Model save
Browse files- README.md +79 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +739 -0
README.md
ADDED
|
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
model-index:
|
| 9 |
+
- name: zephyr-7b-dpo-full-prometheus-high-bleu-3-epochs
|
| 10 |
+
results: []
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 14 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 15 |
+
|
| 16 |
+
# zephyr-7b-dpo-full-prometheus-high-bleu-3-epochs
|
| 17 |
+
|
| 18 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.7540
|
| 21 |
+
- Rewards/chosen: -3.8924
|
| 22 |
+
- Rewards/rejected: -4.8360
|
| 23 |
+
- Rewards/accuracies: 0.6810
|
| 24 |
+
- Rewards/margins: 0.9436
|
| 25 |
+
- Logps/rejected: -731.8801
|
| 26 |
+
- Logps/chosen: -649.2000
|
| 27 |
+
- Logits/rejected: 1.8046
|
| 28 |
+
- Logits/chosen: 1.3830
|
| 29 |
+
|
| 30 |
+
## Model description
|
| 31 |
+
|
| 32 |
+
More information needed
|
| 33 |
+
|
| 34 |
+
## Intended uses & limitations
|
| 35 |
+
|
| 36 |
+
More information needed
|
| 37 |
+
|
| 38 |
+
## Training and evaluation data
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Training procedure
|
| 43 |
+
|
| 44 |
+
### Training hyperparameters
|
| 45 |
+
|
| 46 |
+
The following hyperparameters were used during training:
|
| 47 |
+
- learning_rate: 5e-07
|
| 48 |
+
- train_batch_size: 8
|
| 49 |
+
- eval_batch_size: 8
|
| 50 |
+
- seed: 55
|
| 51 |
+
- distributed_type: multi-GPU
|
| 52 |
+
- num_devices: 8
|
| 53 |
+
- gradient_accumulation_steps: 2
|
| 54 |
+
- total_train_batch_size: 128
|
| 55 |
+
- total_eval_batch_size: 64
|
| 56 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 57 |
+
- lr_scheduler_type: cosine
|
| 58 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 59 |
+
- num_epochs: 3
|
| 60 |
+
|
| 61 |
+
### Training results
|
| 62 |
+
|
| 63 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 64 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 65 |
+
| 0.6701 | 0.3802 | 50 | 0.6608 | -0.0605 | -0.1235 | 0.6379 | 0.0630 | -260.6238 | -266.0125 | -2.5430 | -2.5790 |
|
| 66 |
+
| 0.6369 | 0.7605 | 100 | 0.6256 | -0.4573 | -0.6699 | 0.6379 | 0.2125 | -315.2623 | -305.6931 | -2.2506 | -2.3145 |
|
| 67 |
+
| 0.4762 | 1.1407 | 150 | 0.6095 | -1.0277 | -1.3947 | 0.6638 | 0.3669 | -387.7436 | -362.7340 | -2.0713 | -2.1464 |
|
| 68 |
+
| 0.4416 | 1.5209 | 200 | 0.6303 | -1.5256 | -2.0301 | 0.6897 | 0.5044 | -451.2823 | -412.5244 | -1.8055 | -1.8831 |
|
| 69 |
+
| 0.4058 | 1.9011 | 250 | 0.6470 | -2.1413 | -2.7297 | 0.6724 | 0.5884 | -521.2467 | -474.0945 | -0.8046 | -0.9765 |
|
| 70 |
+
| 0.2288 | 2.2814 | 300 | 0.7265 | -3.4237 | -4.3014 | 0.6724 | 0.8777 | -678.4208 | -602.3348 | 1.1516 | 0.7332 |
|
| 71 |
+
| 0.21 | 2.6616 | 350 | 0.7540 | -3.8924 | -4.8360 | 0.6810 | 0.9436 | -731.8801 | -649.2000 | 1.8046 | 1.3830 |
|
| 72 |
+
|
| 73 |
+
|
| 74 |
+
### Framework versions
|
| 75 |
+
|
| 76 |
+
- Transformers 4.44.0.dev0
|
| 77 |
+
- Pytorch 2.1.2
|
| 78 |
+
- Datasets 2.20.0
|
| 79 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.988593155893536,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.440452757075846,
|
| 5 |
+
"train_runtime": 10425.4174,
|
| 6 |
+
"train_samples": 16800,
|
| 7 |
+
"train_samples_per_second": 4.834,
|
| 8 |
+
"train_steps_per_second": 0.038
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"transformers_version": "4.44.0.dev0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.988593155893536,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.440452757075846,
|
| 5 |
+
"train_runtime": 10425.4174,
|
| 6 |
+
"train_samples": 16800,
|
| 7 |
+
"train_samples_per_second": 4.834,
|
| 8 |
+
"train_steps_per_second": 0.038
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,739 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 2.988593155893536,
|
| 5 |
+
"eval_steps": 50,
|
| 6 |
+
"global_step": 393,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.07604562737642585,
|
| 13 |
+
"grad_norm": 8.750013078952279,
|
| 14 |
+
"learning_rate": 1.25e-07,
|
| 15 |
+
"logits/chosen": -2.757716655731201,
|
| 16 |
+
"logits/rejected": -2.75109601020813,
|
| 17 |
+
