Commit
•
53f7f4a
1
Parent(s):
e81e7c6
Model save
Browse files- README.md +75 -0
- adapter_model.safetensors +1 -1
- all_results.json +8 -0
- train_results.json +8 -0
- trainer_state.json +663 -0
README.md
ADDED
|
@@ -0,0 +1,75 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
library_name: peft
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
base_model: norallm/normistral-7b-warm
|
| 9 |
+
model-index:
|
| 10 |
+
- name: ap-normistral-7b-align-scan
|
| 11 |
+
results: []
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
+
|
| 17 |
+
# ap-normistral-7b-align-scan
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [norallm/normistral-7b-warm](https://huggingface.co/norallm/normistral-7b-warm) on the None dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.7249
|
| 22 |
+
- Rewards/chosen: -0.1096
|
| 23 |
+
- Rewards/rejected: -0.2129
|
| 24 |
+
- Rewards/accuracies: 0.5282
|
| 25 |
+
- Rewards/margins: 0.1033
|
| 26 |
+
- Logps/rejected: -36.3214
|
| 27 |
+
- Logps/chosen: -32.6259
|
| 28 |
+
- Logits/rejected: 98.6240
|
| 29 |
+
- Logits/chosen: 98.6482
|
| 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: 4
|
| 50 |
+
- eval_batch_size: 8
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- gradient_accumulation_steps: 2
|
| 54 |
+
- total_train_batch_size: 8
|
| 55 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 56 |
+
- lr_scheduler_type: cosine
|
| 57 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 58 |
+
- num_epochs: 1
|
| 59 |
+
|
| 60 |
+
### Training results
|
| 61 |
+
|
| 62 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 63 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 64 |
+
| 0.6845 | 0.26 | 100 | 0.7399 | 0.0002 | -0.0442 | 0.5245 | 0.0444 | -36.0401 | -32.4428 | 98.7048 | 98.7141 |
|
| 65 |
+
| 0.6115 | 0.52 | 200 | 0.7253 | -0.1275 | -0.2129 | 0.5303 | 0.0854 | -36.3214 | -32.6557 | 98.6043 | 98.6285 |
|
| 66 |
+
| 0.5545 | 0.78 | 300 | 0.7249 | -0.1096 | -0.2129 | 0.5282 | 0.1033 | -36.3214 | -32.6259 | 98.6240 | 98.6482 |
|
| 67 |
+
|
| 68 |
+
|
| 69 |
+
### Framework versions
|
| 70 |
+
|
| 71 |
+
- PEFT 0.10.0
|
| 72 |
+
- Transformers 4.39.0.dev0
|
| 73 |
+
- Pytorch 2.1.2+cu121
|
| 74 |
+
- Datasets 2.14.6
|
| 75 |
+
- Tokenizers 0.15.1
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 671150064
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8e1c4fb40cab2ed298164fbfba66ecca63a7920ee8c467f1203b62898385f2bc
|
| 3 |
size 671150064
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.5886486524111265,
|
| 4 |
+
"train_runtime": 2556.9439,
|
| 5 |
+
"train_samples": 3079,
|
| 6 |
+
"train_samples_per_second": 1.204,
|
| 7 |
+
"train_steps_per_second": 0.151
|
| 8 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.5886486524111265,
|
| 4 |
+
"train_runtime": 2556.9439,
|
| 5 |
+
"train_samples": 3079,
|
| 6 |
+
"train_samples_per_second": 1.204,
|
| 7 |
+
"train_steps_per_second": 0.151
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,663 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 385,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0,
|
| 13 |
+
"grad_norm": 39.25,
|
| 14 |
+
"learning_rate": 1.282051282051282e-07,
|
| 15 |
+
"logits/chosen": 88.18099975585938,
|
| 16 |
+
"logits/rejected": 88.25153350830078,
|
| 17 |
+
"logps/chosen": -29.073104858398438,
|
| 18 |
+
"logps/rejected": -26.25731658935547,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.0,
|
| 21 |
+
"rewards/chosen": 0.0,
|
| 22 |
+
"rewards/margins": 0.0,
