Model save
Browse files- README.md +77 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +826 -0
README.md
ADDED
|
@@ -0,0 +1,77 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- dpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: rlced_conifer_zephyr-7b-dpo-full
|
| 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 |
+
# rlced_conifer_zephyr-7b-dpo-full
|
| 18 |
+
|
| 19 |
+
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.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.2007
|
| 22 |
+
- Rewards/chosen: -4.2176
|
| 23 |
+
- Rewards/rejected: -11.7584
|
| 24 |
+
- Rewards/accuracies: 0.8650
|
| 25 |
+
- Rewards/margins: 7.5408
|
| 26 |
+
- Logps/rejected: -1544.5891
|
| 27 |
+
- Logps/chosen: -781.9304
|
| 28 |
+
- Logits/rejected: -1.6459
|
| 29 |
+
- Logits/chosen: -2.0690
|
| 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-07
|
| 49 |
+
- train_batch_size: 8
|
| 50 |
+
- eval_batch_size: 8
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 8
|
| 54 |
+
- gradient_accumulation_steps: 4
|
| 55 |
+
- total_train_batch_size: 256
|
| 56 |
+
- total_eval_batch_size: 64
|
| 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 | Logits/rejected | Logits/chosen |
|
| 65 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 66 |
+
| 0.2558 | 0.2107 | 100 | 0.2515 | -3.3402 | -7.6336 | 0.8363 | 4.2934 | -1132.1079 | -694.1878 | -3.1129 | -3.1245 |
|
| 67 |
+
| 0.2204 | 0.4215 | 200 | 0.2260 | -3.7625 | -9.3814 | 0.8587 | 5.6190 | -1306.8950 | -736.4189 | -2.6948 | -2.7329 |
|
| 68 |
+
| 0.204 | 0.6322 | 300 | 0.2096 | -3.3959 | -9.7494 | 0.8650 | 6.3535 | -1343.6858 | -699.7554 | -2.1723 | -2.4109 |
|
| 69 |
+
| 0.1992 | 0.8430 | 400 | 0.2007 | -4.2176 | -11.7584 | 0.8650 | 7.5408 | -1544.5891 | -781.9304 | -1.6459 | -2.0690 |
|
| 70 |
+
|
| 71 |
+
|
| 72 |
+
### Framework versions
|
| 73 |
+
|
| 74 |
+
- Transformers 4.44.1
|
| 75 |
+
- Pytorch 2.1.2+cu121
|
| 76 |
+
- Datasets 2.21.0
|
| 77 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9989462592202318,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.26123517437323235,
|
| 5 |
+
"train_runtime": 15193.8056,
|
| 6 |
+
"train_samples": 121428,
|
| 7 |
+
"train_samples_per_second": 7.992,
|
| 8 |
+
"train_steps_per_second": 0.031
|
| 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": "4.44.1"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9989462592202318,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.26123517437323235,
|
| 5 |
+
"train_runtime": 15193.8056,
|
| 6 |
+
"train_samples": 121428,
|
| 7 |
+
"train_samples_per_second": 7.992,
|
| 8 |
+
"train_steps_per_second": 0.031
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,826 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9989462592202318,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 474,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.002107481559536354,
|
| 13 |
+
"grad_norm": 6.614315138552694,
|
| 14 |
+
"learning_rate": 1.0416666666666666e-08,
|
| 15 |
+
"logits/chosen": -2.9142751693725586,
|
| 16 |
+
"logits/rejected": -2.8893983364105225,
|
| 17 |
+
"logps/chosen": -285.5545654296875,
|
| 18 |
+
"logps/rejected": -301.5740966796875,
|
| 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": 0.02107481559536354,
|
| 28 |
+
"grad_norm": 6.815845823783859,
|
| 29 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 30 |