"logps/chosen": -260.8774719238281,
|
| 18 |
+
"logps/rejected": -271.0572814941406,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.42500001192092896,
|
| 21 |
+
"rewards/chosen": 5.311033601174131e-05,
|
| 22 |
+
"rewards/margins": -0.00021072864183224738,
|
| 23 |
+
"rewards/rejected": 0.0002638388832565397,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.1520912547528517,
|
| 28 |
+
"grad_norm": 7.9922636295097265,
|
| 29 |
+
"learning_rate": 2.5e-07,
|
| 30 |
+
"logits/chosen": -2.7602779865264893,
|
| 31 |
+
"logits/rejected": -2.7576041221618652,
|
| 32 |
+
"logps/chosen": -286.5116882324219,
|
| 33 |
+
"logps/rejected": -283.8603820800781,
|
| 34 |
+
"loss": 0.6926,
|
| 35 |
+
"rewards/accuracies": 0.53125,
|
| 36 |
+
"rewards/chosen": 0.0009672940941527486,
|
| 37 |
+
"rewards/margins": 0.0008663859916850924,
|
| 38 |
+
"rewards/rejected": 0.00010090797150041908,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.22813688212927757,
|
| 43 |
+
"grad_norm": 8.25949567460557,
|
| 44 |
+
"learning_rate": 3.75e-07,
|
| 45 |
+
"logits/chosen": -2.7091896533966064,
|
| 46 |
+
"logits/rejected": -2.707089900970459,
|
| 47 |
+
"logps/chosen": -268.6849670410156,
|
| 48 |
+
"logps/rejected": -269.9923400878906,
|
| 49 |
+
"loss": 0.6901,
|
| 50 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 51 |
+
"rewards/chosen": 0.01067989133298397,
|
| 52 |
+
"rewards/margins": 0.006726422347128391,
|
| 53 |
+
"rewards/rejected": 0.003953468985855579,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.3041825095057034,
|
| 58 |
+
"grad_norm": 7.8540741959718,
|
| 59 |
+
"learning_rate": 5e-07,
|
| 60 |
+
"logits/chosen": -2.7155396938323975,
|
| 61 |
+
"logits/rejected": -2.716829776763916,
|
| 62 |
+
"logps/chosen": -292.66705322265625,
|
| 63 |
+
"logps/rejected": -287.9569396972656,
|
| 64 |
+
"loss": 0.6838,
|
| 65 |
+
"rewards/accuracies": 0.59375,
|
| 66 |
+
"rewards/chosen": 0.03730246424674988,
|
| 67 |
+
"rewards/margins": 0.019583439454436302,
|
| 68 |
+
"rewards/rejected": 0.017719022929668427,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.38022813688212925,
|
| 73 |
+
"grad_norm": 8.712893470844076,
|
| 74 |
+
"learning_rate": 4.990105959637203e-07,
|
| 75 |
+
"logits/chosen": -2.688120126724243,
|
| 76 |
+
"logits/rejected": -2.6780331134796143,
|
| 77 |
+
"logps/chosen": -275.4263610839844,
|
| 78 |
+
"logps/rejected": -280.61688232421875,
|
| 79 |
+
"loss": 0.6701,
|
| 80 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 81 |
+
"rewards/chosen": 0.011463982984423637,
|
| 82 |
+
"rewards/margins": 0.03627394884824753,
|
| 83 |
+
"rewards/rejected": -0.024809962138533592,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.38022813688212925,
|
| 88 |
+
"eval_logits/chosen": -2.5790276527404785,
|
| 89 |
+
"eval_logits/rejected": -2.5429863929748535,
|
| 90 |
+
"eval_logps/chosen": -266.01251220703125,
|
| 91 |
+
"eval_logps/rejected": -260.623779296875,
|
| 92 |
+
"eval_loss": 0.6608449220657349,
|
| 93 |
+
"eval_rewards/accuracies": 0.6379310488700867,
|
| 94 |
+
"eval_rewards/chosen": -0.06051206588745117,
|
| 95 |
+
"eval_rewards/margins": 0.06295385211706161,
|
| 96 |
+
"eval_rewards/rejected": -0.12346591800451279,
|
| 97 |
+
"eval_runtime": 93.5809,
|
| 98 |
+
"eval_samples_per_second": 19.566,
|
| 99 |
+
"eval_steps_per_second": 0.31,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.45627376425855515,
|
| 104 |
+
"grad_norm": 9.48204753541487,
|
| 105 |
+
"learning_rate": 4.960502152176573e-07,
|
| 106 |
+
"logits/chosen": -2.6288909912109375,
|
| 107 |
+
"logits/rejected": -2.6260056495666504,
|
| 108 |
+
"logps/chosen": -305.30902099609375,
|
| 109 |
+
"logps/rejected": -323.2539367675781,
|
| 110 |
+
"loss": 0.6562,
|
| 111 |
+
"rewards/accuracies": 0.668749988079071,
|
| 112 |
+
"rewards/chosen": -0.08925994485616684,
|
| 113 |
+
"rewards/margins": 0.11333100497722626,
|
| 114 |
+
"rewards/rejected": -0.2025909423828125,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.532319391634981,
|
| 119 |
+
"grad_norm": 10.474383395672094,
|
| 120 |
+
"learning_rate": 4.911422898630837e-07,
|
| 121 |
+
"logits/chosen": -2.5303704738616943,
