|
| 100 |
+
"loss": 0.6697,
|
| 101 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 102 |
+
"rewards/chosen": 0.0860854759812355,
|
| 103 |
+
"rewards/margins": 0.049294885247945786,
|
| 104 |
+
"rewards/rejected": 0.03679059445858002,
|
| 105 |
+
"step": 50
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"epoch": 0.06281078251766553,
|
| 109 |
+
"grad_norm": 2.875,
|
| 110 |
+
"learning_rate": 3.125e-06,
|
| 111 |
+
"logits/chosen": -0.5696572661399841,
|
| 112 |
+
"logits/rejected": -0.5703103542327881,
|
| 113 |
+
"logps/chosen": -290.3211975097656,
|
| 114 |
+
"logps/ref_response": -0.5527787804603577,
|
| 115 |
+
"logps/rejected": -254.42190551757812,
|
| 116 |
+
"loss": 0.6511,
|
| 117 |
+
"rewards/accuracies": 0.574999988079071,
|
| 118 |
+
"rewards/chosen": 0.13902851939201355,
|
| 119 |
+
"rewards/margins": 0.06114862486720085,
|
| 120 |
+
"rewards/rejected": 0.0778798907995224,
|
| 121 |
+
"step": 60
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.07327924627060979,
|
| 125 |
+
"grad_norm": 2.9375,
|
| 126 |
+
"learning_rate": 3.6458333333333333e-06,
|
| 127 |
+
"logits/chosen": -0.559634268283844,
|
| 128 |
+
"logits/rejected": -0.5745820999145508,
|
| 129 |
+
"logps/chosen": -285.9539489746094,
|
| 130 |
+
"logps/ref_response": -0.5369429588317871,
|
| 131 |
+
"logps/rejected": -263.0733947753906,
|
| 132 |
+
"loss": 0.6325,
|
| 133 |
+
"rewards/accuracies": 0.6875,
|
| 134 |
+
"rewards/chosen": 0.234585240483284,
|
| 135 |
+
"rewards/margins": 0.16143682599067688,
|
| 136 |
+
"rewards/rejected": 0.07314838469028473,
|
| 137 |
+
"step": 70
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 0.08374771002355404,
|
| 141 |
+
"grad_norm": 2.765625,
|
| 142 |
+
"learning_rate": 4.166666666666667e-06,
|
| 143 |
+
"logits/chosen": -0.4994782507419586,
|
| 144 |
+
"logits/rejected": -0.5257306694984436,
|
| 145 |
+
"logps/chosen": -287.14178466796875,
|
| 146 |
+
"logps/ref_response": -0.46965378522872925,
|
| 147 |
+
"logps/rejected": -273.8600158691406,
|
| 148 |
+
"loss": 0.6093,
|
| 149 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 150 |
+
"rewards/chosen": 0.3439296782016754,
|
| 151 |
+
"rewards/margins": 0.2807127833366394,
|
| 152 |
+
"rewards/rejected": 0.06321687251329422,
|
| 153 |
+
"step": 80
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"epoch": 0.0942161737764983,
|
| 157 |
+
"grad_norm": 2.65625,
|
| 158 |
+
"learning_rate": 4.6875000000000004e-06,
|
| 159 |
+
"logits/chosen": -0.530420184135437,
|
| 160 |
+
"logits/rejected": -0.5506101846694946,
|
| 161 |
+
"logps/chosen": -330.380615234375,
|
| 162 |
+
"logps/ref_response": -0.4922845959663391,
|
| 163 |
+
"logps/rejected": -296.19439697265625,
|
| 164 |
+
"loss": 0.5874,
|
| 165 |
+
"rewards/accuracies": 0.75,
|
| 166 |
+
"rewards/chosen": 0.27289265394210815,
|
| 167 |
+
"rewards/margins": 0.35931333899497986,
|
| 168 |
+
"rewards/rejected": -0.0864206999540329,
|
| 169 |
+
"step": 90
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"epoch": 0.10468463752944256,
|
| 173 |
+
"grad_norm": 2.671875,
|
| 174 |
+
"learning_rate": 4.9997324926814375e-06,
|
| 175 |
+
"logits/chosen": -0.5689066648483276,
|
| 176 |
+
"logits/rejected": -0.5622434020042419,
|
| 177 |
+
"logps/chosen": -276.24310302734375,
|
| 178 |
+
"logps/ref_response": -0.533843994140625,
|
| 179 |
+
"logps/rejected": -291.2969970703125,
|
| 180 |
+
"loss": 0.6044,
|
| 181 |
+
"rewards/accuracies": 0.699999988079071,
|
| 182 |
+
"rewards/chosen": 0.2586072087287903,
|
| 183 |
+
"rewards/margins": 0.3670012056827545,
|
| 184 |
+
"rewards/rejected": -0.10839401185512543,
|
| 185 |
+
"step": 100
|
| 186 |
+
},
|
| 187 |
+
{
|
| 188 |
+
"epoch": 0.10468463752944256,
|
| 189 |
+
"eval_logits/chosen": -0.5446628332138062,
|
| 190 |
+
"eval_logits/rejected": -0.5368726253509521,
|
| 191 |
+
"eval_logps/chosen": -291.36065673828125,
|
| 192 |
+
"eval_logps/ref_response": -0.536393404006958,
|
| 193 |
+
"eval_logps/rejected": -279.3489685058594,
|
| 194 |
+
"eval_loss": 0.5888689160346985,
|
| 195 |
+
"eval_rewards/accuracies": 0.6840000152587891,
|
| 196 |
+
"eval_rewards/chosen": 0.11864880472421646,
|
| 197 |
+
"eval_rewards/margins": 0.38588646054267883,
|
| 198 |
+
"eval_rewards/rejected": -0.26723766326904297,
|
| 199 |
+
"eval_runtime": 351.4888,
|
| 200 |
+
"eval_samples_per_second": 5.69,
|
| 201 |
+
"eval_steps_per_second": 0.356,
|
| 202 |
+
"step": 100
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"epoch": 0.11515310128238682,
|
| 206 |
+
"grad_norm": 2.1875,
|
| 207 |
+
"learning_rate": 4.996723692767927e-06,
|
| 208 |
+
"logits/chosen": -0.6064401865005493,
|
| 209 |
+
"logits/rejected": -0.6292127966880798,
|
| 210 |
+
"logps/chosen": -290.3874206542969,
|
| 211 |
+
"logps/ref_response": -0.5667906999588013,
|
| 212 |
+
"logps/rejected": -279.1947937011719,
|
| 213 |
+
"loss": 0.5769,
|
| 214 |
+
"rewards/accuracies": 0.75,
|
| 215 |
+
"rewards/chosen": 0.12466597557067871,
|
| 216 |
+
"rewards/margins": 0.48810654878616333,
|
| 217 |
+
"rewards/rejected": -0.3634406626224518,
|
| 218 |
+
"step": 110
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"epoch": 0.12562156503533106,
|
| 222 |
+
"grad_norm": 2.328125,
|
| 223 |
+
"learning_rate": 4.9903757462135984e-06,
|
| 224 |
+
"logits/chosen": -0.5538973808288574,
|
| 225 |
+
"logits/rejected": -0.563797652721405,
|
| 226 |
+
"logps/chosen": -263.09698486328125,
|
| 227 |
+
"logps/ref_response": -0.5169209837913513,
|
| 228 |
+
"logps/rejected": -255.43795776367188,
|
| 229 |
+
"loss": 0.5649,
|
| 230 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 231 |
+
"rewards/chosen": -0.05361238867044449,
|
| 232 |
+
"rewards/margins": 0.42452582716941833,
|
| 233 |
+
"rewards/rejected": -0.47813814878463745,
|
| 234 |
+
"step": 120
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"epoch": 0.1360900287882753,
|
| 238 |
+
"grad_norm": 2.390625,
|
| 239 |
+
"learning_rate": 4.980697142834315e-06,
|
| 240 |
+
"logits/chosen": -0.5228760838508606,
|
| 241 |
+
"logits/rejected": -0.539161205291748,
|
| 242 |
+
"logps/chosen": -303.0,
|
| 243 |
+
"logps/ref_response": -0.4790240228176117,
|
| 244 |
+
"logps/rejected": -340.1643981933594,
|
| 245 |
+
"loss": 0.5679,
|
| 246 |
+
"rewards/accuracies": 0.699999988079071,
|
| 247 |
+
"rewards/chosen": -0.004994773771613836,
|
| 248 |
+
"rewards/margins": 0.45376062393188477,
|
| 249 |
+
"rewards/rejected": -0.45875534415245056,
|
| 250 |
+
"step": 130
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.14655849254121958,
|
| 254 |
+
"grad_norm": 2.09375,
|
| 255 |
+
"learning_rate": 4.967700826904229e-06,
|
| 256 |
+
"logits/chosen": -0.5957541465759277,
|
| 257 |
+
"logits/rejected": -0.6013976335525513,
|
| 258 |
+
"logps/chosen": -283.49285888671875,
|
| 259 |
+
"logps/ref_response": -0.5482783913612366,
|
| 260 |
+
"logps/rejected": -278.0264587402344,
|
| 261 |
+
"loss": 0.5477,
|
| 262 |
+
"rewards/accuracies": 0.699999988079071,
|
| 263 |
+
"rewards/chosen": -0.040642134845256805,
|
| 264 |
+
"rewards/margins": 0.6047846674919128,
|
| 265 |
+
"rewards/rejected": -0.6454268097877502,
|
| 266 |
+
"step": 140
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.15702695629416383,
|
| 270 |
+
"grad_norm": 2.484375,
|
| 271 |
+
"learning_rate": 4.951404179843963e-06,
|
| 272 |
+
"logits/chosen": -0.5976008176803589,
|
| 273 |
+
"logits/rejected": -0.5590274930000305,
|
| 274 |
+
"logps/chosen": -308.1587829589844,
|
| 275 |
+
"logps/ref_response": -0.5423828363418579,
|
| 276 |
+
"logps/rejected": -281.4222412109375,
|
| 277 |
+
"loss": 0.5468,
|
| 278 |
+
"rewards/accuracies": 0.71875,
|
| 279 |
+
"rewards/chosen": 0.2197073996067047,
|
| 280 |
+
"rewards/margins": 0.6635927557945251,
|
| 281 |
+
"rewards/rejected": -0.4438853859901428,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.16749542004710807,
|
| 286 |
+
"grad_norm": 2.109375,
|
| 287 |
+
"learning_rate": 4.931828996974498e-06,
|
| 288 |
+
"logits/chosen": -0.5375654697418213,
|
| 289 |
+
"logits/rejected": -0.5237765908241272,
|
| 290 |
+
"logps/chosen": -296.62542724609375,
|
| 291 |
+
"logps/ref_response": -0.4895528256893158,
|
| 292 |
+
"logps/rejected": -272.4287414550781,
|
| 293 |
+
"loss": 0.5462,
|
| 294 |
+
"rewards/accuracies": 0.762499988079071,
|
| 295 |
+
"rewards/chosen": 0.23916181921958923,
|
| 296 |
+
"rewards/margins": 0.6912265419960022,
|
| 297 |
+
"rewards/rejected": -0.4520646631717682,
|
| 298 |
+
"step": 160
|
| 299 |
+
},
|
| 300 |
+
{
|
| 301 |
+
"epoch": 0.17796388380005235,
|
| 302 |
+
"grad_norm": 2.46875,
|
| 303 |
+
"learning_rate": 4.909001458367867e-06,
|
| 304 |
+
"logits/chosen": -0.6169471740722656,
|
| 305 |
+
"logits/rejected": -0.5980030298233032,