+
"logits/chosen": -2.927753448486328,
|
| 31 |
+
"logits/rejected": -2.894536018371582,
|
| 32 |
+
"logps/chosen": -286.5064392089844,
|
| 33 |
+
"logps/rejected": -292.89776611328125,
|
| 34 |
+
"loss": 0.693,
|
| 35 |
+
"rewards/accuracies": 0.4548611044883728,
|
| 36 |
+
"rewards/chosen": -8.25391907710582e-05,
|
| 37 |
+
"rewards/margins": 0.0003568828688003123,
|
| 38 |
+
"rewards/rejected": -0.00043942214688286185,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.04214963119072708,
|
| 43 |
+
"grad_norm": 7.858560008275229,
|
| 44 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 45 |
+
"logits/chosen": -2.9354426860809326,
|
| 46 |
+
"logits/rejected": -2.9020581245422363,
|
| 47 |
+
"logps/chosen": -306.3351135253906,
|
| 48 |
+
"logps/rejected": -318.660888671875,
|
| 49 |
+
"loss": 0.6894,
|
| 50 |
+
"rewards/accuracies": 0.71875,
|
| 51 |
+
"rewards/chosen": 0.005721195600926876,
|
| 52 |
+
"rewards/margins": 0.0084818284958601,
|
| 53 |
+
"rewards/rejected": -0.0027606328949332237,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.06322444678609063,
|
| 58 |
+
"grad_norm": 7.846241639500232,
|
| 59 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 60 |
+
"logits/chosen": -2.920705795288086,
|
| 61 |
+
"logits/rejected": -2.904344081878662,
|
| 62 |
+
"logps/chosen": -298.88543701171875,
|
| 63 |
+
"logps/rejected": -309.48614501953125,
|
| 64 |
+
"loss": 0.6698,
|
| 65 |
+
"rewards/accuracies": 0.8125,
|
| 66 |
+
"rewards/chosen": 0.026271456852555275,
|
| 67 |
+
"rewards/margins": 0.04990110918879509,
|
| 68 |
+
"rewards/rejected": -0.023629654198884964,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.08429926238145416,
|
| 73 |
+
"grad_norm": 8.807848819773,
|
| 74 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 75 |
+
"logits/chosen": -2.92798113822937,
|
| 76 |
+
"logits/rejected": -2.9023869037628174,
|
| 77 |
+
"logps/chosen": -287.96044921875,
|
| 78 |
+
"logps/rejected": -320.7830505371094,
|
| 79 |
+
"loss": 0.6116,
|
| 80 |
+
"rewards/accuracies": 0.8218749761581421,
|
| 81 |
+
"rewards/chosen": -0.0031547886319458485,
|
| 82 |
+
"rewards/margins": 0.1678742617368698,
|
| 83 |
+
"rewards/rejected": -0.17102906107902527,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.1053740779768177,
|
| 88 |
+
"grad_norm": 9.9820742658112,
|
| 89 |
+
"learning_rate": 4.999728079043996e-07,
|
| 90 |
+
"logits/chosen": -3.0243992805480957,
|
| 91 |
+
"logits/rejected": -2.9876396656036377,
|
| 92 |
+
"logps/chosen": -386.5646057128906,
|
| 93 |
+
"logps/rejected": -462.30694580078125,
|
| 94 |
+
"loss": 0.4878,
|
| 95 |
+
"rewards/accuracies": 0.828125,
|
| 96 |
+
"rewards/chosen": -0.7237198948860168,
|
| 97 |
+
"rewards/margins": 0.6255429983139038,
|
| 98 |
+
"rewards/rejected": -1.3492629528045654,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.12644889357218125,
|
| 103 |
+
"grad_norm": 13.622612239694844,
|
| 104 |
+
"learning_rate": 4.990217055187362e-07,
|
| 105 |
+
"logits/chosen": -3.105769395828247,
|
| 106 |
+
"logits/rejected": -3.083045482635498,
|
| 107 |
+
"logps/chosen": -568.8998413085938,
|
| 108 |
+
"logps/rejected": -771.6845703125,
|
| 109 |
+
"loss": 0.3721,
|
| 110 |
+
"rewards/accuracies": 0.828125,
|
| 111 |
+
"rewards/chosen": -2.756446361541748,
|
| 112 |
+
"rewards/margins": 1.8887875080108643,
|
| 113 |
+