|
| 122 |
+
"logits/rejected": -2.5113658905029297,
|
| 123 |
+
"logps/chosen": -289.95233154296875,
|
| 124 |
+
"logps/rejected": -313.9202880859375,
|
| 125 |
+
"loss": 0.6516,
|
| 126 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 127 |
+
"rewards/chosen": -0.24351218342781067,
|
| 128 |
+
"rewards/margins": 0.08410634845495224,
|
| 129 |
+
"rewards/rejected": -0.3276185095310211,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.6083650190114068,
|
| 134 |
+
"grad_norm": 13.987853639442921,
|
| 135 |
+
"learning_rate": 4.84325667269244e-07,
|
| 136 |
+
"logits/chosen": -2.4371495246887207,
|
| 137 |
+
"logits/rejected": -2.4378981590270996,
|
| 138 |
+
"logps/chosen": -334.60919189453125,
|
| 139 |
+
"logps/rejected": -343.8853454589844,
|
| 140 |
+
"loss": 0.6459,
|
| 141 |
+
"rewards/accuracies": 0.625,
|
| 142 |
+
"rewards/chosen": -0.37069252133369446,
|
| 143 |
+
"rewards/margins": 0.14995035529136658,
|
| 144 |
+
"rewards/rejected": -0.520642876625061,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.6844106463878327,
|
| 149 |
+
"grad_norm": 13.701357719917418,
|
| 150 |
+
"learning_rate": 4.7565430258740336e-07,
|
| 151 |
+
"logits/chosen": -2.43092942237854,
|
| 152 |
+
"logits/rejected": -2.398475170135498,
|
| 153 |
+
"logps/chosen": -332.1542053222656,
|
| 154 |
+
"logps/rejected": -322.47540283203125,
|
| 155 |
+
"loss": 0.6262,
|
| 156 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 157 |
+
"rewards/chosen": -0.4322943687438965,
|
| 158 |
+
"rewards/margins": 0.17964716255664825,
|
| 159 |
+
"rewards/rejected": -0.6119415163993835,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.7604562737642585,
|
| 164 |
+
"grad_norm": 14.672531746982555,
|
| 165 |
+
"learning_rate": 4.6519683168329195e-07,
|
| 166 |
+
"logits/chosen": -2.4015495777130127,
|
| 167 |
+
"logits/rejected": -2.371605396270752,
|
| 168 |
+
"logps/chosen": -324.8261413574219,
|
| 169 |
+
"logps/rejected": -323.9524841308594,
|
| 170 |
+
"loss": 0.6369,
|
| 171 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 172 |
+
"rewards/chosen": -0.5212644338607788,
|
| 173 |
+
"rewards/margins": 0.19606027007102966,
|
| 174 |
+
"rewards/rejected": -0.7173247933387756,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.7604562737642585,
|
| 179 |
+
"eval_logits/chosen": -2.3145029544830322,
|
| 180 |
+
"eval_logits/rejected": -2.25056529045105,
|
| 181 |
+
"eval_logps/chosen": -305.69305419921875,
|
| 182 |
+
"eval_logps/rejected": -315.26226806640625,
|
| 183 |
+
"eval_loss": 0.6255849003791809,
|
| 184 |
+
"eval_rewards/accuracies": 0.6379310488700867,
|
| 185 |
+
"eval_rewards/chosen": -0.457317590713501,
|
| 186 |
+
"eval_rewards/margins": 0.21253280341625214,
|
| 187 |
+
"eval_rewards/rejected": -0.6698502898216248,
|
| 188 |
+
"eval_runtime": 94.9787,
|
| 189 |
+
"eval_samples_per_second": 19.278,
|
| 190 |
+
"eval_steps_per_second": 0.305,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.8365019011406845,
|
| 195 |
+
"grad_norm": 14.036217300902836,
|
| 196 |
+
"learning_rate": 4.530360278682841e-07,
|
| 197 |
+
"logits/chosen": -2.405102252960205,
|
| 198 |
+
"logits/rejected": -2.3873496055603027,
|
| 199 |
+
"logps/chosen": -297.6337890625,
|
| 200 |
+
"logps/rejected": -314.28936767578125,
|
| 201 |
+
"loss": 0.6236,
|
| 202 |
+
"rewards/accuracies": 0.75,
|
| 203 |
+
"rewards/chosen": -0.2411026507616043,
|
| 204 |
+
"rewards/margins": 0.2816271483898163,
|
| 205 |
+
"rewards/rejected": -0.522729754447937,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.9125475285171103,
|
| 210 |
+
"grad_norm": 18.73190288761437,
|
| 211 |
+
"learning_rate": 4.3926814672941166e-07,
|
| 212 |
+
"logits/chosen": -2.415482759475708,
|
| 213 |
+
"logits/rejected": -2.409000873565674,
|
| 214 |
+
"logps/chosen": -335.9554138183594,
|
| 215 |
+
"logps/rejected": -366.6188049316406,
|
| 216 |
+
"loss": 0.6232,
|
| 217 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 218 |
+
"rewards/chosen": -0.41517263650894165,
|
| 219 |
+
"rewards/margins": 0.29164889454841614,
|
| 220 |
+
"rewards/rejected": -0.7068215012550354,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.9885931558935361,