|
| 306 |
+
"logps/chosen": -289.12432861328125,
|
| 307 |
+
"logps/ref_response": -0.5753272771835327,
|
| 308 |
+
"logps/rejected": -278.28826904296875,
|
| 309 |
+
"loss": 0.551,
|
| 310 |
+
"rewards/accuracies": 0.737500011920929,
|
| 311 |
+
"rewards/chosen": -0.07445995509624481,
|
| 312 |
+
"rewards/margins": 0.5914220213890076,
|
| 313 |
+
"rewards/rejected": -0.6658819913864136,
|
| 314 |
+
"step": 170
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 0.1884323475529966,
|
| 318 |
+
"grad_norm": 2.109375,
|
| 319 |
+
"learning_rate": 4.882952093833628e-06,
|
| 320 |
+
"logits/chosen": -0.6207358241081238,
|
| 321 |
+
"logits/rejected": -0.590207040309906,
|
| 322 |
+
"logps/chosen": -304.12164306640625,
|
| 323 |
+
"logps/ref_response": -0.5761692523956299,
|
| 324 |
+
"logps/rejected": -268.29193115234375,
|
| 325 |
+
"loss": 0.5446,
|
| 326 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 327 |
+
"rewards/chosen": -0.1806192249059677,
|
| 328 |
+
"rewards/margins": 0.5938898324966431,
|
| 329 |
+
"rewards/rejected": -0.774509072303772,
|
| 330 |
+
"step": 180
|
| 331 |
+
},
|
| 332 |
+
{
|
| 333 |
+
"epoch": 0.19890081130594087,
|
| 334 |
+
"grad_norm": 2.96875,
|
| 335 |
+
"learning_rate": 4.853715742087947e-06,
|
| 336 |
+
"logits/chosen": -0.5569428205490112,
|
| 337 |
+
"logits/rejected": -0.5339682102203369,
|
| 338 |
+
"logps/chosen": -276.7432556152344,
|
| 339 |
+
"logps/ref_response": -0.5028859972953796,
|
| 340 |
+
"logps/rejected": -284.9825134277344,
|
| 341 |
+
"loss": 0.5467,
|
| 342 |
+
"rewards/accuracies": 0.699999988079071,
|
| 343 |
+
"rewards/chosen": 0.025777459144592285,
|
| 344 |
+
"rewards/margins": 0.5920640230178833,
|
| 345 |
+
"rewards/rejected": -0.5662865042686462,
|
| 346 |
+
"step": 190
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"epoch": 0.2093692750588851,
|
| 350 |
+
"grad_norm": 3.78125,
|
| 351 |
+
"learning_rate": 4.821331504159906e-06,
|
| 352 |
+
"logits/chosen": -0.561359167098999,
|
| 353 |
+
"logits/rejected": -0.5725305676460266,
|
| 354 |
+
"logps/chosen": -297.00811767578125,
|
| 355 |
+
"logps/ref_response": -0.5163358449935913,
|
| 356 |
+
"logps/rejected": -256.9476013183594,
|
| 357 |
+
"loss": 0.5438,
|
| 358 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 359 |
+
"rewards/chosen": 0.04708755761384964,
|
| 360 |
+
"rewards/margins": 0.6549776196479797,
|
| 361 |
+
"rewards/rejected": -0.6078900098800659,
|
| 362 |
+
"step": 200
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"epoch": 0.2093692750588851,
|
| 366 |
+
"eval_logits/chosen": -0.4850742220878601,
|
| 367 |
+
"eval_logits/rejected": -0.4631035327911377,
|
| 368 |
+
"eval_logps/chosen": -292.0068664550781,
|
| 369 |
+
"eval_logps/ref_response": -0.5363935232162476,
|
| 370 |
+
"eval_logps/rejected": -282.95562744140625,
|
| 371 |
+
"eval_loss": 0.545185923576355,
|
| 372 |
+
"eval_rewards/accuracies": 0.7179999947547913,
|
| 373 |
+
"eval_rewards/chosen": 0.05402619019150734,
|
| 374 |
+
"eval_rewards/margins": 0.6819319725036621,
|
| 375 |
+
"eval_rewards/rejected": -0.6279057860374451,
|
| 376 |
+
"eval_runtime": 349.4927,
|
| 377 |
+
"eval_samples_per_second": 5.723,
|
| 378 |
+
"eval_steps_per_second": 0.358,
|
| 379 |
+
"step": 200
|
| 380 |
+
},
|
| 381 |
+
{
|
| 382 |
+
"epoch": 0.21983773881182936,
|
| 383 |
+
"grad_norm": 3.5625,
|
| 384 |
+
"learning_rate": 4.7858426910973435e-06,
|
| 385 |
+
"logits/chosen": -0.6022263169288635,
|
| 386 |
+
"logits/rejected": -0.6005181074142456,
|
| 387 |
+
"logps/chosen": -279.6543273925781,
|
| 388 |
+
"logps/ref_response": -0.5563252568244934,
|
| 389 |
+
"logps/rejected": -274.255126953125,
|
| 390 |
+
"loss": 0.5428,
|
| 391 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 392 |
+
"rewards/chosen": 0.05802757292985916,
|
| 393 |
+
"rewards/margins": 0.6256381869316101,
|
| 394 |
+
"rewards/rejected": -0.5676106214523315,
|
| 395 |
+
"step": 210
|
| 396 |
+
},
|
| 397 |
+
{
|
| 398 |
+
"epoch": 0.23030620256477363,
|
| 399 |
+
"grad_norm": 2.265625,
|
| 400 |
+
"learning_rate": 4.747296766042161e-06,
|
| 401 |
+
"logits/chosen": -0.5727165937423706,
|
| 402 |
+
"logits/rejected": -0.5475823879241943,
|
| 403 |
+
"logps/chosen": -319.9413146972656,
|
| 404 |
+
"logps/ref_response": -0.525614857673645,
|
| 405 |
+
"logps/rejected": -273.38226318359375,
|
| 406 |
+
"loss": 0.5473,
|
| 407 |
+
"rewards/accuracies": 0.731249988079071,
|
| 408 |
+
"rewards/chosen": 0.04127366468310356,
|
| 409 |
+
"rewards/margins": 0.6884390711784363,
|
| 410 |
+
"rewards/rejected": -0.6471654176712036,
|
| 411 |
+
"step": 220
|
| 412 |
+
},
|
| 413 |
+
{
|
| 414 |
+
"epoch": 0.24077466631771788,
|
| 415 |
+
"grad_norm": 2.765625,
|
| 416 |
+
"learning_rate": 4.705745280752586e-06,
|
| 417 |
+
"logits/chosen": -0.6053592562675476,
|
| 418 |
+
"logits/rejected": -0.5618892312049866,
|
| 419 |
+
"logps/chosen": -293.00775146484375,
|
| 420 |
+
"logps/ref_response": -0.5675605535507202,
|
| 421 |
+
"logps/rejected": -291.1396179199219,
|
| 422 |
+
"loss": 0.5482,
|
| 423 |
+
"rewards/accuracies": 0.75,
|
| 424 |
+
"rewards/chosen": -0.07024725526571274,
|
| 425 |
+
"rewards/margins": 0.6758569478988647,
|
| 426 |
+
"rewards/rejected": -0.7461041212081909,
|
| 427 |
+
"step": 230
|
| 428 |
+
},
|
| 429 |
+
{
|
| 430 |
+
"epoch": 0.2512431300706621,
|
| 431 |
+
"grad_norm": 1.9453125,
|
| 432 |
+
"learning_rate": 4.661243806657256e-06,
|
| 433 |
+
"logits/chosen": -0.5776160955429077,
|
| 434 |
+
"logits/rejected": -0.5337271690368652,
|
| 435 |
+
"logps/chosen": -300.6571044921875,
|
| 436 |
+
"logps/ref_response": -0.5330287218093872,
|
| 437 |
+
"logps/rejected": -265.2437744140625,
|
| 438 |
+
"loss": 0.544,
|
| 439 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 440 |
+
"rewards/chosen": -0.01873040571808815,
|
| 441 |
+
"rewards/margins": 0.5816887617111206,
|
| 442 |
+
"rewards/rejected": -0.6004191637039185,
|
| 443 |
+
"step": 240
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 0.26171159382360637,
|
| 447 |
+
"grad_norm": 2.25,
|
| 448 |
+
"learning_rate": 4.613851860533367e-06,
|
| 449 |
+
"logits/chosen": -0.5833350419998169,
|
| 450 |
+
"logits/rejected": -0.5385982394218445,
|
| 451 |
+
"logps/chosen": -294.57745361328125,
|
| 452 |
+
"logps/ref_response": -0.5492520928382874,
|
| 453 |
+
"logps/rejected": -261.46356201171875,
|
| 454 |
+
"loss": 0.5475,
|
| 455 |
+
"rewards/accuracies": 0.706250011920929,
|
| 456 |
+
"rewards/chosen": 0.042012982070446014,
|
| 457 |
+
"rewards/margins": 0.5611510276794434,
|
| 458 |
+
"rewards/rejected": -0.5191380381584167,
|
| 459 |
+
"step": 250
|
| 460 |
+
},
|
| 461 |
+
{
|
| 462 |
+
"epoch": 0.2721800575765506,
|
| 463 |
+
"grad_norm": 2.140625,
|
| 464 |
+
"learning_rate": 4.563632824908252e-06,
|
| 465 |
+
"logits/chosen": -0.5507840514183044,
|
| 466 |
+
"logits/rejected": -0.5077590346336365,
|
| 467 |
+
"logps/chosen": -294.02032470703125,
|
| 468 |
+
"logps/ref_response": -0.5089389085769653,
|
| 469 |
+
"logps/rejected": -281.4302978515625,
|
| 470 |
+
"loss": 0.5203,
|
| 471 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 472 |
+
"rewards/chosen": 0.23180679976940155,
|
| 473 |
+
"rewards/margins": 0.9116535186767578,
|
| 474 |
+
"rewards/rejected": -0.6798466444015503,
|
| 475 |
+
"step": 260
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 0.2826485213294949,
|
| 479 |
+
"grad_norm": 3.15625,
|
| 480 |
+
"learning_rate": 4.510653863290871e-06,
|
| 481 |
+
"logits/chosen": -0.5404945611953735,
|
| 482 |
+
"logits/rejected": -0.5136505961418152,
|
| 483 |
+
"logps/chosen": -297.4787292480469,
|
| 484 |
+
"logps/ref_response": -0.5091123580932617,
|
| 485 |
+
"logps/rejected": -306.3871765136719,
|
| 486 |
+
"loss": 0.5336,
|
| 487 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 488 |
+
"rewards/chosen": -0.061466194689273834,
|
| 489 |
+
"rewards/margins": 0.6878874897956848,
|
| 490 |
+
"rewards/rejected": -0.7493537068367004,
|
| 491 |
+
"step": 270
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.29311698508243916,
|
| 495 |
+
"grad_norm": 2.078125,
|
| 496 |
+
"learning_rate": 4.454985830346574e-06,
|
| 497 |
+
"logits/chosen": -0.6031721830368042,
|
| 498 |
+
"logits/rejected": -0.5645710229873657,
|
| 499 |
+
"logps/chosen": -303.2027282714844,
|
| 500 |
+
"logps/ref_response": -0.5748014450073242,
|
| 501 |
+
"logps/rejected": -287.49359130859375,
|
| 502 |
+
"loss": 0.5502,
|
| 503 |
+
"rewards/accuracies": 0.643750011920929,
|
| 504 |
+
"rewards/chosen": -0.08863335102796555,
|
| 505 |
+
"rewards/margins": 0.5962889194488525,
|
| 506 |
+
"rewards/rejected": -0.6849222183227539,
|
| 507 |
+
"step": 280
|
| 508 |
+
},
|
| 509 |
+
{
|
| 510 |