|
| 23 |
+
"rewards/rejected": 0.0,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.03,
|
| 28 |
+
"grad_norm": 37.75,
|
| 29 |
+
"learning_rate": 1.282051282051282e-06,
|
| 30 |
+
"logits/chosen": 81.07136535644531,
|
| 31 |
+
"logits/rejected": 80.77804565429688,
|
| 32 |
+
"logps/chosen": -34.25458526611328,
|
| 33 |
+
"logps/rejected": -33.03440475463867,
|
| 34 |
+
"loss": 0.699,
|
| 35 |
+
"rewards/accuracies": 0.4444444477558136,
|
| 36 |
+
"rewards/chosen": -0.007714875973761082,
|
| 37 |
+
"rewards/margins": 0.03788409009575844,
|
| 38 |
+
"rewards/rejected": -0.045598965138196945,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.05,
|
| 43 |
+
"grad_norm": 26.25,
|
| 44 |
+
"learning_rate": 2.564102564102564e-06,
|
| 45 |
+
"logits/chosen": 80.65422058105469,
|
| 46 |
+
"logits/rejected": 80.54401397705078,
|
| 47 |
+
"logps/chosen": -33.63849639892578,
|
| 48 |
+
"logps/rejected": -30.794116973876953,
|
| 49 |
+
"loss": 0.708,
|
| 50 |
+
"rewards/accuracies": 0.512499988079071,
|
| 51 |
+
"rewards/chosen": 0.030845394358038902,
|
| 52 |
+
"rewards/margins": 0.04082341492176056,
|
| 53 |
+
"rewards/rejected": -0.009978031739592552,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.08,
|
| 58 |
+
"grad_norm": 38.25,
|
| 59 |
+
"learning_rate": 3.846153846153847e-06,
|
| 60 |
+
"logits/chosen": 82.5073013305664,
|
| 61 |
+
"logits/rejected": 82.5381088256836,
|
| 62 |
+
"logps/chosen": -33.88646697998047,
|
| 63 |
+
"logps/rejected": -31.181421279907227,
|
| 64 |
+
"loss": 0.7746,
|
| 65 |
+
"rewards/accuracies": 0.44999998807907104,
|
| 66 |
+
"rewards/chosen": 0.07581041753292084,
|
| 67 |
+
"rewards/margins": -0.06963472068309784,
|
| 68 |
+
"rewards/rejected": 0.14544512331485748,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.1,
|
| 73 |
+
"grad_norm": 31.625,
|
| 74 |
+
"learning_rate": 4.999896948438434e-06,
|
| 75 |
+
"logits/chosen": 81.06532287597656,
|
| 76 |
+
"logits/rejected": 81.06108093261719,
|
| 77 |
+
"logps/chosen": -32.81906509399414,
|
| 78 |
+
"logps/rejected": -33.26140594482422,
|
| 79 |
+
"loss": 0.6847,
|
| 80 |
+
"rewards/accuracies": 0.574999988079071,
|
| 81 |
+
"rewards/chosen": 0.21299926936626434,
|
| 82 |
+
"rewards/margins": 0.14872975647449493,
|
| 83 |
+
"rewards/rejected": 0.0642695277929306,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.13,
|
| 88 |
+
"grad_norm": 23.0,
|
| 89 |
+
"learning_rate": 4.987541037542187e-06,
|
| 90 |
+
"logits/chosen": 78.69737243652344,
|
| 91 |
+
"logits/rejected": 78.7103500366211,
|
| 92 |
+
"logps/chosen": -30.65850257873535,
|
| 93 |
+
"logps/rejected": -30.81766128540039,
|
| 94 |
+
"loss": 0.6962,
|
| 95 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 96 |
+
"rewards/chosen": 0.3280490040779114,
|
| 97 |
+
"rewards/margins": 0.17467446625232697,
|
| 98 |
+
"rewards/rejected": 0.1533745527267456,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.16,
|
| 103 |
+
"grad_norm": 31.625,
|
| 104 |
+
"learning_rate": 4.954691471941119e-06,
|
| 105 |
+
"logits/chosen": 83.20633697509766,
|
| 106 |
+
"logits/rejected": 83.25883483886719,
|
| 107 |
+
"logps/chosen": -30.961681365966797,
|
| 108 |
+
"logps/rejected": -29.538171768188477,
|
| 109 |
+
"loss": 0.703,
|
| 110 |
+
"rewards/accuracies": 0.550000011920929,
|
| 111 |
+
"rewards/chosen": 0.12808682024478912,
|
| 112 |
+
"rewards/margins": 0.09667714685201645,
|
| 113 |
+
"rewards/rejected": 0.03140967711806297,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.18,