|
| 225 |
+
"grad_norm": 14.277531847256165,
|
| 226 |
+
"learning_rate": 4.240021642440332e-07,
|
| 227 |
+
"logits/chosen": -2.3934361934661865,
|
| 228 |
+
"logits/rejected": -2.3797760009765625,
|
| 229 |
+
"logps/chosen": -326.6287841796875,
|
| 230 |
+
"logps/rejected": -334.6332092285156,
|
| 231 |
+
"loss": 0.6044,
|
| 232 |
+
"rewards/accuracies": 0.65625,
|
| 233 |
+
"rewards/chosen": -0.5393946170806885,
|
| 234 |
+
"rewards/margins": 0.20507295429706573,
|
| 235 |
+
"rewards/rejected": -0.7444676160812378,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 1.064638783269962,
|
| 240 |
+
"grad_norm": 12.399932098202758,
|
| 241 |
+
"learning_rate": 4.073589142096592e-07,
|
| 242 |
+
"logits/chosen": -2.3695178031921387,
|
| 243 |
+
"logits/rejected": -2.362006425857544,
|
| 244 |
+
"logps/chosen": -363.754150390625,
|
| 245 |
+
"logps/rejected": -403.7615661621094,
|
| 246 |
+
"loss": 0.5188,
|
| 247 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 248 |
+
"rewards/chosen": -0.4886489808559418,
|
| 249 |
+
"rewards/margins": 0.4631797671318054,
|
| 250 |
+
"rewards/rejected": -0.9518287777900696,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 1.1406844106463878,
|
| 255 |
+
"grad_norm": 14.307776858677078,
|
| 256 |
+
"learning_rate": 3.8947013181637624e-07,
|
| 257 |
+
"logits/chosen": -2.2751259803771973,
|
| 258 |
+
"logits/rejected": -2.2668895721435547,
|
| 259 |
+
"logps/chosen": -347.95452880859375,
|
| 260 |
+
"logps/rejected": -414.0625915527344,
|
| 261 |
+
"loss": 0.4762,
|
| 262 |
+
"rewards/accuracies": 0.800000011920929,
|
| 263 |
+
"rewards/chosen": -0.5590722560882568,
|
| 264 |
+
"rewards/margins": 0.6314972043037415,
|
| 265 |
+
"rewards/rejected": -1.1905694007873535,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 1.1406844106463878,
|
| 270 |
+
"eval_logits/chosen": -2.146411180496216,
|
| 271 |
+
"eval_logits/rejected": -2.071317434310913,
|
| 272 |
+
"eval_logps/chosen": -362.7339782714844,
|
| 273 |
+
"eval_logps/rejected": -387.7436218261719,
|
| 274 |
+
"eval_loss": 0.6095313429832458,
|
| 275 |
+
"eval_rewards/accuracies": 0.6637930870056152,
|
| 276 |
+
"eval_rewards/chosen": -1.027726411819458,
|
| 277 |
+
"eval_rewards/margins": 0.3669382333755493,
|
| 278 |
+
"eval_rewards/rejected": -1.3946646451950073,
|
| 279 |
+
"eval_runtime": 92.9404,
|
| 280 |
+
"eval_samples_per_second": 19.701,
|
| 281 |
+
"eval_steps_per_second": 0.312,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 1.2167300380228137,
|
| 286 |
+
"grad_norm": 17.154986114362913,
|
| 287 |
+
"learning_rate": 3.7047741093221656e-07,
|
| 288 |
+
"logits/chosen": -2.259256601333618,
|
| 289 |
+
"logits/rejected": -2.2275631427764893,
|
| 290 |
+
"logps/chosen": -399.64996337890625,
|
| 291 |
+
"logps/rejected": -466.52056884765625,
|
| 292 |
+
"loss": 0.4675,
|
| 293 |
+
"rewards/accuracies": 0.824999988079071,
|
| 294 |
+
"rewards/chosen": -0.9498268961906433,
|
| 295 |
+
"rewards/margins": 0.8195298910140991,
|
| 296 |
+
"rewards/rejected": -1.7693569660186768,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 1.2927756653992395,
|
| 301 |
+
"grad_norm": 17.94492731274679,
|
| 302 |
+
"learning_rate": 3.5053108335480205e-07,
|
| 303 |
+
"logits/chosen": -2.110236406326294,
|
| 304 |
+
"logits/rejected": -2.0996217727661133,
|
| 305 |
+
"logps/chosen": -406.27825927734375,
|
| 306 |
+
"logps/rejected": -483.40478515625,
|
| 307 |
+
"loss": 0.4492,
|
| 308 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 309 |
+
"rewards/chosen": -1.0264767408370972,
|
| 310 |
+
"rewards/margins": 0.876954197883606,
|
| 311 |
+
"rewards/rejected": -1.903430700302124,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 1.3688212927756653,
|
| 316 |
+
"grad_norm": 22.10036946673983,
|
| 317 |
+
"learning_rate": 3.29789028900245e-07,
|
| 318 |
+
"logits/chosen": -2.075085163116455,
|
| 319 |
+
"logits/rejected": -2.0645432472229004,
|
| 320 |
+
"logps/chosen": -446.86572265625,
|
| 321 |
+
"logps/rejected": -529.196044921875,
|
| 322 |
+
"loss": 0.4381,
|
| 323 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 324 |