+
"epoch": 0.3035854488353834,
|
| 511 |
+
"grad_norm": 2.34375,
|
| 512 |
+
"learning_rate": 4.396703177135262e-06,
|
| 513 |
+
"logits/chosen": -0.5598694086074829,
|
| 514 |
+
"logits/rejected": -0.5239233374595642,
|
| 515 |
+
"logps/chosen": -287.7163391113281,
|
| 516 |
+
"logps/ref_response": -0.5320878624916077,
|
| 517 |
+
"logps/rejected": -259.9132385253906,
|
| 518 |
+
"loss": 0.5259,
|
| 519 |
+
"rewards/accuracies": 0.75,
|
| 520 |
+
"rewards/chosen": 0.1362595558166504,
|
| 521 |
+
"rewards/margins": 0.7511795163154602,
|
| 522 |
+
"rewards/rejected": -0.6149200201034546,
|
| 523 |
+
"step": 290
|
| 524 |
+
},
|
| 525 |
+
{
|
| 526 |
+
"epoch": 0.31405391258832765,
|
| 527 |
+
"grad_norm": 2.78125,
|
| 528 |
+
"learning_rate": 4.335883851539693e-06,
|
| 529 |
+
"logits/chosen": -0.5759841799736023,
|
| 530 |
+
"logits/rejected": -0.5336117148399353,
|
| 531 |
+
"logps/chosen": -296.90283203125,
|
| 532 |
+
"logps/ref_response": -0.5529105067253113,
|
| 533 |
+
"logps/rejected": -294.5677185058594,
|
| 534 |
+
"loss": 0.5367,
|
| 535 |
+
"rewards/accuracies": 0.706250011920929,
|
| 536 |
+
"rewards/chosen": -0.07880517095327377,
|
| 537 |
+
"rewards/margins": 0.7312062978744507,
|
| 538 |
+
"rewards/rejected": -0.8100113868713379,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.31405391258832765,
|
| 543 |
+
"eval_logits/chosen": -0.4076842665672302,
|
| 544 |
+
"eval_logits/rejected": -0.37772703170776367,
|
| 545 |
+
"eval_logps/chosen": -293.4178161621094,
|
| 546 |
+
"eval_logps/ref_response": -0.5363935232162476,
|
| 547 |
+
"eval_logps/rejected": -285.21820068359375,
|
| 548 |
+
"eval_loss": 0.5322972536087036,
|
| 549 |
+
"eval_rewards/accuracies": 0.7239999771118164,
|
| 550 |
+
"eval_rewards/chosen": -0.08706536889076233,
|
| 551 |
+
"eval_rewards/margins": 0.7670957446098328,
|
| 552 |
+
"eval_rewards/rejected": -0.8541611433029175,
|
| 553 |
+
"eval_runtime": 349.5592,
|
| 554 |
+
"eval_samples_per_second": 5.721,
|
| 555 |
+
"eval_steps_per_second": 0.358,
|
| 556 |
+
"step": 300
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"epoch": 0.3245223763412719,
|
| 560 |
+
"grad_norm": 2.71875,
|
| 561 |
+
"learning_rate": 4.2726091940171055e-06,
|
| 562 |
+
"logits/chosen": -0.5224083065986633,
|
| 563 |
+
"logits/rejected": -0.5524710416793823,
|
| 564 |
+
"logps/chosen": -295.6260681152344,
|
| 565 |
+
"logps/ref_response": -0.5006662607192993,
|
| 566 |
+
"logps/rejected": -342.54486083984375,
|
| 567 |
+
"loss": 0.517,
|
| 568 |
+
"rewards/accuracies": 0.800000011920929,
|
| 569 |
+
"rewards/chosen": 0.12224096059799194,
|
| 570 |
+
"rewards/margins": 0.872395396232605,
|
| 571 |
+
"rewards/rejected": -0.7501543760299683,
|
| 572 |
+
"step": 310
|
| 573 |
+
},
|
| 574 |
+
{
|
| 575 |
+
"epoch": 0.33499084009421615,
|
| 576 |
+
"grad_norm": 1.796875,
|
| 577 |
+
"learning_rate": 4.206963828813555e-06,
|
| 578 |
+
"logits/chosen": -0.5745652914047241,
|
| 579 |
+
"logits/rejected": -0.5253760814666748,
|
| 580 |
+
"logps/chosen": -296.783447265625,
|
| 581 |
+
"logps/ref_response": -0.5563712120056152,
|
| 582 |
+
"logps/rejected": -281.023681640625,
|
| 583 |
+
"loss": 0.5186,
|
| 584 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 585 |
+
"rewards/chosen": 0.0009771495824679732,
|
| 586 |
+
"rewards/margins": 0.8379068374633789,
|
| 587 |
+
"rewards/rejected": -0.8369296193122864,
|
| 588 |
+
"step": 320
|
| 589 |
+
},
|
| 590 |
+
{
|
| 591 |
+
"epoch": 0.34545930384716045,
|
| 592 |
+
"grad_norm": 2.03125,
|
| 593 |
+
"learning_rate": 4.139035550786495e-06,
|
| 594 |
+
"logits/chosen": -0.6098914742469788,
|
| 595 |
+
"logits/rejected": -0.5377870798110962,
|
| 596 |
+
"logps/chosen": -290.60955810546875,
|
| 597 |
+
"logps/ref_response": -0.5800708532333374,
|
| 598 |
+
"logps/rejected": -262.45794677734375,
|
| 599 |
+
"loss": 0.5298,
|
| 600 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 601 |
+
"rewards/chosen": -0.1382167935371399,
|
| 602 |
+
"rewards/margins": 0.6818909645080566,
|
| 603 |
+
"rewards/rejected": -0.820107638835907,
|
| 604 |
+
"step": 330
|
| 605 |
+
},
|
| 606 |
+
{
|
| 607 |
+
"epoch": 0.3559277676001047,
|
| 608 |
+
"grad_norm": 2.40625,
|
| 609 |
+
"learning_rate": 4.068915207986931e-06,
|
| 610 |
+
"logits/chosen": -0.5661717653274536,
|
| 611 |
+
"logits/rejected": -0.49055665731430054,
|
| 612 |
+
"logps/chosen": -298.7618713378906,
|
| 613 |
+
"logps/ref_response": -0.5407181978225708,
|
| 614 |
+
"logps/rejected": -259.991943359375,
|
| 615 |
+
"loss": 0.5262,
|
| 616 |
+
"rewards/accuracies": 0.75,
|
| 617 |
+
"rewards/chosen": -0.3091699182987213,
|
| 618 |
+
"rewards/margins": 0.8509536981582642,
|
| 619 |
+
"rewards/rejected": -1.1601234674453735,
|
| 620 |
+
"step": 340
|
| 621 |
+
},
|
| 622 |
+
{
|
| 623 |
+
"epoch": 0.36639623135304894,
|
| 624 |
+
"grad_norm": 2.25,
|
| 625 |
+
"learning_rate": 3.996696580158211e-06,
|
| 626 |
+
"logits/chosen": -0.5107800364494324,
|
| 627 |
+
"logits/rejected": -0.477796733379364,
|
| 628 |
+
"logps/chosen": -338.06329345703125,
|
| 629 |
+
"logps/ref_response": -0.486247718334198,
|
| 630 |
+
"logps/rejected": -293.3652648925781,
|
| 631 |
+
"loss": 0.5389,
|
| 632 |
+
"rewards/accuracies": 0.71875,
|
| 633 |
+
"rewards/chosen": -0.1124631017446518,
|
| 634 |
+
"rewards/margins": 0.7731461524963379,
|
| 635 |
+
"rewards/rejected": -0.8856091499328613,
|
| 636 |
+
"step": 350
|
| 637 |
+
},
|
| 638 |
+
{
|
| 639 |
+
"epoch": 0.3768646951059932,
|
| 640 |
+
"grad_norm": 2.0,
|
| 641 |
+
"learning_rate": 3.922476253313921e-06,
|
| 642 |
+
"logits/chosen": -0.48775357007980347,
|
| 643 |
+
"logits/rejected": -0.4903165400028229,
|
| 644 |
+
"logps/chosen": -275.6280822753906,
|
| 645 |
+
"logps/ref_response": -0.48780474066734314,
|
| 646 |
+
"logps/rejected": -299.45269775390625,
|
| 647 |
+
"loss": 0.523,
|
| 648 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 649 |
+
"rewards/chosen": -0.24213580787181854,
|
| 650 |
+
"rewards/margins": 0.7627736330032349,
|
| 651 |
+
"rewards/rejected": -1.0049093961715698,
|
| 652 |
+
"step": 360
|
| 653 |
+
},
|
| 654 |
+
{
|
| 655 |
+
"epoch": 0.38733315885893743,
|
| 656 |
+
"grad_norm": 3.171875,
|
| 657 |
+
"learning_rate": 3.846353490562664e-06,
|
| 658 |
+
"logits/chosen": -0.5095352530479431,
|
| 659 |
+
"logits/rejected": -0.5028492212295532,
|
| 660 |
+
"logps/chosen": -289.87359619140625,
|
| 661 |
+
"logps/ref_response": -0.491685152053833,
|
| 662 |
+
"logps/rejected": -264.93988037109375,
|
| 663 |
+
"loss": 0.508,
|
| 664 |
+
"rewards/accuracies": 0.762499988079071,
|
| 665 |
+
"rewards/chosen": 0.03148346394300461,
|
| 666 |
+
"rewards/margins": 0.9142478704452515,
|
| 667 |
+
"rewards/rejected": -0.8827645182609558,
|
| 668 |
+
"step": 370
|
| 669 |
+
},
|
| 670 |
+
{
|
| 671 |
+
"epoch": 0.39780162261188173,
|
| 672 |
+
"grad_norm": 2.171875,
|
| 673 |
+
"learning_rate": 3.768430099352445e-06,
|
| 674 |
+
"logits/chosen": -0.528128981590271,
|
| 675 |
+
"logits/rejected": -0.5178142786026001,
|
| 676 |
+
"logps/chosen": -307.6114501953125,
|
| 677 |
+
"logps/ref_response": -0.5215914845466614,
|
| 678 |
+
"logps/rejected": -281.41888427734375,
|
| 679 |
+
"loss": 0.5227,
|
| 680 |
+
"rewards/accuracies": 0.768750011920929,
|
| 681 |
+
"rewards/chosen": -0.09869714826345444,
|
| 682 |
+
"rewards/margins": 1.0406019687652588,
|
| 683 |
+
"rewards/rejected": -1.1392991542816162,
|
| 684 |
+
"step": 380
|
| 685 |
+
},
|
| 686 |
+
{
|
| 687 |
+
"epoch": 0.408270086364826,
|
| 688 |
+
"grad_norm": 3.234375,
|
| 689 |
+
"learning_rate": 3.6888102953122307e-06,
|
| 690 |
+
"logits/chosen": -0.5722348093986511,
|
| 691 |
+
"logits/rejected": -0.525412380695343,
|
| 692 |
+
"logps/chosen": -266.93212890625,
|
| 693 |
+
"logps/ref_response": -0.5661150813102722,
|
| 694 |
+
"logps/rejected": -268.66094970703125,
|
| 695 |
+
"loss": 0.5332,
|
| 696 |
+
"rewards/accuracies": 0.768750011920929,
|
| 697 |
+
"rewards/chosen": -0.3913359045982361,
|
| 698 |
+
"rewards/margins": 0.8005081415176392,
|
| 699 |
+
"rewards/rejected": -1.1918439865112305,
|
| 700 |
+
"step": 390
|
| 701 |
+
},
|
| 702 |
+
{
|
| 703 |
+
"epoch": 0.4187385501177702,
|
| 704 |
+
"grad_norm": 2.828125,
|
| 705 |
+
"learning_rate": 3.607600562872785e-06,
|
| 706 |
+
"logits/chosen": -0.5220564603805542,
|
| 707 |
+
"logits/rejected": -0.4830014705657959,
|
| 708 |
+
"logps/chosen": -286.83056640625,
|
| 709 |
+
"logps/ref_response": -0.5258094072341919,
|
| 710 |
+
"logps/rejected": -278.11456298828125,
|
| 711 |
+
"loss": 0.5196,
|
| 712 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 713 |
+
"rewards/chosen": -0.10876519978046417,
|
| 714 |
+
"rewards/margins": 0.8259785771369934,