|
| 118 |
+
"grad_norm": 53.25,
|
| 119 |
+
"learning_rate": 4.901618883413549e-06,
|
| 120 |
+
"logits/chosen": 83.81951141357422,
|
| 121 |
+
"logits/rejected": 83.84638977050781,
|
| 122 |
+
"logps/chosen": -30.67291259765625,
|
| 123 |
+
"logps/rejected": -33.11872482299805,
|
| 124 |
+
"loss": 0.755,
|
| 125 |
+
"rewards/accuracies": 0.4749999940395355,
|
| 126 |
+
"rewards/chosen": -0.026334354653954506,
|
| 127 |
+
"rewards/margins": 0.02227923832833767,
|
| 128 |
+
"rewards/rejected": -0.04861358925700188,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.21,
|
| 133 |
+
"grad_norm": 31.75,
|
| 134 |
+
"learning_rate": 4.828760511501322e-06,
|
| 135 |
+
"logits/chosen": 81.4664306640625,
|
| 136 |
+
"logits/rejected": 81.44920349121094,
|
| 137 |
+
"logps/chosen": -31.316049575805664,
|
| 138 |
+
"logps/rejected": -31.0085391998291,
|
| 139 |
+
"loss": 0.6446,
|
| 140 |
+
"rewards/accuracies": 0.550000011920929,
|
| 141 |
+
"rewards/chosen": 0.11333731561899185,
|
| 142 |
+
"rewards/margins": 0.2638704478740692,
|
| 143 |
+
"rewards/rejected": -0.15053315460681915,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.23,
|
| 148 |
+
"grad_norm": 37.0,
|
| 149 |
+
"learning_rate": 4.7367166013034295e-06,
|
| 150 |
+
"logits/chosen": 78.19766998291016,
|
| 151 |
+
"logits/rejected": 78.16535186767578,
|
| 152 |
+
"logps/chosen": -32.48051071166992,
|
| 153 |
+
"logps/rejected": -31.223648071289062,
|
| 154 |
+
"loss": 0.6567,
|
| 155 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 156 |
+
"rewards/chosen": 0.09460089355707169,
|
| 157 |
+
"rewards/margins": 0.2579067349433899,
|
| 158 |
+
"rewards/rejected": -0.1633058488368988,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.26,
|
| 163 |
+
"grad_norm": 31.5,
|
| 164 |
+
"learning_rate": 4.626245458345211e-06,
|
| 165 |
+
"logits/chosen": 83.43191528320312,
|
| 166 |
+
"logits/rejected": 83.45047760009766,
|
| 167 |
+
"logps/chosen": -34.02558135986328,
|
| 168 |
+
"logps/rejected": -31.787883758544922,
|
| 169 |
+
"loss": 0.6845,
|
| 170 |
+
"rewards/accuracies": 0.574999988079071,
|
| 171 |
+
"rewards/chosen": 0.16764816641807556,
|
| 172 |
+
"rewards/margins": 0.1900513470172882,
|
| 173 |
+
"rewards/rejected": -0.022403212264180183,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.26,
|
| 178 |
+
"eval_logits/chosen": 98.71414947509766,
|
| 179 |
+
"eval_logits/rejected": 98.70475769042969,
|
| 180 |
+
"eval_logps/chosen": -32.44282531738281,
|
| 181 |
+
"eval_logps/rejected": -36.040138244628906,
|
| 182 |
+
"eval_loss": 0.7398820519447327,
|
| 183 |
+
"eval_rewards/accuracies": 0.5245016813278198,
|
| 184 |
+
"eval_rewards/chosen": 0.00021068855130579323,
|
| 185 |
+
"eval_rewards/margins": 0.04437926039099693,
|
| 186 |
+
"eval_rewards/rejected": -0.04416857287287712,
|
| 187 |
+
"eval_runtime": 104.2075,
|
| 188 |
+
"eval_samples_per_second": 3.292,
|
| 189 |
+
"eval_steps_per_second": 0.413,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.29,
|
| 194 |
+
"grad_norm": 40.25,
|
| 195 |
+
"learning_rate": 4.498257201263691e-06,
|
| 196 |
+
"logits/chosen": 83.59847259521484,
|
| 197 |
+
"logits/rejected": 83.49092102050781,
|
| 198 |
+
"logps/chosen": -32.43052673339844,
|
| 199 |
+
"logps/rejected": -32.78325271606445,
|
| 200 |
+
"loss": 0.6135,
|
| 201 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 202 |
+
"rewards/chosen": 0.3553674817085266,
|
| 203 |
+
"rewards/margins": 0.43178611993789673,
|
| 204 |
+