+
"rewards/chosen": -1.4222126007080078,
|
| 325 |
+
"rewards/margins": 0.9729466438293457,
|
| 326 |
+
"rewards/rejected": -2.3951592445373535,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 1.4448669201520912,
|
| 331 |
+
"grad_norm": 22.48679557744493,
|
| 332 |
+
"learning_rate": 3.084154257477301e-07,
|
| 333 |
+
"logits/chosen": -2.159637928009033,
|
| 334 |
+
"logits/rejected": -2.144123077392578,
|
| 335 |
+
"logps/chosen": -454.2430114746094,
|
| 336 |
+
"logps/rejected": -546.5646362304688,
|
| 337 |
+
"loss": 0.4343,
|
| 338 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 339 |
+
"rewards/chosen": -1.5697638988494873,
|
| 340 |
+
"rewards/margins": 0.8866235017776489,
|
| 341 |
+
"rewards/rejected": -2.456387519836426,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 1.5209125475285172,
|
| 346 |
+
"grad_norm": 19.33120207165494,
|
| 347 |
+
"learning_rate": 2.865794509310888e-07,
|
| 348 |
+
"logits/chosen": -1.9821497201919556,
|
| 349 |
+
"logits/rejected": -1.9777543544769287,
|
| 350 |
+
"logps/chosen": -427.1321716308594,
|
| 351 |
+
"logps/rejected": -510.4873046875,
|
| 352 |
+
"loss": 0.4416,
|
| 353 |
+
"rewards/accuracies": 0.824999988079071,
|
| 354 |
+
"rewards/chosen": -1.4423638582229614,
|
| 355 |
+
"rewards/margins": 0.8751496076583862,
|
| 356 |
+
"rewards/rejected": -2.3175134658813477,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 1.5209125475285172,
|
| 361 |
+
"eval_logits/chosen": -1.8830668926239014,
|
| 362 |
+
"eval_logits/rejected": -1.8054617643356323,
|
| 363 |
+
"eval_logps/chosen": -412.5243835449219,
|
| 364 |
+
"eval_logps/rejected": -451.2822570800781,
|
| 365 |
+
"eval_loss": 0.6302618384361267,
|
| 366 |
+
"eval_rewards/accuracies": 0.6896551847457886,
|
| 367 |
+
"eval_rewards/chosen": -1.525631070137024,
|
| 368 |
+
"eval_rewards/margins": 0.5044201612472534,
|
| 369 |
+
"eval_rewards/rejected": -2.0300514698028564,
|
| 370 |
+
"eval_runtime": 94.2551,
|
| 371 |
+
"eval_samples_per_second": 19.426,
|
| 372 |
+
"eval_steps_per_second": 0.308,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 1.5969581749049429,
|
| 377 |
+
"grad_norm": 21.341696689301088,
|
| 378 |
+
"learning_rate": 2.644539412632786e-07,
|
| 379 |
+
"logits/chosen": -2.0418550968170166,
|
| 380 |
+
"logits/rejected": -1.9922387599945068,
|
| 381 |
+
"logps/chosen": -456.74951171875,
|
| 382 |
+
"logps/rejected": -536.8170166015625,
|
| 383 |
+
"loss": 0.4216,
|
| 384 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 385 |
+
"rewards/chosen": -1.3003852367401123,
|
| 386 |
+
"rewards/margins": 1.0898964405059814,
|
| 387 |
+
"rewards/rejected": -2.390282154083252,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 1.673003802281369,
|
| 392 |
+
"grad_norm": 24.250284409575745,
|
| 393 |
+
"learning_rate": 2.422140252928601e-07,
|
| 394 |
+
"logits/chosen": -1.964264154434204,
|
| 395 |
+
"logits/rejected": -1.914783239364624,
|
| 396 |
+
"logps/chosen": -446.67950439453125,
|
| 397 |
+
"logps/rejected": -542.49169921875,
|
| 398 |
+
"loss": 0.4058,
|
| 399 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 400 |
+
"rewards/chosen": -1.5265235900878906,
|
| 401 |
+
"rewards/margins": 1.0695818662643433,
|
| 402 |
+
"rewards/rejected": -2.5961055755615234,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 1.7490494296577945,
|
| 407 |
+
"grad_norm": 19.00774265939697,
|
| 408 |
+
"learning_rate": 2.2003573712085455e-07,
|
| 409 |
+
"logits/chosen": -1.8126968145370483,
|
| 410 |
+
"logits/rejected": -1.7982728481292725,
|
| 411 |
+
"logps/chosen": -407.3929748535156,
|
| 412 |
+
"logps/rejected": -507.03704833984375,
|
| 413 |
+
"loss": 0.4308,
|
| 414 |
+
"rewards/accuracies": 0.824999988079071,
|
| 415 |
+
"rewards/chosen": -1.3489201068878174,
|
| 416 |
+
"rewards/margins": 0.9808903932571411,
|
| 417 |
+
"rewards/rejected": -2.329810619354248,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 1.8250950570342206,
|
| 422 |
+
"grad_norm": 22.60307202010589,
|
| 423 |
+
"learning_rate": 1.980946230499431e-07,
|
| 424 |
+
"logits/chosen": -1.6764023303985596,
|
| 425 |
+
"logits/rejected": -1.666276216506958,
|