|
| 715 |
+
"rewards/rejected": -0.9347437024116516,
|
| 716 |
+
"step": 400
|
| 717 |
+
},
|
| 718 |
+
{
|
| 719 |
+
"epoch": 0.4187385501177702,
|
| 720 |
+
"eval_logits/chosen": -0.3280640244483948,
|
| 721 |
+
"eval_logits/rejected": -0.2898561656475067,
|
| 722 |
+
"eval_logps/chosen": -292.925537109375,
|
| 723 |
+
"eval_logps/ref_response": -0.5363935232162476,
|
| 724 |
+
"eval_logps/rejected": -285.290283203125,
|
| 725 |
+
"eval_loss": 0.5235576629638672,
|
| 726 |
+
"eval_rewards/accuracies": 0.7319999933242798,
|
| 727 |
+
"eval_rewards/chosen": -0.03783903643488884,
|
| 728 |
+
"eval_rewards/margins": 0.823529839515686,
|
| 729 |
+
"eval_rewards/rejected": -0.8613688349723816,
|
| 730 |
+
"eval_runtime": 349.5237,
|
| 731 |
+
"eval_samples_per_second": 5.722,
|
| 732 |
+
"eval_steps_per_second": 0.358,
|
| 733 |
+
"step": 400
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.42920701387071447,
|
| 737 |
+
"grad_norm": 2.53125,
|
| 738 |
+
"learning_rate": 3.5249095128531863e-06,
|
| 739 |
+
"logits/chosen": -0.54010409116745,
|
| 740 |
+
"logits/rejected": -0.485682874917984,
|
| 741 |
+
"logps/chosen": -278.5982360839844,
|
| 742 |
+
"logps/ref_response": -0.5564926862716675,
|
| 743 |
+
"logps/rejected": -277.0348205566406,
|
| 744 |
+
"loss": 0.5119,
|
| 745 |
+
"rewards/accuracies": 0.768750011920929,
|
| 746 |
+
"rewards/chosen": -0.04259440302848816,
|
| 747 |
+
"rewards/margins": 0.8908407092094421,
|
| 748 |
+
"rewards/rejected": -0.9334350824356079,
|
| 749 |
+
"step": 410
|
| 750 |
+
},
|
| 751 |
+
{
|
| 752 |
+
"epoch": 0.4396754776236587,
|
| 753 |
+
"grad_norm": 2.25,
|
| 754 |
+
"learning_rate": 3.4408477372034743e-06,
|
| 755 |
+
"logits/chosen": -0.5260821580886841,
|
| 756 |
+
"logits/rejected": -0.4941217303276062,
|
| 757 |
+
"logps/chosen": -310.8683776855469,
|
| 758 |
+
"logps/ref_response": -0.5361344218254089,
|
| 759 |
+
"logps/rejected": -299.2928161621094,
|
| 760 |
+
"loss": 0.5345,
|
| 761 |
+
"rewards/accuracies": 0.762499988079071,
|
| 762 |
+
"rewards/chosen": -0.19507813453674316,
|
| 763 |
+
"rewards/margins": 0.7008574604988098,
|
| 764 |
+
"rewards/rejected": -0.8959355354309082,
|
| 765 |
+
"step": 420
|
| 766 |
+
},
|
| 767 |
+
{
|
| 768 |
+
"epoch": 0.45014394137660296,
|
| 769 |
+
"grad_norm": 3.484375,
|
| 770 |
+
"learning_rate": 3.355527661097728e-06,
|
| 771 |
+
"logits/chosen": -0.5269330739974976,
|
| 772 |
+
"logits/rejected": -0.5198745727539062,
|
| 773 |
+
"logps/chosen": -282.8704833984375,
|
| 774 |
+
"logps/ref_response": -0.5477866530418396,
|
| 775 |
+
"logps/rejected": -284.66448974609375,
|
| 776 |
+
"loss": 0.5363,
|
| 777 |
+
"rewards/accuracies": 0.75,
|
| 778 |
+
"rewards/chosen": -0.3844759464263916,
|
| 779 |
+
"rewards/margins": 0.7087138295173645,
|
| 780 |
+
"rewards/rejected": -1.0931897163391113,
|
| 781 |
+
"step": 430
|
| 782 |
+
},
|
| 783 |
+
{
|
| 784 |
+
"epoch": 0.46061240512954726,
|
| 785 |
+
"grad_norm": 2.265625,
|
| 786 |
+
"learning_rate": 3.269063392575352e-06,
|
| 787 |
+
"logits/chosen": -0.4928794503211975,
|
| 788 |
+
"logits/rejected": -0.4792874753475189,
|
| 789 |
+
"logps/chosen": -329.7843017578125,
|
| 790 |
+
"logps/ref_response": -0.5050511360168457,
|
| 791 |
+
"logps/rejected": -308.486572265625,
|
| 792 |
+
"loss": 0.5184,
|
| 793 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 794 |
+
"rewards/chosen": -0.10263900458812714,
|
| 795 |
+
"rewards/margins": 0.7831138372421265,
|
| 796 |
+
"rewards/rejected": -0.8857528567314148,
|
| 797 |
+
"step": 440
|
| 798 |
+
},
|
| 799 |
+
{
|
| 800 |
+
"epoch": 0.4710808688824915,
|
| 801 |
+
"grad_norm": 1.9453125,
|
| 802 |
+
"learning_rate": 3.181570569931697e-06,
|
| 803 |
+
"logits/chosen": -0.5173367857933044,
|
| 804 |
+
"logits/rejected": -0.49148645997047424,
|
| 805 |
+
"logps/chosen": -287.93524169921875,
|
| 806 |
+
"logps/ref_response": -0.5224987864494324,
|
| 807 |
+
"logps/rejected": -284.7456970214844,
|
| 808 |
+
"loss": 0.5167,
|
| 809 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 810 |
+
"rewards/chosen": -0.2799225449562073,
|
| 811 |
+
"rewards/margins": 0.795501708984375,
|
| 812 |
+
"rewards/rejected": -1.0754241943359375,
|
| 813 |
+
"step": 450
|
| 814 |
+
},
|
| 815 |
+
{
|
| 816 |
+
"epoch": 0.48154933263543576,
|
| 817 |
+
"grad_norm": 1.7109375,
|
| 818 |
+
"learning_rate": 3.09316620706208e-06,
|
| 819 |
+
"logits/chosen": -0.4698413014411926,
|
| 820 |
+
"logits/rejected": -0.4720715582370758,
|
| 821 |
+
"logps/chosen": -310.1413879394531,
|
| 822 |
+
"logps/ref_response": -0.4874509274959564,
|
| 823 |
+
"logps/rejected": -292.0880126953125,
|
| 824 |
+
"loss": 0.5003,
|
| 825 |
+
"rewards/accuracies": 0.793749988079071,
|
| 826 |
+
"rewards/chosen": -0.2300359308719635,
|
| 827 |
+
"rewards/margins": 1.019820213317871,
|
| 828 |
+
"rewards/rejected": -1.2498562335968018,
|
| 829 |
+
"step": 460
|
| 830 |
+
},
|
| 831 |
+
{
|
| 832 |
+
"epoch": 0.49201779638838,
|
| 833 |
+
"grad_norm": 2.28125,
|
| 834 |
+
"learning_rate": 3.0039685369660785e-06,
|
| 835 |
+
"logits/chosen": -0.4714682102203369,
|
| 836 |
+
"logits/rejected": -0.4186578392982483,
|
| 837 |
+
"logps/chosen": -284.99407958984375,
|
| 838 |
+
"logps/ref_response": -0.4861488938331604,
|
| 839 |
+
"logps/rejected": -270.0592041015625,
|
| 840 |
+
"loss": 0.5301,
|
| 841 |
+
"rewards/accuracies": 0.71875,
|
| 842 |
+
"rewards/chosen": -0.1522492617368698,
|
| 843 |
+
"rewards/margins": 0.8750549554824829,
|
| 844 |
+
"rewards/rejected": -1.0273042917251587,
|
| 845 |
+
"step": 470
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 0.5024862601413242,
|
| 849 |
+
"grad_norm": 2.15625,
|
| 850 |
+
"learning_rate": 2.91409685362137e-06,
|
| 851 |
+
"logits/chosen": -0.4800891876220703,
|
| 852 |
+
"logits/rejected": -0.4620879590511322,
|
| 853 |
+
"logps/chosen": -282.891845703125,
|
| 854 |
+
"logps/ref_response": -0.5061747431755066,
|
| 855 |
+
"logps/rejected": -280.4508972167969,
|
| 856 |
+
"loss": 0.506,
|
| 857 |
+
"rewards/accuracies": 0.731249988079071,
|
| 858 |
+
"rewards/chosen": -0.3358300030231476,
|
| 859 |
+
"rewards/margins": 0.7954305410385132,
|
| 860 |
+
"rewards/rejected": -1.1312605142593384,
|
| 861 |
+
"step": 480
|
| 862 |
+
},
|
| 863 |
+
{
|
| 864 |
+
"epoch": 0.5129547238942685,
|
| 865 |
+
"grad_norm": 1.4921875,
|
| 866 |
+
"learning_rate": 2.8236713524386085e-06,
|
| 867 |
+
"logits/chosen": -0.5464354753494263,
|
| 868 |
+
"logits/rejected": -0.4956323504447937,
|
| 869 |
+
"logps/chosen": -283.87847900390625,
|
| 870 |
+
"logps/ref_response": -0.5583964586257935,
|
| 871 |
+
"logps/rejected": -262.64935302734375,
|
| 872 |
+
"loss": 0.5042,
|
| 873 |
+
"rewards/accuracies": 0.78125,
|
| 874 |
+
"rewards/chosen": -0.2944543957710266,
|
| 875 |
+
"rewards/margins": 0.9288061857223511,
|
| 876 |
+
"rewards/rejected": -1.223260760307312,
|
| 877 |
+
"step": 490
|
| 878 |
+
},
|
| 879 |
+
{
|
| 880 |
+
"epoch": 0.5234231876472127,
|
| 881 |
+
"grad_norm": 2.28125,
|
| 882 |
+
"learning_rate": 2.7328129695107205e-06,
|
| 883 |
+
"logits/chosen": -0.45227164030075073,
|
| 884 |
+
"logits/rejected": -0.4433063864707947,
|
| 885 |
+
"logps/chosen": -268.2086181640625,
|
| 886 |
+
"logps/ref_response": -0.46682921051979065,
|
| 887 |
+
"logps/rejected": -278.5611572265625,
|
| 888 |
+
"loss": 0.509,
|
| 889 |
+
"rewards/accuracies": 0.8125,
|
| 890 |
+
"rewards/chosen": -0.18779410421848297,
|
| 891 |
+
"rewards/margins": 1.08889639377594,
|
| 892 |
+
"rewards/rejected": -1.2766902446746826,
|
| 893 |
+
"step": 500
|
| 894 |
+
},
|
| 895 |
+
{
|
| 896 |
+
"epoch": 0.5234231876472127,
|
| 897 |
+
"eval_logits/chosen": -0.2739206552505493,
|
| 898 |
+
"eval_logits/rejected": -0.22961243987083435,
|
| 899 |
+
"eval_logps/chosen": -295.23968505859375,
|
| 900 |
+
"eval_logps/ref_response": -0.536393404006958,
|
| 901 |
+
"eval_logps/rejected": -287.97900390625,
|
| 902 |
+
"eval_loss": 0.5184563994407654,
|
| 903 |
+
"eval_rewards/accuracies": 0.7360000014305115,
|
| 904 |
+
"eval_rewards/chosen": -0.2692505419254303,
|
| 905 |
+
"eval_rewards/margins": 0.8609901070594788,
|
| 906 |
+
"eval_rewards/rejected": -1.1302406787872314,
|
| 907 |
+
"eval_runtime": 349.5683,
|
| 908 |
+
"eval_samples_per_second": 5.721,
|
| 909 |
+
"eval_steps_per_second": 0.358,
|
| 910 |
+
"step": 500
|
| 911 |
+
},
|
| 912 |
+
{
|
| 913 |
+
"epoch": 0.533891651400157,
|
| 914 |
+
"grad_norm": 1.8515625,
|
| 915 |
+
"learning_rate": 2.641643219871597e-06,
|
| 916 |
+
"logits/chosen": -0.4826118052005768,
|
| 917 |
+
"logits/rejected": -0.4357355237007141,
|
| 918 |
+
"logps/chosen": -315.3731689453125,
|
| 919 |
+
"logps/ref_response": -0.5090646743774414,
|
| 920 |
+