"rewards/rejected": -4.645233631134033,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.1475237091675448,
|
| 118 |
+
"grad_norm": 13.789853181411452,
|
| 119 |
+
"learning_rate": 4.967169078520476e-07,
|
| 120 |
+
"logits/chosen": -3.179182291030884,
|
| 121 |
+
"logits/rejected": -3.1583375930786133,
|
| 122 |
+
"logps/chosen": -573.2684326171875,
|
| 123 |
+
"logps/rejected": -837.7855224609375,
|
| 124 |
+
"loss": 0.3064,
|
| 125 |
+
"rewards/accuracies": 0.828125,
|
| 126 |
+
"rewards/chosen": -2.7872049808502197,
|
| 127 |
+
"rewards/margins": 2.5140316486358643,
|
| 128 |
+
"rewards/rejected": -5.301236629486084,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.16859852476290832,
|
| 133 |
+
"grad_norm": 22.075616248232635,
|
| 134 |
+
"learning_rate": 4.930709439074527e-07,
|
| 135 |
+
"logits/chosen": -3.175481081008911,
|
| 136 |
+
"logits/rejected": -3.1578330993652344,
|
| 137 |
+
"logps/chosen": -679.1819458007812,
|
| 138 |
+
"logps/rejected": -1066.6025390625,
|
| 139 |
+
"loss": 0.2837,
|
| 140 |
+
"rewards/accuracies": 0.840624988079071,
|
| 141 |
+
"rewards/chosen": -3.896761417388916,
|
| 142 |
+
"rewards/margins": 3.724388599395752,
|
| 143 |
+
"rewards/rejected": -7.62114953994751,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.18967334035827185,
|
| 148 |
+
"grad_norm": 15.685694072094627,
|
| 149 |
+
"learning_rate": 4.881036333395328e-07,
|
| 150 |
+
"logits/chosen": -3.170954942703247,
|
| 151 |
+
"logits/rejected": -3.155890941619873,
|
| 152 |
+
"logps/chosen": -609.4580688476562,
|
| 153 |
+
"logps/rejected": -974.7374267578125,
|
| 154 |
+
"loss": 0.268,
|
| 155 |
+
"rewards/accuracies": 0.840624988079071,
|
| 156 |
+
"rewards/chosen": -3.0708985328674316,
|
| 157 |
+
"rewards/margins": 3.6035964488983154,
|
| 158 |
+
"rewards/rejected": -6.674495697021484,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.2107481559536354,
|
| 163 |
+
"grad_norm": 18.90111640126689,
|
| 164 |
+
"learning_rate": 4.818419787136311e-07,
|
| 165 |
+
"logits/chosen": -3.083106279373169,
|
| 166 |
+
"logits/rejected": -3.054839611053467,
|
| 167 |
+
"logps/chosen": -578.0242919921875,
|
| 168 |
+
"logps/rejected": -1001.0924682617188,
|
| 169 |
+
"loss": 0.2558,
|
| 170 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 171 |
+
"rewards/chosen": -2.942237138748169,
|
| 172 |
+
"rewards/margins": 4.090949058532715,
|
| 173 |
+
"rewards/rejected": -7.033186912536621,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.2107481559536354,
|
| 178 |
+
"eval_logits/chosen": -3.124534845352173,
|
| 179 |
+
"eval_logits/rejected": -3.112900733947754,
|
| 180 |
+
"eval_logps/chosen": -694.1878051757812,
|
| 181 |
+
"eval_logps/rejected": -1132.10791015625,
|
| 182 |
+
"eval_loss": 0.25145235657691956,
|
| 183 |
+
"eval_rewards/accuracies": 0.8362500071525574,
|
| 184 |
+
"eval_rewards/chosen": -3.3401803970336914,
|
| 185 |
+
"eval_rewards/margins": 4.293397426605225,
|
| 186 |
+
"eval_rewards/rejected": -7.633577823638916,
|
| 187 |
+
"eval_runtime": 294.2421,
|
| 188 |
+
"eval_samples_per_second": 21.724,
|
| 189 |
+
"eval_steps_per_second": 0.34,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.23182297154899895,
|
| 194 |
+
"grad_norm": 16.556417816253123,
|
| 195 |
+
"learning_rate": 4.7432001871846694e-07,
|
| 196 |
+
"logits/chosen": -3.0214481353759766,
|
| 197 |
+