| 426 |
+
"logps/chosen": -441.1431579589844,
|
| 427 |
+
"logps/rejected": -543.8973388671875,
|
| 428 |
+
"loss": 0.4012,
|
| 429 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 430 |
+
"rewards/chosen": -1.6486518383026123,
|
| 431 |
+
"rewards/margins": 1.0365753173828125,
|
| 432 |
+
"rewards/rejected": -2.685227155685425,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 1.9011406844106464,
|
| 437 |
+
"grad_norm": 21.996162773488756,
|
| 438 |
+
"learning_rate": 1.7656435209470376e-07,
|
| 439 |
+
"logits/chosen": -1.4262855052947998,
|
| 440 |
+
"logits/rejected": -1.4329298734664917,
|
| 441 |
+
"logps/chosen": -472.7010803222656,
|
| 442 |
+
"logps/rejected": -587.4404296875,
|
| 443 |
+
"loss": 0.4058,
|
| 444 |
+
"rewards/accuracies": 0.8125,
|
| 445 |
+
"rewards/chosen": -1.8240299224853516,
|
| 446 |
+
"rewards/margins": 1.0630748271942139,
|
| 447 |
+
"rewards/rejected": -2.8871047496795654,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 1.9011406844106464,
|
| 452 |
+
"eval_logits/chosen": -0.976457953453064,
|
| 453 |
+
"eval_logits/rejected": -0.8045514225959778,
|
| 454 |
+
"eval_logps/chosen": -474.0944519042969,
|
| 455 |
+
"eval_logps/rejected": -521.2467041015625,
|
| 456 |
+
"eval_loss": 0.6470248103141785,
|
| 457 |
+
"eval_rewards/accuracies": 0.6724137663841248,
|
| 458 |
+
"eval_rewards/chosen": -2.141331672668457,
|
| 459 |
+
"eval_rewards/margins": 0.5883632898330688,
|
| 460 |
+
"eval_rewards/rejected": -2.7296955585479736,
|
| 461 |
+
"eval_runtime": 93.8915,
|
| 462 |
+
"eval_samples_per_second": 19.501,
|
| 463 |
+
"eval_steps_per_second": 0.309,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 1.9771863117870723,
|
| 468 |
+
"grad_norm": 26.930679059566526,
|
| 469 |
+
"learning_rate": 1.5561534135101884e-07,
|
| 470 |
+
"logits/chosen": -1.0850579738616943,
|
| 471 |
+
"logits/rejected": -0.9344271421432495,
|
| 472 |
+
"logps/chosen": -493.2483825683594,
|
| 473 |
+
"logps/rejected": -594.302734375,
|
| 474 |
+
"loss": 0.4126,
|
| 475 |
+
"rewards/accuracies": 0.856249988079071,
|
| 476 |
+
"rewards/chosen": -1.9806413650512695,
|
| 477 |
+
"rewards/margins": 1.1219998598098755,
|
| 478 |
+
"rewards/rejected": -3.1026411056518555,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 2.053231939163498,
|
| 483 |
+
"grad_norm": 31.653928530377552,
|
| 484 |
+
"learning_rate": 1.3541340710517546e-07,
|
| 485 |
+
"logits/chosen": -0.8464574813842773,
|
| 486 |
+
"logits/rejected": -0.7507003545761108,
|
| 487 |
+
"logps/chosen": -435.60260009765625,
|
| 488 |
+
"logps/rejected": -586.5030517578125,
|
| 489 |
+
"loss": 0.3086,
|
| 490 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 491 |
+
"rewards/chosen": -1.743119239807129,
|
| 492 |
+
"rewards/margins": 1.482635259628296,
|
| 493 |
+
"rewards/rejected": -3.225754499435425,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 2.129277566539924,
|
| 498 |
+
"grad_norm": 19.6643061541321,
|
| 499 |
+
"learning_rate": 1.1611845235944143e-07,
|
| 500 |
+
"logits/chosen": -0.4659014642238617,
|
| 501 |
+
"logits/rejected": -0.35695117712020874,
|
| 502 |
+
"logps/chosen": -510.74456787109375,
|
| 503 |
+
"logps/rejected": -693.2734985351562,
|
| 504 |
+
"loss": 0.2494,
|
| 505 |
+
"rewards/accuracies": 0.9375,
|
| 506 |
+
"rewards/chosen": -2.120767116546631,
|
| 507 |
+
"rewards/margins": 1.895451307296753,
|
| 508 |
+
"rewards/rejected": -4.016218662261963,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 2.20532319391635,
|
| 513 |
+
"grad_norm": 24.595513209110543,
|
| 514 |
+
"learning_rate": 9.788320116265892e-08,
|
| 515 |
+
"logits/chosen": -0.13620242476463318,
|
| 516 |
+
"logits/rejected": 0.04458779841661453,
|
| 517 |
+
"logps/chosen": -530.4153442382812,
|
| 518 |
+
"logps/rejected": -696.87109375,
|
| 519 |
+
"loss": 0.2381,
|
| 520 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 521 |
+
"rewards/chosen": -2.3323657512664795,
|
| 522 |
+
"rewards/margins": 1.861975073814392,
|
| 523 |
+
"rewards/rejected": -4.194340705871582,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 2.2813688212927756,
|
| 528 |
+
"grad_norm": 21.633771054738194,