"logps/rejected": -300.0047607421875,
|
| 921 |
+
"loss": 0.5249,
|
| 922 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 923 |
+
"rewards/chosen": -0.24246744811534882,
|
| 924 |
+
"rewards/margins": 0.7875919342041016,
|
| 925 |
+
"rewards/rejected": -1.0300593376159668,
|
| 926 |
+
"step": 510
|
| 927 |
+
},
|
| 928 |
+
{
|
| 929 |
+
"epoch": 0.5443601151531012,
|
| 930 |
+
"grad_norm": 2.65625,
|
| 931 |
+
"learning_rate": 2.5502840349805074e-06,
|
| 932 |
+
"logits/chosen": -0.4724315106868744,
|
| 933 |
+
"logits/rejected": -0.4490731656551361,
|
| 934 |
+
"logps/chosen": -312.1684265136719,
|
| 935 |
+
"logps/ref_response": -0.5057616829872131,
|
| 936 |
+
"logps/rejected": -300.3685302734375,
|
| 937 |
+
"loss": 0.531,
|
| 938 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 939 |
+
"rewards/chosen": -0.21491758525371552,
|
| 940 |
+
"rewards/margins": 0.9349055290222168,
|
| 941 |
+
"rewards/rejected": -1.1498230695724487,
|
| 942 |
+
"step": 520
|
| 943 |
+
},
|
| 944 |
+
{
|
| 945 |
+
"epoch": 0.5548285789060455,
|
| 946 |
+
"grad_norm": 2.6875,
|
| 947 |
+
"learning_rate": 2.4588575996495797e-06,
|
| 948 |
+
"logits/chosen": -0.43103843927383423,
|
| 949 |
+
"logits/rejected": -0.4253179430961609,
|
| 950 |
+
"logps/chosen": -275.5965881347656,
|
| 951 |
+
"logps/ref_response": -0.45075368881225586,
|
| 952 |
+
"logps/rejected": -266.48785400390625,
|
| 953 |
+
"loss": 0.5277,
|
| 954 |
+
"rewards/accuracies": 0.762499988079071,
|
| 955 |
+
"rewards/chosen": -0.5141997337341309,
|
| 956 |
+
"rewards/margins": 0.9270822405815125,
|
| 957 |
+
"rewards/rejected": -1.441282033920288,
|
| 958 |
+
"step": 530
|
| 959 |
+
},
|
| 960 |
+
{
|
| 961 |
+
"epoch": 0.5652970426589898,
|
| 962 |
+
"grad_norm": 2.984375,
|
| 963 |
+
"learning_rate": 2.367486188632446e-06,
|
| 964 |
+
"logits/chosen": -0.47084522247314453,
|
| 965 |
+
"logits/rejected": -0.46598607301712036,
|
| 966 |
+
"logps/chosen": -288.1656188964844,
|
| 967 |
+
"logps/ref_response": -0.5035119652748108,
|
| 968 |
+
"logps/rejected": -328.1705322265625,
|
| 969 |
+
"loss": 0.5151,
|
| 970 |
+
"rewards/accuracies": 0.731249988079071,
|
| 971 |
+
"rewards/chosen": -0.2480819970369339,
|
| 972 |
+
"rewards/margins": 0.912807822227478,
|
| 973 |
+
"rewards/rejected": -1.1608898639678955,
|
| 974 |
+
"step": 540
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 0.575765506411934,
|
| 978 |
+
"grad_norm": 2.234375,
|
| 979 |
+
"learning_rate": 2.276292003092593e-06,
|
| 980 |
+
"logits/chosen": -0.4974418580532074,
|
| 981 |
+
"logits/rejected": -0.46066489815711975,
|
| 982 |
+
"logps/chosen": -259.3627624511719,
|
| 983 |
+
"logps/ref_response": -0.5067554712295532,
|
| 984 |
+
"logps/rejected": -267.8301086425781,
|
| 985 |
+
"loss": 0.5027,
|
| 986 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 987 |
+
"rewards/chosen": -0.1483170986175537,
|
| 988 |
+
"rewards/margins": 0.9989498257637024,
|
| 989 |
+
"rewards/rejected": -1.1472669839859009,
|
| 990 |
+
"step": 550
|
| 991 |
+
},
|
| 992 |
+
{
|
| 993 |
+
"epoch": 0.5862339701648783,
|
| 994 |
+
"grad_norm": 2.0625,
|
| 995 |
+
"learning_rate": 2.1853970071701415e-06,
|
| 996 |
+
"logits/chosen": -0.4845849871635437,
|
| 997 |
+
"logits/rejected": -0.4430512487888336,
|
| 998 |
+
"logps/chosen": -281.133544921875,
|
| 999 |
+
"logps/ref_response": -0.5059608817100525,
|
| 1000 |
+
"logps/rejected": -283.225830078125,
|
| 1001 |
+
"loss": 0.5108,
|
| 1002 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1003 |
+
"rewards/chosen": -0.17830374836921692,
|
| 1004 |
+
"rewards/margins": 0.9193024635314941,
|
| 1005 |
+
"rewards/rejected": -1.0976061820983887,
|
| 1006 |
+
"step": 560
|
| 1007 |
+
},
|
| 1008 |
+
{
|
| 1009 |
+
"epoch": 0.5967024339178225,
|
| 1010 |
+
"grad_norm": 2.671875,
|
| 1011 |
+
"learning_rate": 2.0949227648656194e-06,
|
| 1012 |
+
"logits/chosen": -0.5085529088973999,
|
| 1013 |
+
"logits/rejected": -0.45733585953712463,
|
| 1014 |
+
"logps/chosen": -298.2741394042969,
|
| 1015 |
+
"logps/ref_response": -0.5283219218254089,
|
| 1016 |
+
"logps/rejected": -266.26641845703125,
|
| 1017 |
+
"loss": 0.5149,
|
| 1018 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 1019 |
+
"rewards/chosen": -0.31216269731521606,
|
| 1020 |
+
"rewards/margins": 0.9785518646240234,
|
| 1021 |
+
"rewards/rejected": -1.2907145023345947,
|
| 1022 |
+
"step": 570
|
| 1023 |
+
},
|
| 1024 |
+
{
|
| 1025 |
+
"epoch": 0.6071708976707668,
|
| 1026 |
+
"grad_norm": 1.765625,
|
| 1027 |
+
"learning_rate": 2.00499027745888e-06,
|
| 1028 |
+
"logits/chosen": -0.4733489453792572,
|
| 1029 |
+
"logits/rejected": -0.4443192481994629,
|
| 1030 |
+
"logps/chosen": -302.96319580078125,
|
| 1031 |
+
"logps/ref_response": -0.5130476355552673,
|
| 1032 |
+
"logps/rejected": -302.777099609375,
|
| 1033 |
+
"loss": 0.524,
|
| 1034 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1035 |
+
"rewards/chosen": -0.31352347135543823,
|
| 1036 |
+
"rewards/margins": 0.8776998519897461,
|
| 1037 |
+
"rewards/rejected": -1.191223382949829,
|
| 1038 |
+
"step": 580
|
| 1039 |
+
},
|
| 1040 |
+
{
|
| 1041 |
+
"epoch": 0.6176393614237111,
|
| 1042 |
+
"grad_norm": 1.953125,
|
| 1043 |
+
"learning_rate": 1.915719821680624e-06,
|
| 1044 |
+
"logits/chosen": -0.5032647848129272,
|
| 1045 |
+
"logits/rejected": -0.425483763217926,
|
| 1046 |
+
"logps/chosen": -291.1716613769531,
|
| 1047 |
+
"logps/ref_response": -0.5210384130477905,
|
| 1048 |
+
"logps/rejected": -287.4841003417969,
|
| 1049 |
+
"loss": 0.5085,
|
| 1050 |
+
"rewards/accuracies": 0.78125,
|
| 1051 |
+
"rewards/chosen": -0.12319432199001312,
|
| 1052 |
+
"rewards/margins": 0.930014967918396,
|
| 1053 |
+
"rewards/rejected": -1.0532093048095703,
|
| 1054 |
+
"step": 590
|
| 1055 |
+
},
|
| 1056 |
+
{
|
| 1057 |
+
"epoch": 0.6281078251766553,
|
| 1058 |
+
"grad_norm": 1.734375,
|
| 1059 |
+
"learning_rate": 1.8272307888529276e-06,
|
| 1060 |
+
"logits/chosen": -0.4218737483024597,
|
| 1061 |
+
"logits/rejected": -0.3671064078807831,
|
| 1062 |
+
"logps/chosen": -267.94598388671875,
|
| 1063 |
+
"logps/ref_response": -0.4653666913509369,
|
| 1064 |
+
"logps/rejected": -285.668212890625,
|
| 1065 |
+
"loss": 0.5012,
|
| 1066 |
+
"rewards/accuracies": 0.768750011920929,
|
| 1067 |
+
"rewards/chosen": -0.35736727714538574,
|
| 1068 |
+
"rewards/margins": 0.9875767827033997,
|
| 1069 |
+
"rewards/rejected": -1.3449440002441406,
|
| 1070 |
+
"step": 600
|
| 1071 |
+
},
|
| 1072 |
+
{
|
| 1073 |
+
"epoch": 0.6281078251766553,
|
| 1074 |
+
"eval_logits/chosen": -0.23965981602668762,
|
| 1075 |
+
"eval_logits/rejected": -0.19264473021030426,
|
| 1076 |
+
"eval_logps/chosen": -296.0675354003906,
|
| 1077 |
+
"eval_logps/ref_response": -0.5363935232162476,
|
| 1078 |
+
"eval_logps/rejected": -289.1474914550781,
|
| 1079 |
+
"eval_loss": 0.5152395963668823,
|
| 1080 |
+
"eval_rewards/accuracies": 0.7480000257492065,
|
| 1081 |
+
"eval_rewards/chosen": -0.3520371615886688,
|
| 1082 |
+
"eval_rewards/margins": 0.895053505897522,
|
| 1083 |
+
"eval_rewards/rejected": -1.2470906972885132,
|
| 1084 |
+
"eval_runtime": 349.5867,
|
| 1085 |
+
"eval_samples_per_second": 5.721,
|
| 1086 |
+
"eval_steps_per_second": 0.358,
|
| 1087 |
+
"step": 600
|
| 1088 |
+
},
|
| 1089 |
+
{
|
| 1090 |
+
"epoch": 0.6385762889295996,
|
| 1091 |
+
"grad_norm": 1.9140625,
|
| 1092 |
+
"learning_rate": 1.739641525213929e-06,
|
| 1093 |
+
"logits/chosen": -0.4473685324192047,
|
| 1094 |
+
"logits/rejected": -0.4271601736545563,
|
| 1095 |
+
"logps/chosen": -269.72998046875,
|
| 1096 |
+
"logps/ref_response": -0.500705897808075,
|
| 1097 |
+
"logps/rejected": -275.93499755859375,
|
| 1098 |
+
"loss": 0.5048,
|
| 1099 |
+
"rewards/accuracies": 0.75,
|
| 1100 |
+
"rewards/chosen": -0.3331184685230255,
|
| 1101 |
+
"rewards/margins": 0.9662030339241028,
|
| 1102 |
+
"rewards/rejected": -1.2993214130401611,
|
| 1103 |
+
"step": 610
|
| 1104 |
+
},
|
| 1105 |
+
{
|
| 1106 |
+
"epoch": 0.6490447526825438,
|
| 1107 |
+
"grad_norm": 2.09375,
|
| 1108 |
+
"learning_rate": 1.6530691736402317e-06,
|
| 1109 |
+
"logits/chosen": -0.4652767777442932,
|
| 1110 |
+
"logits/rejected": -0.41162675619125366,
|
| 1111 |
+
"logps/chosen": -296.17120361328125,
|
| 1112 |
+
"logps/ref_response": -0.502620279788971,
|
| 1113 |
+
"logps/rejected": -286.85284423828125,
|
| 1114 |
+
"loss": 0.5152,
|
| 1115 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1116 |
+
"rewards/chosen": -0.32981282472610474,
|
| 1117 |
+
"rewards/margins": 1.0166393518447876,
|
| 1118 |
+
"rewards/rejected": -1.3464521169662476,
|
| 1119 |
+
"step": 620
|
| 1120 |
+
},
|
| 1121 |
+
{
|
| 1122 |
+
"epoch": 0.6595132164354881,
|
| 1123 |
+