"logits/rejected": -2.9939751625061035,
|
| 198 |
+
"logps/chosen": -565.4187622070312,
|
| 199 |
+
"logps/rejected": -990.65478515625,
|
| 200 |
+
"loss": 0.2415,
|
| 201 |
+
"rewards/accuracies": 0.846875011920929,
|
| 202 |
+
"rewards/chosen": -2.8635013103485107,
|
| 203 |
+
"rewards/margins": 4.140561103820801,
|
| 204 |
+
"rewards/rejected": -7.004061698913574,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.2528977871443625,
|
| 209 |
+
"grad_norm": 36.18904544734523,
|
| 210 |
+
"learning_rate": 4.655786431300069e-07,
|
| 211 |
+
"logits/chosen": -3.0037097930908203,
|
| 212 |
+
"logits/rejected": -2.979717493057251,
|
| 213 |
+
"logps/chosen": -620.5673828125,
|
| 214 |
+
"logps/rejected": -1151.0147705078125,
|
| 215 |
+
"loss": 0.2443,
|
| 216 |
+
"rewards/accuracies": 0.862500011920929,
|
| 217 |
+
"rewards/chosen": -3.3321540355682373,
|
| 218 |
+
"rewards/margins": 5.147377967834473,
|
| 219 |
+
"rewards/rejected": -8.479532241821289,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.273972602739726,
|
| 224 |
+
"grad_norm": 18.70901795335142,
|
| 225 |
+
"learning_rate": 4.55665370532461e-07,
|
| 226 |
+
"logits/chosen": -2.97050404548645,
|
| 227 |
+
"logits/rejected": -2.9448835849761963,
|
| 228 |
+
"logps/chosen": -572.4600830078125,
|
| 229 |
+
"logps/rejected": -1059.3128662109375,
|
| 230 |
+
"loss": 0.2432,
|
| 231 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 232 |
+
"rewards/chosen": -2.720818519592285,
|
| 233 |
+
"rewards/margins": 4.68039608001709,
|
| 234 |
+
"rewards/rejected": -7.401214599609375,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.2950474183350896,
|
| 239 |
+
"grad_norm": 18.2112429436409,
|
| 240 |
+
"learning_rate": 4.446340900047223e-07,
|
| 241 |
+
"logits/chosen": -2.942176342010498,
|
| 242 |
+
"logits/rejected": -2.9192707538604736,
|
| 243 |
+
"logps/chosen": -543.0445556640625,
|
| 244 |
+
"logps/rejected": -1057.8526611328125,
|
| 245 |
+
"loss": 0.2351,
|
| 246 |
+
"rewards/accuracies": 0.856249988079071,
|
| 247 |
+
"rewards/chosen": -2.618699789047241,
|
| 248 |
+
"rewards/margins": 4.893606185913086,
|
| 249 |
+
"rewards/rejected": -7.512306213378906,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.31612223393045313,
|
| 254 |
+
"grad_norm": 19.519419751801294,
|
| 255 |
+
"learning_rate": 4.325447681764586e-07,
|
| 256 |
+
"logits/chosen": -2.861685276031494,
|
| 257 |
+
"logits/rejected": -2.836148738861084,
|
| 258 |
+
"logps/chosen": -572.3717041015625,
|
| 259 |
+
"logps/rejected": -1162.0958251953125,
|
| 260 |
+
"loss": 0.2207,
|
| 261 |
+
"rewards/accuracies": 0.875,
|
| 262 |
+
"rewards/chosen": -3.064368486404419,
|
| 263 |
+
"rewards/margins": 5.642943382263184,
|
| 264 |
+
"rewards/rejected": -8.707311630249023,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.33719704952581664,
|
| 269 |
+
"grad_norm": 21.374254957458422,
|
| 270 |
+
"learning_rate": 4.1946312324631276e-07,
|
| 271 |
+
"logits/chosen": -2.8398401737213135,
|
| 272 |
+
"logits/rejected": -2.79673433303833,
|
| 273 |
+
"logps/chosen": -608.8257446289062,
|
| 274 |
+
"logps/rejected": -1237.2921142578125,
|
| 275 |
+
"loss": 0.2219,
|
| 276 |
+
"rewards/accuracies": 0.878125011920929,
|
| 277 |
+
"rewards/chosen": -3.0954906940460205,
|
| 278 |
+
"rewards/margins": 6.1136980056762695,
|
| 279 |
+
"rewards/rejected": -9.209188461303711,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.3582718651211802,