"rewards/rejected": -0.07641863822937012,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.31,
|
| 209 |
+
"grad_norm": 46.5,
|
| 210 |
+
"learning_rate": 4.353806263777678e-06,
|
| 211 |
+
"logits/chosen": 83.7637710571289,
|
| 212 |
+
"logits/rejected": 83.87000274658203,
|
| 213 |
+
"logps/chosen": -28.259990692138672,
|
| 214 |
+
"logps/rejected": -35.35393524169922,
|
| 215 |
+
"loss": 0.6375,
|
| 216 |
+
"rewards/accuracies": 0.612500011920929,
|
| 217 |
+
"rewards/chosen": 0.40175461769104004,
|
| 218 |
+
"rewards/margins": 0.33862805366516113,
|
| 219 |
+
"rewards/rejected": 0.06312654912471771,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.34,
|
| 224 |
+
"grad_norm": 24.875,
|
| 225 |
+
"learning_rate": 4.1940827077152755e-06,
|
| 226 |
+
"logits/chosen": 80.89453125,
|
| 227 |
+
"logits/rejected": 80.9158706665039,
|
| 228 |
+
"logps/chosen": -30.432043075561523,
|
| 229 |
+
"logps/rejected": -32.080535888671875,
|
| 230 |
+
"loss": 0.6294,
|
| 231 |
+
"rewards/accuracies": 0.675000011920929,
|
| 232 |
+
"rewards/chosen": 0.2851874530315399,
|
| 233 |
+
"rewards/margins": 0.3745357096195221,
|
| 234 |
+
"rewards/rejected": -0.08934825658798218,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.36,
|
| 239 |
+
"grad_norm": 25.5,
|
| 240 |
+
"learning_rate": 4.0204024186666215e-06,
|
| 241 |
+
"logits/chosen": 82.0683822631836,
|
| 242 |
+
"logits/rejected": 82.07270812988281,
|
| 243 |
+
"logps/chosen": -27.02596092224121,
|
| 244 |
+
"logps/rejected": -33.121150970458984,
|
| 245 |
+
"loss": 0.5365,
|
| 246 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 247 |
+
"rewards/chosen": 0.2528177499771118,
|
| 248 |
+
"rewards/margins": 0.6714814305305481,
|
| 249 |
+
"rewards/rejected": -0.4186636805534363,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.39,
|
| 254 |
+
"grad_norm": 25.375,
|
| 255 |
+
"learning_rate": 3.834196265035119e-06,
|
| 256 |
+
"logits/chosen": 80.59815979003906,
|
| 257 |
+
"logits/rejected": 80.57023620605469,
|
| 258 |
+
"logps/chosen": -28.871845245361328,
|
| 259 |
+
"logps/rejected": -33.09119415283203,
|
| 260 |
+
"loss": 0.5456,
|
| 261 |
+
"rewards/accuracies": 0.699999988079071,
|
| 262 |
+
"rewards/chosen": 0.31036004424095154,
|
| 263 |
+
"rewards/margins": 0.6251744627952576,
|
| 264 |
+
"rewards/rejected": -0.3148145079612732,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.42,
|
| 269 |
+
"grad_norm": 44.25,
|
| 270 |
+
"learning_rate": 3.636998309800573e-06,
|
| 271 |
+
"logits/chosen": 82.46113586425781,
|
| 272 |
+
"logits/rejected": 82.46646118164062,
|
| 273 |
+
"logps/chosen": -33.629737854003906,
|
| 274 |
+
"logps/rejected": -30.432525634765625,
|
| 275 |
+
"loss": 0.6101,
|
| 276 |
+
"rewards/accuracies": 0.737500011920929,
|
| 277 |
+
"rewards/chosen": 0.30420786142349243,
|
| 278 |
+
"rewards/margins": 0.5921996235847473,
|
| 279 |
+
"rewards/rejected": -0.2879917025566101,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.44,
|
| 284 |
+
"grad_norm": 33.0,
|
| 285 |
+
"learning_rate": 3.4304331721118078e-06,
|
| 286 |
+
"logits/chosen": 83.26214599609375,
|
| 287 |
+
"logits/rejected": 83.21092224121094,
|
| 288 |
+
"logps/chosen": -30.77018165588379,
|
| 289 |
+
"logps/rejected": -32.57013702392578,
|
| 290 |
+
"loss": 0.573,
|
| 291 |
+
"rewards/accuracies": 0.6875,
|
| 292 |
+
"rewards/chosen": 0.2934645712375641,
|
| 293 |
+
"rewards/margins": 0.6235076189041138,
|
| 294 |
+
"rewards/rejected": -0.33004307746887207,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.47,