"grad_norm": 2.15625,
|
| 1124 |
+
"learning_rate": 1.5676295169786864e-06,
|
| 1125 |
+
"logits/chosen": -0.48260921239852905,
|
| 1126 |
+
"logits/rejected": -0.43052974343299866,
|
| 1127 |
+
"logps/chosen": -288.948486328125,
|
| 1128 |
+
"logps/ref_response": -0.522256076335907,
|
| 1129 |
+
"logps/rejected": -275.0885009765625,
|
| 1130 |
+
"loss": 0.5087,
|
| 1131 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1132 |
+
"rewards/chosen": -0.2533746659755707,
|
| 1133 |
+
"rewards/margins": 0.9306432604789734,
|
| 1134 |
+
"rewards/rejected": -1.1840178966522217,
|
| 1135 |
+
"step": 630
|
| 1136 |
+
},
|
| 1137 |
+
{
|
| 1138 |
+
"epoch": 0.6699816801884323,
|
| 1139 |
+
"grad_norm": 1.7109375,
|
| 1140 |
+
"learning_rate": 1.4834368231970922e-06,
|
| 1141 |
+
"logits/chosen": -0.5107460618019104,
|
| 1142 |
+
"logits/rejected": -0.4472557604312897,
|
| 1143 |
+
"logps/chosen": -287.7373352050781,
|
| 1144 |
+
"logps/ref_response": -0.5478745698928833,
|
| 1145 |
+
"logps/rejected": -274.25653076171875,
|
| 1146 |
+
"loss": 0.4972,
|
| 1147 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 1148 |
+
"rewards/chosen": -0.11735512316226959,
|
| 1149 |
+
"rewards/margins": 0.8366876840591431,
|
| 1150 |
+
"rewards/rejected": -0.9540427327156067,
|
| 1151 |
+
"step": 640
|
| 1152 |
+
},
|
| 1153 |
+
{
|
| 1154 |
+
"epoch": 0.6804501439413766,
|
| 1155 |
+
"grad_norm": 2.046875,
|
| 1156 |
+
"learning_rate": 1.4006036925609245e-06,
|
| 1157 |
+
"logits/chosen": -0.48332634568214417,
|
| 1158 |
+
"logits/rejected": -0.41887950897216797,
|
| 1159 |
+
"logps/chosen": -300.3069152832031,
|
| 1160 |
+
"logps/ref_response": -0.5103051662445068,
|
| 1161 |
+
"logps/rejected": -251.2999725341797,
|
| 1162 |
+
"loss": 0.5163,
|
| 1163 |
+
"rewards/accuracies": 0.768750011920929,
|
| 1164 |
+
"rewards/chosen": -0.16883328557014465,
|
| 1165 |
+
"rewards/margins": 0.9550157785415649,
|
| 1166 |
+
"rewards/rejected": -1.1238490343093872,
|
| 1167 |
+
"step": 650
|
| 1168 |
+
},
|
| 1169 |
+
{
|
| 1170 |
+
"epoch": 0.6909186076943209,
|
| 1171 |
+
"grad_norm": 2.125,
|
| 1172 |
+
"learning_rate": 1.3192409070404582e-06,
|
| 1173 |
+
"logits/chosen": -0.4920195937156677,
|
| 1174 |
+
"logits/rejected": -0.45738130807876587,
|
| 1175 |
+
"logps/chosen": -304.8662414550781,
|
| 1176 |
+
"logps/ref_response": -0.5286127328872681,
|
| 1177 |
+
"logps/rejected": -307.9287109375,
|
| 1178 |
+
"loss": 0.5109,
|
| 1179 |
+
"rewards/accuracies": 0.71875,
|
| 1180 |
+
"rewards/chosen": -0.044912584125995636,
|
| 1181 |
+
"rewards/margins": 0.9029264450073242,
|
| 1182 |
+
"rewards/rejected": -0.9478389620780945,
|
| 1183 |
+
"step": 660
|
| 1184 |
+
},
|
| 1185 |
+
{
|
| 1186 |
+
"epoch": 0.7013870714472651,
|
| 1187 |
+
"grad_norm": 2.640625,
|
| 1188 |
+
"learning_rate": 1.2394572821496953e-06,
|
| 1189 |
+
"logits/chosen": -0.4847448468208313,
|
| 1190 |
+
"logits/rejected": -0.44459033012390137,
|
| 1191 |
+
"logps/chosen": -278.811279296875,
|
| 1192 |
+
"logps/ref_response": -0.5491371154785156,
|
| 1193 |
+
"logps/rejected": -260.7911682128906,
|
| 1194 |
+
"loss": 0.5158,
|
| 1195 |
+
"rewards/accuracies": 0.71875,
|
| 1196 |
+
"rewards/chosen": -0.17895063757896423,
|
| 1197 |
+
"rewards/margins": 0.8761111497879028,
|
| 1198 |
+
"rewards/rejected": -1.0550616979599,
|
| 1199 |
+
"step": 670
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 0.7118555352002094,
|
| 1203 |
+
"grad_norm": 1.703125,
|
| 1204 |
+
"learning_rate": 1.1613595214152713e-06,
|
| 1205 |
+
"logits/chosen": -0.5145822763442993,
|
| 1206 |
+
"logits/rejected": -0.45838356018066406,
|
| 1207 |
+
"logps/chosen": -288.93072509765625,
|
| 1208 |
+
"logps/ref_response": -0.5694643259048462,
|
| 1209 |
+
"logps/rejected": -278.52642822265625,
|
| 1210 |
+
"loss": 0.5018,
|
| 1211 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1212 |
+
"rewards/chosen": -0.2978869080543518,
|
| 1213 |
+
"rewards/margins": 0.903253436088562,
|
| 1214 |
+
"rewards/rejected": -1.2011405229568481,
|
| 1215 |
+
"step": 680
|
| 1216 |
+
},
|
| 1217 |
+
{
|
| 1218 |
+
"epoch": 0.7223239989531536,
|
| 1219 |
+
"grad_norm": 1.7578125,
|
| 1220 |
+
"learning_rate": 1.0850520736699362e-06,
|
| 1221 |
+
"logits/chosen": -0.46343159675598145,
|
| 1222 |
+
"logits/rejected": -0.40929287672042847,
|
| 1223 |
+
"logps/chosen": -342.8109130859375,
|
| 1224 |
+
"logps/ref_response": -0.4945286810398102,
|
| 1225 |
+
"logps/rejected": -318.6613464355469,
|
| 1226 |
+
"loss": 0.516,
|
| 1227 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 1228 |
+
"rewards/chosen": -0.27149656414985657,
|
| 1229 |
+
"rewards/margins": 0.9835799336433411,
|
| 1230 |
+
"rewards/rejected": -1.25507652759552,
|
| 1231 |
+
"step": 690
|
| 1232 |
+
},
|
| 1233 |
+
{
|
| 1234 |
+
"epoch": 0.7327924627060979,
|
| 1235 |
+
"grad_norm": 1.96875,
|
| 1236 |
+
"learning_rate": 1.0106369933615043e-06,
|
| 1237 |
+
"logits/chosen": -0.5036773681640625,
|
| 1238 |
+
"logits/rejected": -0.42831772565841675,
|
| 1239 |
+
"logps/chosen": -317.4853515625,
|
| 1240 |
+
"logps/ref_response": -0.5506774187088013,
|
| 1241 |
+
"logps/rejected": -265.49896240234375,
|
| 1242 |
+
"loss": 0.5168,
|
| 1243 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 1244 |
+
"rewards/chosen": -0.30581605434417725,
|
| 1245 |
+
"rewards/margins": 0.7802382707595825,
|
| 1246 |
+
"rewards/rejected": -1.0860543251037598,
|
| 1247 |
+
"step": 700
|
| 1248 |
+
},
|
| 1249 |
+
{
|
| 1250 |
+
"epoch": 0.7327924627060979,
|
| 1251 |
+
"eval_logits/chosen": -0.2157570868730545,
|
| 1252 |
+
"eval_logits/rejected": -0.16649317741394043,
|
| 1253 |
+
"eval_logps/chosen": -295.0681457519531,
|
| 1254 |
+
"eval_logps/ref_response": -0.5363935232162476,
|
| 1255 |
+
"eval_logps/rejected": -288.23870849609375,
|
| 1256 |
+
"eval_loss": 0.5139358639717102,
|
| 1257 |
+
"eval_rewards/accuracies": 0.7440000176429749,
|
| 1258 |
+
"eval_rewards/chosen": -0.2521001994609833,
|
| 1259 |
+
"eval_rewards/margins": 0.9041155576705933,
|
| 1260 |
+
"eval_rewards/rejected": -1.156215786933899,
|
| 1261 |
+
"eval_runtime": 349.6297,
|
| 1262 |
+
"eval_samples_per_second": 5.72,
|
| 1263 |
+
"eval_steps_per_second": 0.358,
|
| 1264 |
+
"step": 700
|
| 1265 |
+
},
|
| 1266 |
+
{
|
| 1267 |
+
"epoch": 0.7432609264590422,
|
| 1268 |
+
"grad_norm": 2.265625,
|
| 1269 |
+
"learning_rate": 9.382138040640714e-07,
|
| 1270 |
+
"logits/chosen": -0.509280800819397,
|
| 1271 |
+
"logits/rejected": -0.45220834016799927,
|
| 1272 |
+
"logps/chosen": -266.75555419921875,
|
| 1273 |
+
"logps/ref_response": -0.5634459257125854,
|
| 1274 |
+
"logps/rejected": -281.2605285644531,
|
| 1275 |
+
"loss": 0.5042,
|
| 1276 |
+
"rewards/accuracies": 0.78125,
|
| 1277 |
+
"rewards/chosen": -0.24380390346050262,
|
| 1278 |
+
"rewards/margins": 0.9118436574935913,
|
| 1279 |
+
"rewards/rejected": -1.1556475162506104,
|
| 1280 |
+
"step": 710
|
| 1281 |
+
},
|
| 1282 |
+
{
|
| 1283 |
+
"epoch": 0.7537293902119864,
|
| 1284 |
+
"grad_norm": 1.890625,
|
| 1285 |
+
"learning_rate": 8.678793653740633e-07,
|
| 1286 |
+
"logits/chosen": -0.43401581048965454,
|
| 1287 |
+
"logits/rejected": -0.41436678171157837,
|
| 1288 |
+
"logps/chosen": -264.42425537109375,
|
| 1289 |
+
"logps/ref_response": -0.49243393540382385,
|
| 1290 |
+
"logps/rejected": -265.80517578125,
|
| 1291 |
+
"loss": 0.5067,
|
| 1292 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1293 |
+
"rewards/chosen": -0.12135882675647736,
|
| 1294 |
+
"rewards/margins": 1.0111067295074463,
|
| 1295 |
+
"rewards/rejected": -1.1324656009674072,
|
| 1296 |
+
"step": 720
|
| 1297 |
+
},
|
| 1298 |
+
{
|
| 1299 |
+
"epoch": 0.7641978539649307,
|
| 1300 |
+
"grad_norm": 1.5078125,
|
| 1301 |
+
"learning_rate": 7.997277433690984e-07,
|
| 1302 |
+
"logits/chosen": -0.4582897126674652,
|
| 1303 |
+
"logits/rejected": -0.38019412755966187,
|
| 1304 |
+
"logps/chosen": -303.3408203125,
|
| 1305 |
+
"logps/ref_response": -0.4944031834602356,
|
| 1306 |
+
"logps/rejected": -290.0057067871094,
|
| 1307 |
+
"loss": 0.5008,
|
| 1308 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1309 |
+
"rewards/chosen": -0.24817109107971191,
|
| 1310 |
+
"rewards/margins": 0.8901812434196472,
|
| 1311 |
+
"rewards/rejected": -1.1383522748947144,
|
| 1312 |
+
"step": 730
|
| 1313 |
+
},
|
| 1314 |
+
{
|
| 1315 |
+
"epoch": 0.7746663177178749,
|
| 1316 |
+
"grad_norm": 1.53125,
|
| 1317 |
+
"learning_rate": 7.338500848029603e-07,
|
| 1318 |
+
"logits/chosen": -0.4196249544620514,
|
| 1319 |
+
"logits/rejected": -0.41058415174484253,
|
| 1320 |
+
"logps/chosen": -293.5699157714844,
|
| 1321 |
+
"logps/ref_response": -0.4282347559928894,
|
| 1322 |
+
"logps/rejected": -278.59942626953125,