|
| 284 |
+
"grad_norm": 22.91268457823563,
|
| 285 |
+
"learning_rate": 4.0546026773426835e-07,
|
| 286 |
+
"logits/chosen": -2.7913198471069336,
|
| 287 |
+
"logits/rejected": -2.751511335372925,
|
| 288 |
+
"logps/chosen": -636.1942749023438,
|
| 289 |
+
"logps/rejected": -1279.91357421875,
|
| 290 |
+
"loss": 0.2176,
|
| 291 |
+
"rewards/accuracies": 0.856249988079071,
|
| 292 |
+
"rewards/chosen": -3.3635153770446777,
|
| 293 |
+
"rewards/margins": 6.284361839294434,
|
| 294 |
+
"rewards/rejected": -9.647878646850586,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.3793466807165437,
|
| 299 |
+
"grad_norm": 17.21799076325381,
|
| 300 |
+
"learning_rate": 3.9061232191019517e-07,
|
| 301 |
+
"logits/chosen": -2.7712769508361816,
|
| 302 |
+
"logits/rejected": -2.7276077270507812,
|
| 303 |
+
"logps/chosen": -575.2244873046875,
|
| 304 |
+
"logps/rejected": -1119.7806396484375,
|
| 305 |
+
"loss": 0.2257,
|
| 306 |
+
"rewards/accuracies": 0.862500011920929,
|
| 307 |
+
"rewards/chosen": -2.81962251663208,
|
| 308 |
+
"rewards/margins": 5.422026634216309,
|
| 309 |
+
"rewards/rejected": -8.241649627685547,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.40042149631190727,
|
| 314 |
+
"grad_norm": 20.672214976979642,
|
| 315 |
+
"learning_rate": 3.75e-07,
|
| 316 |
+
"logits/chosen": -2.659935712814331,
|
| 317 |
+
"logits/rejected": -2.618821144104004,
|
| 318 |
+
"logps/chosen": -634.2330322265625,
|
| 319 |
+
"logps/rejected": -1279.434326171875,
|
| 320 |
+
"loss": 0.225,
|
| 321 |
+
"rewards/accuracies": 0.871874988079071,
|
| 322 |
+
"rewards/chosen": -3.622291088104248,
|
| 323 |
+
"rewards/margins": 6.227897644042969,
|
| 324 |
+
"rewards/rejected": -9.850188255310059,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.4214963119072708,
|
| 329 |
+
"grad_norm": 21.322162159621154,
|
| 330 |
+
"learning_rate": 3.5870817141878733e-07,
|
| 331 |
+
"logits/chosen": -2.649911880493164,
|
| 332 |
+
"logits/rejected": -2.598778247833252,
|
| 333 |
+
"logps/chosen": -625.1468505859375,
|
| 334 |
+
"logps/rejected": -1208.257080078125,
|
| 335 |
+
"loss": 0.2204,
|
| 336 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 337 |
+
"rewards/chosen": -3.3302555084228516,
|
| 338 |
+
"rewards/margins": 5.740489959716797,
|
| 339 |
+
"rewards/rejected": -9.070745468139648,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.4214963119072708,
|
| 344 |
+
"eval_logits/chosen": -2.7329447269439697,
|
| 345 |
+
"eval_logits/rejected": -2.6947696208953857,
|
| 346 |
+
"eval_logps/chosen": -736.4188842773438,
|
| 347 |
+
"eval_logps/rejected": -1306.89501953125,
|
| 348 |
+
"eval_loss": 0.22599634528160095,
|
| 349 |
+
"eval_rewards/accuracies": 0.8587499856948853,
|
| 350 |
+
"eval_rewards/chosen": -3.7624917030334473,
|
| 351 |
+
"eval_rewards/margins": 5.618957042694092,
|
| 352 |
+
"eval_rewards/rejected": -9.381448745727539,
|
| 353 |
+
"eval_runtime": 297.1596,
|
| 354 |
+
"eval_samples_per_second": 21.51,
|
| 355 |
+
"eval_steps_per_second": 0.337,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.44257112750263433,
|
| 360 |
+
"grad_norm": 13.31035102593267,
|
| 361 |
+
"learning_rate": 3.418253994161892e-07,
|
| 362 |
+
"logits/chosen": -2.6757235527038574,
|
| 363 |
+
"logits/rejected": -2.6246249675750732,
|
| 364 |
+