|
| 299 |
+
"grad_norm": 27.125,
|
| 300 |
+
"learning_rate": 3.2162026428305436e-06,
|
| 301 |
+
"logits/chosen": 80.83445739746094,
|
| 302 |
+
"logits/rejected": 80.81375885009766,
|
| 303 |
+
"logps/chosen": -30.401935577392578,
|
| 304 |
+
"logps/rejected": -31.623117446899414,
|
| 305 |
+
"loss": 0.5116,
|
| 306 |
+
"rewards/accuracies": 0.737500011920929,
|
| 307 |
+
"rewards/chosen": 0.4771292805671692,
|
| 308 |
+
"rewards/margins": 0.7566738724708557,
|
| 309 |
+
"rewards/rejected": -0.2795446515083313,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.49,
|
| 314 |
+
"grad_norm": 14.0,
|
| 315 |
+
"learning_rate": 2.996071664294641e-06,
|
| 316 |
+
"logits/chosen": 82.55574035644531,
|
| 317 |
+
"logits/rejected": 82.5384521484375,
|
| 318 |
+
"logps/chosen": -30.206974029541016,
|
| 319 |
+
"logps/rejected": -30.71441078186035,
|
| 320 |
+
"loss": 0.6219,
|
| 321 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 322 |
+
"rewards/chosen": 0.3356670141220093,
|
| 323 |
+
"rewards/margins": 0.4834977686405182,
|
| 324 |
+
"rewards/rejected": -0.14783072471618652,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.52,
|
| 329 |
+
"grad_norm": 15.375,
|
| 330 |
+
"learning_rate": 2.7718537898066833e-06,
|
| 331 |
+
"logits/chosen": 78.06065368652344,
|
| 332 |
+
"logits/rejected": 78.00971984863281,
|
| 333 |
+
"logps/chosen": -33.789581298828125,
|
| 334 |
+
"logps/rejected": -32.68096923828125,
|
| 335 |
+
"loss": 0.6115,
|
| 336 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 337 |
+
"rewards/chosen": 0.577893853187561,
|
| 338 |
+
"rewards/margins": 0.690432071685791,
|
| 339 |
+
"rewards/rejected": -0.11253812164068222,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.52,
|
| 344 |
+
"eval_logits/chosen": 98.62848663330078,
|
| 345 |
+
"eval_logits/rejected": 98.60428619384766,
|
| 346 |
+
"eval_logps/chosen": -32.65570068359375,
|
| 347 |
+
"eval_logps/rejected": -36.321441650390625,
|
| 348 |
+
"eval_loss": 0.7252821922302246,
|
| 349 |
+
"eval_rewards/accuracies": 0.530315637588501,
|
| 350 |
+
"eval_rewards/chosen": -0.12751542031764984,
|
| 351 |
+
"eval_rewards/margins": 0.08543363958597183,
|
| 352 |
+
"eval_rewards/rejected": -0.2129490226507187,
|
| 353 |
+
"eval_runtime": 103.8957,
|
| 354 |
+
"eval_samples_per_second": 3.301,
|
| 355 |
+
"eval_steps_per_second": 0.414,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.55,
|
| 360 |
+
"grad_norm": 52.0,
|
| 361 |
+
"learning_rate": 2.5453962426402006e-06,
|
| 362 |
+
"logits/chosen": 80.63914489746094,
|
| 363 |
+
"logits/rejected": 80.54652404785156,
|
| 364 |
+
"logps/chosen": -33.34014129638672,
|
| 365 |
+
"logps/rejected": -35.32052230834961,
|
| 366 |
+
"loss": 0.5935,
|
| 367 |
+
"rewards/accuracies": 0.737500011920929,
|
| 368 |
+
"rewards/chosen": 0.3633476793766022,
|
| 369 |
+
"rewards/margins": 0.5640031099319458,
|
| 370 |
+
"rewards/rejected": -0.20065537095069885,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.57,
|
| 375 |
+
"grad_norm": 19.625,
|
| 376 |
+
"learning_rate": 2.3185646976551794e-06,
|
| 377 |
+
"logits/chosen": 82.76437377929688,
|
| 378 |
+
"logits/rejected": 82.84717559814453,
|
| 379 |
+
"logps/chosen": -31.025707244873047,
|
| 380 |
+
"logps/rejected": -31.30951499938965,
|
| 381 |
+
"loss": 0.5027,
|
| 382 |
+
"rewards/accuracies": 0.762499988079071,
|
| 383 |
+
"rewards/chosen": 0.553946852684021,
|
| 384 |
+
"rewards/margins": 0.9022806286811829,
|
| 385 |
+