|
| 1323 |
+
"loss": 0.4953,
|
| 1324 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1325 |
+
"rewards/chosen": -0.13130240142345428,
|
| 1326 |
+
"rewards/margins": 0.9627996683120728,
|
| 1327 |
+
"rewards/rejected": -1.0941020250320435,
|
| 1328 |
+
"step": 740
|
| 1329 |
+
},
|
| 1330 |
+
{
|
| 1331 |
+
"epoch": 0.7851347814708192,
|
| 1332 |
+
"grad_norm": 1.9296875,
|
| 1333 |
+
"learning_rate": 6.70334495204884e-07,
|
| 1334 |
+
"logits/chosen": -0.4480054974555969,
|
| 1335 |
+
"logits/rejected": -0.40563225746154785,
|
| 1336 |
+
"logps/chosen": -326.13360595703125,
|
| 1337 |
+
"logps/ref_response": -0.49645256996154785,
|
| 1338 |
+
"logps/rejected": -289.4101257324219,
|
| 1339 |
+
"loss": 0.5131,
|
| 1340 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1341 |
+
"rewards/chosen": -0.11767721176147461,
|
| 1342 |
+
"rewards/margins": 0.9364310503005981,
|
| 1343 |
+
"rewards/rejected": -1.0541083812713623,
|
| 1344 |
+
"step": 750
|
| 1345 |
+
},
|
| 1346 |
+
{
|
| 1347 |
+
"epoch": 0.7956032452237635,
|
| 1348 |
+
"grad_norm": 2.234375,
|
| 1349 |
+
"learning_rate": 6.092659210462232e-07,
|
| 1350 |
+
"logits/chosen": -0.47132453322410583,
|
| 1351 |
+
"logits/rejected": -0.44373002648353577,
|
| 1352 |
+
"logps/chosen": -271.53887939453125,
|
| 1353 |
+
"logps/ref_response": -0.5222411751747131,
|
| 1354 |
+
"logps/rejected": -271.93829345703125,
|
| 1355 |
+
"loss": 0.5203,
|
| 1356 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 1357 |
+
"rewards/chosen": -0.2481314241886139,
|
| 1358 |
+
"rewards/margins": 0.8226064443588257,
|
| 1359 |
+
"rewards/rejected": -1.0707378387451172,
|
| 1360 |
+
"step": 760
|
| 1361 |
+
},
|
| 1362 |
+
{
|
| 1363 |
+
"epoch": 0.8060717089767077,
|
| 1364 |
+
"grad_norm": 1.4609375,
|
| 1365 |
+
"learning_rate": 5.507260361320738e-07,
|
| 1366 |
+
"logits/chosen": -0.45327791571617126,
|
| 1367 |
+
"logits/rejected": -0.4443967342376709,
|
| 1368 |
+
"logps/chosen": -287.787109375,
|
| 1369 |
+
"logps/ref_response": -0.50932776927948,
|
| 1370 |
+
"logps/rejected": -282.6838073730469,
|
| 1371 |
+
"loss": 0.5151,
|
| 1372 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 1373 |
+
"rewards/chosen": -0.15161371231079102,
|
| 1374 |
+
"rewards/margins": 0.7541275024414062,
|
| 1375 |
+
"rewards/rejected": -0.9057412147521973,
|
| 1376 |
+
"step": 770
|
| 1377 |
+
},
|
| 1378 |
+
{
|
| 1379 |
+
"epoch": 0.816540172729652,
|
| 1380 |
+
"grad_norm": 1.5,
|
| 1381 |
+
"learning_rate": 4.947931323697983e-07,
|
| 1382 |
+
"logits/chosen": -0.4503496289253235,
|
| 1383 |
+
"logits/rejected": -0.4044824540615082,
|
| 1384 |
+
"logps/chosen": -289.29718017578125,
|
| 1385 |
+
"logps/ref_response": -0.49121037125587463,
|
| 1386 |
+
"logps/rejected": -283.48651123046875,
|
| 1387 |
+
"loss": 0.5147,
|
| 1388 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1389 |
+
"rewards/chosen": -0.2892809808254242,
|
| 1390 |
+
"rewards/margins": 0.7692901492118835,
|
| 1391 |
+
"rewards/rejected": -1.0585711002349854,
|
| 1392 |
+
"step": 780
|
| 1393 |
+
},
|
| 1394 |
+
{
|
| 1395 |
+
"epoch": 0.8270086364825961,
|
| 1396 |
+
"grad_norm": 2.46875,
|
| 1397 |
+
"learning_rate": 4.4154201506053985e-07,
|
| 1398 |
+
"logits/chosen": -0.4808201789855957,
|
| 1399 |
+
"logits/rejected": -0.437244176864624,
|
| 1400 |
+
"logps/chosen": -303.23907470703125,
|
| 1401 |
+
"logps/ref_response": -0.5042006373405457,
|
| 1402 |
+
"logps/rejected": -268.26165771484375,
|
| 1403 |
+
"loss": 0.5086,
|
| 1404 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1405 |
+
"rewards/chosen": -0.2500911056995392,
|
| 1406 |
+
"rewards/margins": 0.948900043964386,
|
| 1407 |
+
"rewards/rejected": -1.198991060256958,
|
| 1408 |
+
"step": 790
|
| 1409 |
+
},
|
| 1410 |
+
{
|
| 1411 |
+
"epoch": 0.8374771002355405,
|
| 1412 |
+
"grad_norm": 1.109375,
|
| 1413 |
+
"learning_rate": 3.910439028537638e-07,
|
| 1414 |
+
"logits/chosen": -0.4850529730319977,
|
| 1415 |
+
"logits/rejected": -0.4150736927986145,
|
| 1416 |
+
"logps/chosen": -351.47967529296875,
|
| 1417 |
+
"logps/ref_response": -0.5149141550064087,
|
| 1418 |
+
"logps/rejected": -306.39044189453125,
|
| 1419 |
+
"loss": 0.5156,
|
| 1420 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1421 |
+
"rewards/chosen": -0.20133157074451447,
|
| 1422 |
+
"rewards/margins": 0.7480217218399048,
|
| 1423 |
+
"rewards/rejected": -0.9493532180786133,
|
| 1424 |
+
"step": 800
|
| 1425 |
+
},
|
| 1426 |
+
{
|
| 1427 |
+
"epoch": 0.8374771002355405,
|
| 1428 |
+
"eval_logits/chosen": -0.21027056872844696,
|
| 1429 |
+
"eval_logits/rejected": -0.16033047437667847,
|
| 1430 |
+
"eval_logps/chosen": -294.7516174316406,
|
| 1431 |
+
"eval_logps/ref_response": -0.5363935232162476,
|
| 1432 |
+
"eval_logps/rejected": -287.9801330566406,
|
| 1433 |
+
"eval_loss": 0.5134991407394409,
|
| 1434 |
+
"eval_rewards/accuracies": 0.7519999742507935,
|
| 1435 |
+
"eval_rewards/chosen": -0.22044621407985687,
|
| 1436 |
+
"eval_rewards/margins": 0.9099085927009583,
|
| 1437 |
+
"eval_rewards/rejected": -1.1303547620773315,
|
| 1438 |
+
"eval_runtime": 349.5707,
|
| 1439 |
+
"eval_samples_per_second": 5.721,
|
| 1440 |
+
"eval_steps_per_second": 0.358,
|
| 1441 |
+
"step": 800
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 0.8479455639884846,
|
| 1445 |
+
"grad_norm": 1.59375,
|
| 1446 |
+
"learning_rate": 3.4336633249862084e-07,
|
| 1447 |
+
"logits/chosen": -0.5064218640327454,
|
| 1448 |
+
"logits/rejected": -0.40463584661483765,
|
| 1449 |
+
"logps/chosen": -323.1934814453125,
|
| 1450 |
+
"logps/ref_response": -0.5519742369651794,
|
| 1451 |
+
"logps/rejected": -292.6800231933594,
|
| 1452 |
+
"loss": 0.4987,
|
| 1453 |
+
"rewards/accuracies": 0.731249988079071,
|
| 1454 |
+
"rewards/chosen": -0.28738316893577576,
|
| 1455 |
+
"rewards/margins": 0.8722078204154968,
|
| 1456 |
+
"rewards/rejected": -1.1595909595489502,
|
| 1457 |
+
"step": 810
|
| 1458 |
+
},
|
| 1459 |
+
{
|
| 1460 |
+
"epoch": 0.8584140277414289,
|
| 1461 |
+
"grad_norm": 1.78125,
|
| 1462 |
+
"learning_rate": 2.98573068519539e-07,
|
| 1463 |
+
"logits/chosen": -0.4821909964084625,
|
| 1464 |
+
"logits/rejected": -0.44719791412353516,
|
| 1465 |
+
"logps/chosen": -310.1927185058594,
|
| 1466 |
+
"logps/ref_response": -0.5307375192642212,
|
| 1467 |
+
"logps/rejected": -297.6319274902344,
|
| 1468 |
+
"loss": 0.5104,
|
| 1469 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1470 |
+
"rewards/chosen": -0.15440870821475983,
|
| 1471 |
+
"rewards/margins": 0.9249560236930847,
|
| 1472 |
+
"rewards/rejected": -1.0793647766113281,
|
| 1473 |
+
"step": 820
|
| 1474 |
+
},
|
| 1475 |
+
{
|
| 1476 |
+
"epoch": 0.8688824914943732,
|
| 1477 |
+
"grad_norm": 2.234375,
|
| 1478 |
+
"learning_rate": 2.5672401793681854e-07,
|
| 1479 |
+
"logits/chosen": -0.49778860807418823,
|
| 1480 |
+
"logits/rejected": -0.4640674591064453,
|
| 1481 |
+
"logps/chosen": -277.7762145996094,
|
| 1482 |
+
"logps/ref_response": -0.5466696619987488,
|
| 1483 |
+
"logps/rejected": -273.833740234375,
|
| 1484 |
+
"loss": 0.5,
|
| 1485 |
+
"rewards/accuracies": 0.78125,
|
| 1486 |
+
"rewards/chosen": -0.24376177787780762,
|
| 1487 |
+
"rewards/margins": 0.9780392646789551,
|
| 1488 |
+
"rewards/rejected": -1.2218010425567627,
|
| 1489 |
+
"step": 830
|
| 1490 |
+
},
|
| 1491 |
+
{
|
| 1492 |
+
"epoch": 0.8793509552473174,
|
| 1493 |
+
"grad_norm": 1.828125,
|
| 1494 |
+
"learning_rate": 2.178751501463036e-07,
|
| 1495 |
+
"logits/chosen": -0.46538400650024414,
|
| 1496 |
+
"logits/rejected": -0.4340798258781433,
|
| 1497 |
+
"logps/chosen": -317.69232177734375,
|
| 1498 |
+
"logps/ref_response": -0.5086795091629028,
|
| 1499 |
+
"logps/rejected": -311.81475830078125,
|
| 1500 |
+
"loss": 0.4914,
|
| 1501 |
+
"rewards/accuracies": 0.75,
|
| 1502 |
+
"rewards/chosen": -0.2690751850605011,
|
| 1503 |
+
"rewards/margins": 0.9315555691719055,
|
| 1504 |
+
"rewards/rejected": -1.200630784034729,
|
| 1505 |
+
"step": 840
|
| 1506 |
+
},
|
| 1507 |
+
{
|
| 1508 |
+
"epoch": 0.8898194190002617,
|
| 1509 |
+
"grad_norm": 2.015625,
|
| 1510 |
+
"learning_rate": 1.820784220652766e-07,
|
| 1511 |
+
"logits/chosen": -0.5234388113021851,
|
| 1512 |
+
"logits/rejected": -0.4476381838321686,
|
| 1513 |
+
"logps/chosen": -348.9575500488281,
|
| 1514 |
+
"logps/ref_response": -0.5546728372573853,
|
| 1515 |
+
"logps/rejected": -283.19366455078125,
|
| 1516 |
+
"loss": 0.5109,
|
| 1517 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1518 |
+
"rewards/chosen": -0.0391000397503376,
|
| 1519 |
+
"rewards/margins": 0.9498162269592285,
|
| 1520 |
+
"rewards/rejected": -0.9889162182807922,
|
| 1521 |
+
"step": 850
|
| 1522 |
+
},
|