|
| 529 |
+
"learning_rate": 8.085198976392124e-08,
|
| 530 |
+
"logits/chosen": 0.2940121591091156,
|
| 531 |
+
"logits/rejected": 0.4243450164794922,
|
| 532 |
+
"logps/chosen": -531.2306518554688,
|
| 533 |
+
"logps/rejected": -713.4442138671875,
|
| 534 |
+
"loss": 0.2288,
|
| 535 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 536 |
+
"rewards/chosen": -2.508195400238037,
|
| 537 |
+
"rewards/margins": 2.0444493293762207,
|
| 538 |
+
"rewards/rejected": -4.552644729614258,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 2.2813688212927756,
|
| 543 |
+
"eval_logits/chosen": 0.7331591844558716,
|
| 544 |
+
"eval_logits/rejected": 1.1516082286834717,
|
| 545 |
+
"eval_logps/chosen": -602.3347778320312,
|
| 546 |
+
"eval_logps/rejected": -678.4207763671875,
|
| 547 |
+
"eval_loss": 0.7265122532844543,
|
| 548 |
+
"eval_rewards/accuracies": 0.6724137663841248,
|
| 549 |
+
"eval_rewards/chosen": -3.4237349033355713,
|
| 550 |
+
"eval_rewards/margins": 0.877700686454773,
|
| 551 |
+
"eval_rewards/rejected": -4.301435470581055,
|
| 552 |
+
"eval_runtime": 94.112,
|
| 553 |
+
"eval_samples_per_second": 19.456,
|
| 554 |
+
"eval_steps_per_second": 0.308,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 2.3574144486692017,
|
| 559 |
+
"grad_norm": 25.495425543277612,
|
| 560 |
+
"learning_rate": 6.515962415763369e-08,
|
| 561 |
+
"logits/chosen": 0.6170581579208374,
|
| 562 |
+
"logits/rejected": 0.7346482276916504,
|
| 563 |
+
"logps/chosen": -559.1423950195312,
|
| 564 |
+
"logps/rejected": -797.1380615234375,
|
| 565 |
+
"loss": 0.2308,
|
| 566 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 567 |
+
"rewards/chosen": -2.808534860610962,
|
| 568 |
+
"rewards/margins": 2.294363021850586,
|
| 569 |
+
"rewards/rejected": -5.102897644042969,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 2.4334600760456273,
|
| 574 |
+
"grad_norm": 26.06035826594137,
|
| 575 |
+
"learning_rate": 5.093031306275308e-08,
|
| 576 |
+
"logits/chosen": 0.7691527009010315,
|
| 577 |
+
"logits/rejected": 1.0421297550201416,
|
| 578 |
+
"logps/chosen": -611.4459228515625,
|
| 579 |
+
"logps/rejected": -806.7598876953125,
|
| 580 |
+
"loss": 0.2207,
|
| 581 |
+
"rewards/accuracies": 0.918749988079071,
|
| 582 |
+
"rewards/chosen": -3.107365369796753,
|
| 583 |
+
"rewards/margins": 2.2186014652252197,
|
| 584 |
+
"rewards/rejected": -5.325966835021973,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 2.5095057034220534,
|
| 589 |
+
"grad_norm": 22.38621742753542,
|
| 590 |
+
"learning_rate": 3.827668478192578e-08,
|
| 591 |
+
"logits/chosen": 0.8044630289077759,
|
| 592 |
+
"logits/rejected": 0.9414040446281433,
|
| 593 |
+
"logps/chosen": -576.51708984375,
|
| 594 |
+
"logps/rejected": -825.3946533203125,
|
| 595 |
+
"loss": 0.2206,
|
| 596 |
+
"rewards/accuracies": 0.9375,
|
| 597 |
+
"rewards/chosen": -3.0663881301879883,
|
| 598 |
+
"rewards/margins": 2.4700827598571777,
|
| 599 |
+
"rewards/rejected": -5.536470890045166,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 2.585551330798479,
|
| 604 |
+
"grad_norm": 24.44232863700222,
|
| 605 |
+
"learning_rate": 2.729889572230856e-08,
|
| 606 |
+
"logits/chosen": 0.9481533765792847,
|
| 607 |
+
"logits/rejected": 1.2326308488845825,
|
| 608 |
+
"logps/chosen": -563.4014282226562,
|
| 609 |
+
"logps/rejected": -779.8926391601562,
|
| 610 |
+
"loss": 0.2173,
|
| 611 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 612 |
+
"rewards/chosen": -2.942201614379883,
|
| 613 |
+
"rewards/margins": 2.315051555633545,
|
| 614 |
+
"rewards/rejected": -5.257253170013428,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 2.661596958174905,
|
| 619 |
+
"grad_norm": 25.605389790591452,
|
| 620 |
+
"learning_rate": 1.8083837634341766e-08,
|
| 621 |
+
"logits/chosen": 0.5146993398666382,
|
| 622 |
+
"logits/rejected": 0.7431297302246094,
|
| 623 |
+
"logps/chosen": -602.69482421875,
|
| 624 |
+
"logps/rejected": -831.6427001953125,
|
| 625 |
+
"loss": 0.21,
|
| 626 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 627 |
+
"rewards/chosen": -3.007021427154541,
|
| 628 |
+
"rewards/margins": 2.4290623664855957,
|
| 629 |