"rewards/rejected": -0.3483339250087738,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.6,
|
| 390 |
+
"grad_norm": 32.75,
|
| 391 |
+
"learning_rate": 2.0932279108998323e-06,
|
| 392 |
+
"logits/chosen": 79.89958190917969,
|
| 393 |
+
"logits/rejected": 79.95211791992188,
|
| 394 |
+
"logps/chosen": -32.34553146362305,
|
| 395 |
+
"logps/rejected": -34.391754150390625,
|
| 396 |
+
"loss": 0.6272,
|
| 397 |
+
"rewards/accuracies": 0.625,
|
| 398 |
+
"rewards/chosen": 0.2761251628398895,
|
| 399 |
+
"rewards/margins": 0.5038853287696838,
|
| 400 |
+
"rewards/rejected": -0.2277601659297943,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.62,
|
| 405 |
+
"grad_norm": 35.5,
|
| 406 |
+
"learning_rate": 1.8712423238279358e-06,
|
| 407 |
+
"logits/chosen": 82.25331115722656,
|
| 408 |
+
"logits/rejected": 82.53690338134766,
|
| 409 |
+
"logps/chosen": -30.6766357421875,
|
| 410 |
+
"logps/rejected": -31.96030044555664,
|
| 411 |
+
"loss": 0.4539,
|
| 412 |
+
"rewards/accuracies": 0.8125,
|
| 413 |
+
"rewards/chosen": 0.6068586111068726,
|
| 414 |
+
"rewards/margins": 0.8638145327568054,
|
| 415 |
+
"rewards/rejected": -0.25695592164993286,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.65,
|
| 420 |
+
"grad_norm": 30.5,
|
| 421 |
+
"learning_rate": 1.6544367689701824e-06,
|
| 422 |
+
"logits/chosen": 80.93089294433594,
|
| 423 |
+
"logits/rejected": 80.99276733398438,
|
| 424 |
+
"logps/chosen": -27.04372787475586,
|
| 425 |
+
"logps/rejected": -30.084264755249023,
|
| 426 |
+
"loss": 0.6593,
|
| 427 |
+
"rewards/accuracies": 0.574999988079071,
|
| 428 |
+
"rewards/chosen": 0.3313008248806,
|
| 429 |
+
"rewards/margins": 0.441417396068573,
|
| 430 |
+
"rewards/rejected": -0.11011654138565063,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.68,
|
| 435 |
+
"grad_norm": 29.125,
|
| 436 |
+
"learning_rate": 1.4445974030621963e-06,
|
| 437 |
+
"logits/chosen": 78.20941162109375,
|
| 438 |
+
"logits/rejected": 78.33964538574219,
|
| 439 |
+
"logps/chosen": -30.433767318725586,
|
| 440 |
+
"logps/rejected": -36.57436752319336,
|
| 441 |
+
"loss": 0.5,
|
| 442 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 443 |
+
"rewards/chosen": 0.6763362884521484,
|
| 444 |
+
"rewards/margins": 0.9599828720092773,
|
| 445 |
+
"rewards/rejected": -0.28364673256874084,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.7,
|
| 450 |
+
"grad_norm": 21.5,
|
| 451 |
+
"learning_rate": 1.243452991757889e-06,
|
| 452 |
+
"logits/chosen": 77.5750503540039,
|
| 453 |
+
"logits/rejected": 77.60489654541016,
|
| 454 |
+
"logps/chosen": -30.800561904907227,
|
| 455 |
+
"logps/rejected": -31.87221908569336,
|
| 456 |
+
"loss": 0.4973,
|
| 457 |
+
"rewards/accuracies": 0.762499988079071,
|
| 458 |
+
"rewards/chosen": 0.5870199203491211,
|
| 459 |
+
"rewards/margins": 0.807098388671875,
|
| 460 |
+
"rewards/rejected": -0.2200784683227539,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.73,
|
| 465 |
+
"grad_norm": 33.25,
|
| 466 |
+
"learning_rate": 1.0526606671603523e-06,
|
| 467 |
+
"logits/chosen": 80.28849029541016,
|
| 468 |
+
"logits/rejected": 80.06718444824219,
|
| 469 |
+
"logps/chosen": -31.078380584716797,
|
| 470 |
+
"logps/rejected": -29.8966007232666,
|
| 471 |
+
"loss": 0.5973,
|
| 472 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 473 |
+
"rewards/chosen": 0.4389079213142395,
|
| 474 |
+
"rewards/margins": 0.5766692757606506,
|
| 475 |
+
"rewards/rejected": -0.13776138424873352,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.75,