| 1523 |
+
{
|
| 1524 |
+
"epoch": 0.9002878827532059,
|
| 1525 |
+
"grad_norm": 1.9453125,
|
| 1526 |
+
"learning_rate": 1.4938170864468636e-07,
|
| 1527 |
+
"logits/chosen": -0.44146886467933655,
|
| 1528 |
+
"logits/rejected": -0.38905996084213257,
|
| 1529 |
+
"logps/chosen": -292.7279968261719,
|
| 1530 |
+
"logps/ref_response": -0.4814940392971039,
|
| 1531 |
+
"logps/rejected": -273.334228515625,
|
| 1532 |
+
"loss": 0.4883,
|
| 1533 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 1534 |
+
"rewards/chosen": -0.21229040622711182,
|
| 1535 |
+
"rewards/margins": 1.0407958030700684,
|
| 1536 |
+
"rewards/rejected": -1.2530862092971802,
|
| 1537 |
+
"step": 860
|
| 1538 |
+
},
|
| 1539 |
+
{
|
| 1540 |
+
"epoch": 0.9107563465061502,
|
| 1541 |
+
"grad_norm": 2.46875,
|
| 1542 |
+
"learning_rate": 1.1982873884064466e-07,
|
| 1543 |
+
"logits/chosen": -0.40739497542381287,
|
| 1544 |
+
"logits/rejected": -0.3892499804496765,
|
| 1545 |
+
"logps/chosen": -290.3547668457031,
|
| 1546 |
+
"logps/ref_response": -0.463235467672348,
|
| 1547 |
+
"logps/rejected": -281.25054931640625,
|
| 1548 |
+
"loss": 0.5111,
|
| 1549 |
+
"rewards/accuracies": 0.731249988079071,
|
| 1550 |
+
"rewards/chosen": -0.24277010560035706,
|
| 1551 |
+
"rewards/margins": 0.7852509617805481,
|
| 1552 |
+
"rewards/rejected": -1.0280208587646484,
|
| 1553 |
+
"step": 870
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 0.9212248102590945,
|
| 1557 |
+
"grad_norm": 1.84375,
|
| 1558 |
+
"learning_rate": 9.345903713082305e-08,
|
| 1559 |
+
"logits/chosen": -0.4908636510372162,
|
| 1560 |
+
"logits/rejected": -0.4692970812320709,
|
| 1561 |
+
"logps/chosen": -317.8591613769531,
|
| 1562 |
+
"logps/ref_response": -0.5406745672225952,
|
| 1563 |
+
"logps/rejected": -284.82958984375,
|
| 1564 |
+
"loss": 0.5206,
|
| 1565 |
+
"rewards/accuracies": 0.637499988079071,
|
| 1566 |
+
"rewards/chosen": -0.23101326823234558,
|
| 1567 |
+
"rewards/margins": 0.7203452587127686,
|
| 1568 |
+
"rewards/rejected": -0.9513584971427917,
|
| 1569 |
+
"step": 880
|
| 1570 |
+
},
|
| 1571 |
+
{
|
| 1572 |
+
"epoch": 0.9316932740120387,
|
| 1573 |
+
"grad_norm": 2.765625,
|
| 1574 |
+
"learning_rate": 7.030787065396866e-08,
|
| 1575 |
+
"logits/chosen": -0.4572540819644928,
|
| 1576 |
+
"logits/rejected": -0.4023072123527527,
|
| 1577 |
+
"logps/chosen": -322.2685546875,
|
| 1578 |
+
"logps/ref_response": -0.5117658376693726,
|
| 1579 |
+
"logps/rejected": -297.341064453125,
|
| 1580 |
+
"loss": 0.5061,
|
| 1581 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 1582 |
+
"rewards/chosen": -0.18995711207389832,
|
| 1583 |
+
"rewards/margins": 0.8081440925598145,
|
| 1584 |
+
"rewards/rejected": -0.9981012344360352,
|
| 1585 |
+
"step": 890
|
| 1586 |
+
},
|
| 1587 |
+
{
|
| 1588 |
+
"epoch": 0.942161737764983,
|
| 1589 |
+
"grad_norm": 2.078125,
|
| 1590 |
+
"learning_rate": 5.0406202043228604e-08,
|
| 1591 |
+
"logits/chosen": -0.4801081120967865,
|
| 1592 |
+
"logits/rejected": -0.4477986693382263,
|
| 1593 |
+
"logps/chosen": -336.6103820800781,
|
| 1594 |
+
"logps/ref_response": -0.5195636749267578,
|
| 1595 |
+
"logps/rejected": -278.45306396484375,
|
| 1596 |
+
"loss": 0.506,
|
| 1597 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 1598 |
+
"rewards/chosen": -0.05415695160627365,
|
| 1599 |
+
"rewards/margins": 1.0482077598571777,
|
| 1600 |
+
"rewards/rejected": -1.1023646593093872,
|
| 1601 |
+
"step": 900
|
| 1602 |
+
},
|
| 1603 |
+
{
|
| 1604 |
+
"epoch": 0.942161737764983,
|
| 1605 |
+
"eval_logits/chosen": -0.20995385944843292,
|
| 1606 |
+
"eval_logits/rejected": -0.1602318435907364,
|
| 1607 |
+
"eval_logps/chosen": -294.57037353515625,
|
| 1608 |
+
"eval_logps/ref_response": -0.5363935232162476,
|
| 1609 |
+
"eval_logps/rejected": -287.7952575683594,
|
| 1610 |
+
"eval_loss": 0.5134302377700806,
|
| 1611 |
+
"eval_rewards/accuracies": 0.7480000257492065,
|
| 1612 |
+
"eval_rewards/chosen": -0.20232149958610535,
|
| 1613 |
+
"eval_rewards/margins": 0.9095419049263,
|
| 1614 |
+
"eval_rewards/rejected": -1.111863374710083,
|
| 1615 |
+
"eval_runtime": 349.7136,
|
| 1616 |
+
"eval_samples_per_second": 5.719,
|
| 1617 |
+
"eval_steps_per_second": 0.357,
|
| 1618 |
+
"step": 900
|
| 1619 |
+
},
|
| 1620 |
+
{
|
| 1621 |
+
"epoch": 0.9526302015179272,
|
| 1622 |
+
"grad_norm": 1.546875,
|
| 1623 |
+
"learning_rate": 3.378064801637687e-08,
|
| 1624 |
+
"logits/chosen": -0.5161974430084229,
|
| 1625 |
+
"logits/rejected": -0.4518052935600281,
|
| 1626 |
+
"logps/chosen": -317.8857421875,
|
| 1627 |
+
"logps/ref_response": -0.561827540397644,
|
| 1628 |
+
"logps/rejected": -317.4599609375,
|
| 1629 |
+
"loss": 0.5156,
|
| 1630 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1631 |
+
"rewards/chosen": -0.11832026392221451,
|
| 1632 |
+
"rewards/margins": 0.8348624110221863,
|
| 1633 |
+
"rewards/rejected": -0.9531826972961426,
|
| 1634 |
+
"step": 910
|
| 1635 |
+
},
|
| 1636 |
+
{
|
| 1637 |
+
"epoch": 0.9630986652708715,
|
| 1638 |
+
"grad_norm": 1.8046875,
|
| 1639 |
+
"learning_rate": 2.0453443778310766e-08,
|
| 1640 |
+
"logits/chosen": -0.4482875466346741,
|
| 1641 |
+
"logits/rejected": -0.3840841054916382,
|
| 1642 |
+
"logps/chosen": -332.5438232421875,
|
| 1643 |
+
"logps/ref_response": -0.4732615351676941,
|
| 1644 |
+
"logps/rejected": -309.3650207519531,
|
| 1645 |
+
"loss": 0.5037,
|
| 1646 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1647 |
+
"rewards/chosen": -0.15293975174427032,
|
| 1648 |
+
"rewards/margins": 1.0318092107772827,
|
| 1649 |
+
"rewards/rejected": -1.184748888015747,
|
| 1650 |
+
"step": 920
|
| 1651 |
+
},
|
| 1652 |
+
{
|
| 1653 |
+
"epoch": 0.9735671290238157,
|
| 1654 |
+
"grad_norm": 1.7109375,
|
| 1655 |
+
"learning_rate": 1.0442413283435759e-08,
|
| 1656 |
+
"logits/chosen": -0.4494338929653168,
|
| 1657 |
+
"logits/rejected": -0.37092915177345276,
|
| 1658 |
+
"logps/chosen": -321.01531982421875,
|
| 1659 |
+
"logps/ref_response": -0.4792579114437103,
|
| 1660 |
+
"logps/rejected": -282.218017578125,
|
| 1661 |
+
"loss": 0.4975,
|
| 1662 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1663 |
+
"rewards/chosen": -0.13543161749839783,
|
| 1664 |
+
"rewards/margins": 1.1757858991622925,
|
| 1665 |
+
"rewards/rejected": -1.3112175464630127,
|
| 1666 |
+
"step": 930
|
| 1667 |
+
},
|
| 1668 |
+
{
|
| 1669 |
+
"epoch": 0.98403559277676,
|
| 1670 |
+
"grad_norm": 1.453125,
|
| 1671 |
+
"learning_rate": 3.760945397705828e-09,
|
| 1672 |
+
"logits/chosen": -0.4677162170410156,
|
| 1673 |
+
"logits/rejected": -0.39422911405563354,
|
| 1674 |
+
"logps/chosen": -293.9225769042969,
|
| 1675 |
+
"logps/ref_response": -0.5234506726264954,
|
| 1676 |
+
"logps/rejected": -264.69964599609375,
|
| 1677 |
+
"loss": 0.4975,
|
| 1678 |
+
"rewards/accuracies": 0.78125,
|
| 1679 |
+
"rewards/chosen": -0.05600558966398239,
|
| 1680 |
+
"rewards/margins": 1.1023480892181396,
|
| 1681 |
+
"rewards/rejected": -1.1583536863327026,
|
| 1682 |
+
"step": 940
|
| 1683 |
+
},
|
| 1684 |
+
{
|
| 1685 |
+
"epoch": 0.9945040565297043,
|
| 1686 |
+
"grad_norm": 2.015625,
|
| 1687 |
+
"learning_rate": 4.1797599220405605e-10,
|
| 1688 |
+
"logits/chosen": -0.48652324080467224,
|
| 1689 |
+
"logits/rejected": -0.4518989622592926,
|
| 1690 |
+
"logps/chosen": -298.553955078125,
|
| 1691 |
+
"logps/ref_response": -0.5367287397384644,
|
| 1692 |
+
"logps/rejected": -282.85626220703125,
|
| 1693 |
+
"loss": 0.5151,
|
| 1694 |
+
"rewards/accuracies": 0.75,
|
| 1695 |
+
"rewards/chosen": -0.14170411229133606,
|
| 1696 |
+
"rewards/margins": 0.9237847328186035,
|
| 1697 |
+
"rewards/rejected": -1.0654886960983276,
|
| 1698 |
+
"step": 950
|
| 1699 |
+
},
|
| 1700 |
+
{
|
| 1701 |
+
"epoch": 0.9997382884061764,
|
| 1702 |
+
"step": 955,
|
| 1703 |
+
"total_flos": 0.0,
|
| 1704 |
+
"train_loss": 0.5342667900454936,
|
| 1705 |
+
"train_runtime": 19113.1655,
|
| 1706 |
+
"train_samples_per_second": 3.199,
|
| 1707 |
+
"train_steps_per_second": 0.05
|
| 1708 |
+
}
|
| 1709 |
+
],
|
| 1710 |
+
"logging_steps": 10,
|
| 1711 |
+
"max_steps": 955,
|
| 1712 |
+
"num_input_tokens_seen": 0,
|
| 1713 |
+
"num_train_epochs": 1,
|
| 1714 |
+
"save_steps": 100000,
|
| 1715 |
+
"stateful_callbacks": {
|
| 1716 |
+
"TrainerControl": {
|
| 1717 |
+
"args": {
|
| 1718 |
+
"should_epoch_stop": false,
|
| 1719 |
+
"should_evaluate": false,
|
| 1720 |
+
"should_log": false,
|
| 1721 |
+
"should_save": true,
|
| 1722 |
+
"should_training_stop": true
|
| 1723 |
+
},
|
| 1724 |
+
"attributes": {}
|
| 1725 |
+
}
|
| 1726 |
+
},
|
| 1727 |
+
"total_flos": 0.0,
|
| 1728 |
+
"train_batch_size": 1,
|
| 1729 |
+
"trial_name": null,
|
| 1730 |
+
"trial_params": null
|
| 1731 |
+
}
|