"logps/chosen": -629.2353515625,
|
| 365 |
+
"logps/rejected": -1236.3388671875,
|
| 366 |
+
"loss": 0.2229,
|
| 367 |
+
"rewards/accuracies": 0.8531249761581421,
|
| 368 |
+
"rewards/chosen": -3.4933292865753174,
|
| 369 |
+
"rewards/margins": 5.829395294189453,
|
| 370 |
+
"rewards/rejected": -9.322725296020508,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.4636459430979979,
|
| 375 |
+
"grad_norm": 16.25315803809995,
|
| 376 |
+
"learning_rate": 3.244434596418139e-07,
|
| 377 |
+
"logits/chosen": -2.7054009437561035,
|
| 378 |
+
"logits/rejected": -2.658482313156128,
|
| 379 |
+
"logps/chosen": -635.1846313476562,
|
| 380 |
+
"logps/rejected": -1274.047119140625,
|
| 381 |
+
"loss": 0.2029,
|
| 382 |
+
"rewards/accuracies": 0.8531249761581421,
|
| 383 |
+
"rewards/chosen": -3.43379282951355,
|
| 384 |
+
"rewards/margins": 6.226017951965332,
|
| 385 |
+
"rewards/rejected": -9.659811019897461,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.48472075869336145,
|
| 390 |
+
"grad_norm": 18.979777410538098,
|
| 391 |
+
"learning_rate": 3.066568412479167e-07,
|
| 392 |
+
"logits/chosen": -2.7195122241973877,
|
| 393 |
+
"logits/rejected": -2.6623740196228027,
|
| 394 |
+
"logps/chosen": -645.3162841796875,
|
| 395 |
+
"logps/rejected": -1333.099365234375,
|
| 396 |
+
"loss": 0.209,
|
| 397 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 398 |
+
"rewards/chosen": -3.6151633262634277,
|
| 399 |
+
"rewards/margins": 6.788934230804443,
|
| 400 |
+
"rewards/rejected": -10.404096603393555,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.505795574288725,
|
| 405 |
+
"grad_norm": 14.572953436795377,
|
| 406 |
+
"learning_rate": 2.8856223324132555e-07,
|
| 407 |
+
"logits/chosen": -2.73162579536438,
|
| 408 |
+
"logits/rejected": -2.68361234664917,
|
| 409 |
+
"logps/chosen": -623.8787231445312,
|
| 410 |
+
"logps/rejected": -1263.7047119140625,
|
| 411 |
+
"loss": 0.2131,
|
| 412 |
+
"rewards/accuracies": 0.8656250238418579,
|
| 413 |
+
"rewards/chosen": -3.288330554962158,
|
| 414 |
+
"rewards/margins": 6.264595985412598,
|
| 415 |
+
"rewards/rejected": -9.552926063537598,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5268703898840885,
|
| 420 |
+
"grad_norm": 15.727626316245573,
|
| 421 |
+
"learning_rate": 2.7025799887683996e-07,
|
| 422 |
+
"logits/chosen": -2.652764081954956,
|
| 423 |
+
"logits/rejected": -2.596436023712158,
|
| 424 |
+
"logps/chosen": -676.56494140625,
|
| 425 |
+
"logps/rejected": -1391.48095703125,
|
| 426 |
+
"loss": 0.2028,
|
| 427 |
+
"rewards/accuracies": 0.8531249761581421,
|
| 428 |
+
"rewards/chosen": -3.7927825450897217,
|
| 429 |
+
"rewards/margins": 7.004532814025879,
|
| 430 |
+
"rewards/rejected": -10.797314643859863,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.547945205479452,
|
| 435 |
+
"grad_norm": 14.772632800998503,
|
| 436 |
+
"learning_rate": 2.518436409493281e-07,
|
| 437 |
+
"logits/chosen": -2.593949794769287,
|
| 438 |
+
"logits/rejected": -2.5094432830810547,
|
| 439 |
+
"logps/chosen": -597.7491455078125,
|
| 440 |
+
"logps/rejected": -1275.842529296875,
|
| 441 |
+
"loss": 0.2136,
|
| 442 |
+
"rewards/accuracies": 0.893750011920929,
|
| 443 |
+
"rewards/chosen": -3.1343584060668945,
|
| 444 |
+
"rewards/margins": 6.584381103515625,
|
| 445 |
+
"rewards/rejected": -9.718740463256836,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.5690200210748156,