|
| 480 |
+
"grad_norm": 17.75,
|
| 481 |
+
"learning_rate": 8.737922755071455e-07,
|
| 482 |
+
"logits/chosen": 80.41847229003906,
|
| 483 |
+
"logits/rejected": 80.33303833007812,
|
| 484 |
+
"logps/chosen": -32.99018478393555,
|
| 485 |
+
"logps/rejected": -32.6365966796875,
|
| 486 |
+
"loss": 0.4458,
|
| 487 |
+
"rewards/accuracies": 0.762499988079071,
|
| 488 |
+
"rewards/chosen": 0.6684367656707764,
|
| 489 |
+
"rewards/margins": 1.0402761697769165,
|
| 490 |
+
"rewards/rejected": -0.37183937430381775,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.78,
|
| 495 |
+
"grad_norm": 34.25,
|
| 496 |
+
"learning_rate": 7.08321427484816e-07,
|
| 497 |
+
"logits/chosen": 76.02632141113281,
|
| 498 |
+
"logits/rejected": 76.11949920654297,
|
| 499 |
+
"logps/chosen": -32.25402069091797,
|
| 500 |
+
"logps/rejected": -29.283954620361328,
|
| 501 |
+
"loss": 0.5545,
|
| 502 |
+
"rewards/accuracies": 0.699999988079071,
|
| 503 |
+
"rewards/chosen": 0.6910130381584167,
|
| 504 |
+
"rewards/margins": 0.8001711964607239,
|
| 505 |
+
"rewards/rejected": -0.10915807634592056,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.78,
|
| 510 |
+
"eval_logits/chosen": 98.64820861816406,
|
| 511 |
+
"eval_logits/rejected": 98.62397003173828,
|
| 512 |
+
"eval_logps/chosen": -32.62586212158203,
|
| 513 |
+
"eval_logps/rejected": -36.32143783569336,
|
| 514 |
+
"eval_loss": 0.7248644828796387,
|
| 515 |
+
"eval_rewards/accuracies": 0.5282392501831055,
|
| 516 |
+
"eval_rewards/chosen": -0.10961288958787918,
|
| 517 |
+
"eval_rewards/margins": 0.10333485901355743,
|
| 518 |
+
"eval_rewards/rejected": -0.2129477560520172,
|
| 519 |
+
"eval_runtime": 104.0116,
|
| 520 |
+
"eval_samples_per_second": 3.298,
|
| 521 |
+
"eval_steps_per_second": 0.413,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.81,
|
| 526 |
+
"grad_norm": 27.625,
|
| 527 |
+
"learning_rate": 5.576113578589035e-07,
|
| 528 |
+
"logits/chosen": 83.13574981689453,
|
| 529 |
+
"logits/rejected": 83.16615295410156,
|
| 530 |
+
"logps/chosen": -29.959243774414062,
|
| 531 |
+
"logps/rejected": -32.55767059326172,
|
| 532 |
+
"loss": 0.5265,
|
| 533 |
+
"rewards/accuracies": 0.737500011920929,
|
| 534 |
+
"rewards/chosen": 0.542574405670166,
|
| 535 |
+
"rewards/margins": 0.7573197484016418,
|
| 536 |
+
"rewards/rejected": -0.21474528312683105,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.83,
|
| 541 |
+
"grad_norm": 21.625,
|
| 542 |
+
"learning_rate": 4.229036944380913e-07,
|
| 543 |
+
"logits/chosen": 80.65809631347656,
|
| 544 |
+
"logits/rejected": 80.65727233886719,
|
| 545 |
+
"logps/chosen": -30.505443572998047,
|
| 546 |
+
"logps/rejected": -29.11099624633789,
|
| 547 |
+
"loss": 0.5087,
|
| 548 |
+
"rewards/accuracies": 0.737500011920929,
|
| 549 |
+
"rewards/chosen": 0.6558700799942017,
|
| 550 |
+
"rewards/margins": 0.7707726359367371,
|
| 551 |
+
"rewards/rejected": -0.11490253359079361,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.86,
|
| 556 |
+
"grad_norm": 19.25,
|
| 557 |
+
"learning_rate": 3.053082288996112e-07,
|
| 558 |
+
"logits/chosen": 77.81159210205078,
|
| 559 |
+
"logits/rejected": 77.86034393310547,
|
| 560 |
+
"logps/chosen": -29.130138397216797,
|
| 561 |
+
"logps/rejected": -33.010986328125,
|
| 562 |
+
"loss": 0.4483,
|
| 563 |
+
"rewards/accuracies": 0.75,
|
| 564 |
+
"rewards/chosen": 0.7334806323051453,
|
| 565 |
+
"rewards/margins": 0.9591558575630188,
|
| 566 |
+
"rewards/rejected": -0.22567513585090637,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.88,