+
"rewards/rejected": -5.436083793640137,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 2.661596958174905,
|
| 634 |
+
"eval_logits/chosen": 1.3830486536026,
|
| 635 |
+
"eval_logits/rejected": 1.804565668106079,
|
| 636 |
+
"eval_logps/chosen": -649.2000122070312,
|
| 637 |
+
"eval_logps/rejected": -731.8800659179688,
|
| 638 |
+
"eval_loss": 0.7540475726127625,
|
| 639 |
+
"eval_rewards/accuracies": 0.681034505367279,
|
| 640 |
+
"eval_rewards/chosen": -3.8923871517181396,
|
| 641 |
+
"eval_rewards/margins": 0.9436419010162354,
|
| 642 |
+
"eval_rewards/rejected": -4.836028575897217,
|
| 643 |
+
"eval_runtime": 92.8895,
|
| 644 |
+
"eval_samples_per_second": 19.712,
|
| 645 |
+
"eval_steps_per_second": 0.312,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 2.7376425855513307,
|
| 650 |
+
"grad_norm": 33.51553576619823,
|
| 651 |
+
"learning_rate": 1.0704449843359498e-08,
|
| 652 |
+
"logits/chosen": 0.9970355033874512,
|
| 653 |
+
"logits/rejected": 1.2489699125289917,
|
| 654 |
+
"logps/chosen": -617.2589721679688,
|
| 655 |
+
"logps/rejected": -876.3328857421875,
|
| 656 |
+
"loss": 0.2347,
|
| 657 |
+
"rewards/accuracies": 0.918749988079071,
|
| 658 |
+
"rewards/chosen": -3.3343372344970703,
|
| 659 |
+
"rewards/margins": 2.5600481033325195,
|
| 660 |
+
"rewards/rejected": -5.894384860992432,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 2.8136882129277567,
|
| 665 |
+
"grad_norm": 28.342143525311826,
|
| 666 |
+
"learning_rate": 5.2191419178871935e-09,
|
| 667 |
+
"logits/chosen": 0.9573332071304321,
|
| 668 |
+
"logits/rejected": 0.9704620242118835,
|
| 669 |
+
"logps/chosen": -602.1446533203125,
|
| 670 |
+
"logps/rejected": -826.4749145507812,
|
| 671 |
+
"loss": 0.2285,
|
| 672 |
+
"rewards/accuracies": 0.949999988079071,
|
| 673 |
+
"rewards/chosen": -3.1017754077911377,
|
| 674 |
+
"rewards/margins": 2.2921016216278076,
|
| 675 |
+
"rewards/rejected": -5.3938775062561035,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 2.8897338403041823,
|
| 680 |
+
"grad_norm": 25.89340425260963,
|
| 681 |
+
"learning_rate": 1.6713313443384724e-09,
|
| 682 |
+
"logits/chosen": 1.0670582056045532,
|
| 683 |
+
"logits/rejected": 1.103432536125183,
|
| 684 |
+
"logps/chosen": -586.4422607421875,
|
| 685 |
+
"logps/rejected": -816.7635498046875,
|
| 686 |
+
"loss": 0.2255,
|
| 687 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 688 |
+
"rewards/chosen": -3.018024444580078,
|
| 689 |
+
"rewards/margins": 2.285562753677368,
|
| 690 |
+
"rewards/rejected": -5.303586959838867,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 2.9657794676806084,
|
| 695 |
+
"grad_norm": 24.29095712697487,
|
| 696 |
+
"learning_rate": 8.909986752470012e-11,
|
| 697 |
+
"logits/chosen": 0.6790753602981567,
|
| 698 |
+
"logits/rejected": 0.9370797872543335,
|
| 699 |
+
"logps/chosen": -650.6864624023438,
|
| 700 |
+
"logps/rejected": -848.3294067382812,
|
| 701 |
+
"loss": 0.2245,
|
| 702 |
+
"rewards/accuracies": 0.981249988079071,
|
| 703 |
+
"rewards/chosen": -3.326509475708008,
|
| 704 |
+
"rewards/margins": 2.2717742919921875,
|
| 705 |
+
"rewards/rejected": -5.5982842445373535,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 2.988593155893536,
|
| 710 |
+
"step": 393,
|
| 711 |
+
"total_flos": 0.0,
|
| 712 |
+
"train_loss": 0.440452757075846,
|
| 713 |
+
"train_runtime": 10425.4174,
|
| 714 |
+
"train_samples_per_second": 4.834,
|
| 715 |
+
"train_steps_per_second": 0.038
|
| 716 |
+
}
|
| 717 |
+
],
|
| 718 |
+
"logging_steps": 10,
|
| 719 |
+
"max_steps": 393,
|
| 720 |
+
"num_input_tokens_seen": 0,
|
| 721 |
+
"num_train_epochs": 3,
|
| 722 |
+
"save_steps": 100,
|
| 723 |
+
"stateful_callbacks": {
|
| 724 |
+
"TrainerControl": {
|
| 725 |
+
"args": {
|
| 726 |
+
"should_epoch_stop": false,
|
| 727 |
+
"should_evaluate": false,
|
| 728 |
+
"should_log": false,
|
| 729 |
+
"should_save": true,
|
| 730 |
+
"should_training_stop": true
|
| 731 |
+
},
|
| 732 |
+
"attributes": {}
|
| 733 |
+
}
|
| 734 |
+
},
|
| 735 |
+
"total_flos": 0.0,
|
| 736 |
+
"train_batch_size": 8,
|
| 737 |
+
"trial_name": null,
|
| 738 |
+
"trial_params": null
|
| 739 |
+
}
|