|
| 571 |
+
"grad_norm": 41.75,
|
| 572 |
+
"learning_rate": 2.0579377374915805e-07,
|
| 573 |
+
"logits/chosen": 82.1180648803711,
|
| 574 |
+
"logits/rejected": 82.14155578613281,
|
| 575 |
+
"logps/chosen": -32.119606018066406,
|
| 576 |
+
"logps/rejected": -33.77212905883789,
|
| 577 |
+
"loss": 0.5073,
|
| 578 |
+
"rewards/accuracies": 0.737500011920929,
|
| 579 |
+
"rewards/chosen": 0.6555252075195312,
|
| 580 |
+
"rewards/margins": 0.8975871801376343,
|
| 581 |
+
"rewards/rejected": -0.24206197261810303,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.91,
|
| 586 |
+
"grad_norm": 17.25,
|
| 587 |
+
"learning_rate": 1.2518018074041684e-07,
|
| 588 |
+
"logits/chosen": 81.12958526611328,
|
| 589 |
+
"logits/rejected": 81.1399154663086,
|
| 590 |
+
"logps/chosen": -32.4399299621582,
|
| 591 |
+
"logps/rejected": -33.30702590942383,
|
| 592 |
+
"loss": 0.5483,
|
| 593 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 594 |
+
"rewards/chosen": 0.7222862839698792,
|
| 595 |
+
"rewards/margins": 0.8514213562011719,
|
| 596 |
+
"rewards/rejected": -0.12913502752780914,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.94,
|
| 601 |
+
"grad_norm": 24.875,
|
| 602 |
+
"learning_rate": 6.41315865106129e-08,
|
| 603 |
+
"logits/chosen": 82.61198425292969,
|
| 604 |
+
"logits/rejected": 82.64558410644531,
|
| 605 |
+
"logps/chosen": -28.419490814208984,
|
| 606 |
+
"logps/rejected": -31.76764488220215,
|
| 607 |
+
"loss": 0.5254,
|
| 608 |
+
"rewards/accuracies": 0.6875,
|
| 609 |
+
"rewards/chosen": 0.6796320080757141,
|
| 610 |
+
"rewards/margins": 0.7503107786178589,
|
| 611 |
+
"rewards/rejected": -0.07067875564098358,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.96,
|
| 616 |
+
"grad_norm": 30.25,
|
| 617 |
+
"learning_rate": 2.3150941078050325e-08,
|
| 618 |
+
"logits/chosen": 82.08049774169922,
|
| 619 |
+
"logits/rejected": 82.0997543334961,
|
| 620 |
+
"logps/chosen": -31.871307373046875,
|
| 621 |
+
"logps/rejected": -35.636024475097656,
|
| 622 |
+
"loss": 0.5176,
|
| 623 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 624 |
+
"rewards/chosen": 0.6029146909713745,
|
| 625 |
+
"rewards/margins": 0.9189049601554871,
|
| 626 |
+
"rewards/rejected": -0.31599029898643494,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.99,
|
| 631 |
+
"grad_norm": 31.875,
|
| 632 |
+
"learning_rate": 2.575864278703266e-09,
|
| 633 |
+
"logits/chosen": 75.98027038574219,
|
| 634 |
+
"logits/rejected": 75.85136413574219,
|
| 635 |
+
"logps/chosen": -29.75612449645996,
|
| 636 |
+
"logps/rejected": -28.387653350830078,
|
| 637 |
+
"loss": 0.5513,
|
| 638 |
+
"rewards/accuracies": 0.699999988079071,
|
| 639 |
+
"rewards/chosen": 0.49200135469436646,
|
| 640 |
+
"rewards/margins": 0.6282498836517334,
|
| 641 |
+
"rewards/rejected": -0.13624855875968933,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 1.0,
|
| 646 |
+
"step": 385,
|
| 647 |
+
"total_flos": 0.0,
|
| 648 |
+
"train_loss": 0.5886486524111265,
|
| 649 |
+
"train_runtime": 2556.9439,
|
| 650 |
+
"train_samples_per_second": 1.204,
|
| 651 |
+
"train_steps_per_second": 0.151
|
| 652 |
+
}
|
| 653 |
+
],
|
| 654 |
+
"logging_steps": 10,
|
| 655 |
+
"max_steps": 385,
|
| 656 |
+
"num_input_tokens_seen": 0,
|
| 657 |
+
"num_train_epochs": 1,
|
| 658 |
+
"save_steps": 100,
|
| 659 |
+
"total_flos": 0.0,
|
| 660 |
+
"train_batch_size": 4,
|
| 661 |
+
"trial_name": null,
|
| 662 |
+
"trial_params": null
|
| 663 |
+
}
|