|
| 450 |
+
"grad_norm": 14.402069920182464,
|
| 451 |
+
"learning_rate": 2.3341926089122408e-07,
|
| 452 |
+
"logits/chosen": -2.5253357887268066,
|
| 453 |
+
"logits/rejected": -2.403183698654175,
|
| 454 |
+
"logps/chosen": -579.7105712890625,
|
| 455 |
+
"logps/rejected": -1186.1314697265625,
|
| 456 |
+
"loss": 0.2185,
|
| 457 |
+
"rewards/accuracies": 0.871874988079071,
|
| 458 |
+
"rewards/chosen": -2.8341760635375977,
|
| 459 |
+
"rewards/margins": 5.931792259216309,
|
| 460 |
+
"rewards/rejected": -8.76596736907959,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.5900948366701791,
|
| 465 |
+
"grad_norm": 20.389043372443385,
|
| 466 |
+
"learning_rate": 2.1508501461579848e-07,
|
| 467 |
+
"logits/chosen": -2.323107957839966,
|
| 468 |
+
"logits/rejected": -2.1116223335266113,
|
| 469 |
+
"logps/chosen": -674.8060913085938,
|
| 470 |
+
"logps/rejected": -1314.333740234375,
|
| 471 |
+
"loss": 0.1983,
|
| 472 |
+
"rewards/accuracies": 0.862500011920929,
|
| 473 |
+
"rewards/chosen": -3.736111879348755,
|
| 474 |
+
"rewards/margins": 6.262539386749268,
|
| 475 |
+
"rewards/rejected": -9.998651504516602,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.6111696522655427,
|
| 480 |
+
"grad_norm": 11.479399018053,
|
| 481 |
+
"learning_rate": 1.9694056806426927e-07,
|
| 482 |
+
"logits/chosen": -2.2525992393493652,
|
| 483 |
+
"logits/rejected": -1.9792003631591797,
|
| 484 |
+
"logps/chosen": -641.6122436523438,
|
| 485 |
+
"logps/rejected": -1334.318115234375,
|
| 486 |
+
"loss": 0.2017,
|
| 487 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 488 |
+
"rewards/chosen": -3.6301467418670654,
|
| 489 |
+
"rewards/margins": 6.797834873199463,
|
| 490 |
+
"rewards/rejected": -10.427982330322266,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6322444678609063,
|
| 495 |
+
"grad_norm": 18.35346885487197,
|
| 496 |
+
"learning_rate": 1.7908455541642582e-07,
|
| 497 |
+
"logits/chosen": -2.305216073989868,
|
| 498 |
+
"logits/rejected": -1.9890754222869873,
|
| 499 |
+
"logps/chosen": -631.3724365234375,
|
| 500 |
+
"logps/rejected": -1385.84765625,
|
| 501 |
+
"loss": 0.204,
|
| 502 |
+
"rewards/accuracies": 0.8843749761581421,
|
| 503 |
+
"rewards/chosen": -3.2906665802001953,
|
| 504 |
+
"rewards/margins": 7.4196929931640625,
|
| 505 |
+
"rewards/rejected": -10.710360527038574,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.6322444678609063,
|
| 510 |
+
"eval_logits/chosen": -2.410874128341675,
|
| 511 |
+
"eval_logits/rejected": -2.1722898483276367,
|
| 512 |
+
"eval_logps/chosen": -699.75537109375,
|
| 513 |
+
"eval_logps/rejected": -1343.685791015625,
|
| 514 |
+
"eval_loss": 0.2095702439546585,
|
| 515 |
+
"eval_rewards/accuracies": 0.8650000095367432,
|
| 516 |
+
"eval_rewards/chosen": -3.3958559036254883,
|
| 517 |
+
"eval_rewards/margins": 6.353498935699463,
|
| 518 |
+
"eval_rewards/rejected": -9.749356269836426,
|
| 519 |
+
"eval_runtime": 296.8451,
|
| 520 |
+
"eval_samples_per_second": 21.533,
|
| 521 |
+
"eval_steps_per_second": 0.337,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.6533192834562698,
|
| 526 |
+
"grad_norm": 15.446119970622084,
|
| 527 |
+
"learning_rate": 1.616140429099641e-07,
|
| 528 |
+
"logits/chosen": -2.237611770629883,
|
| 529 |
+
"logits/rejected": -1.8744182586669922,
|
| 530 |
+
"logps/chosen": -615.1163330078125,
|
| 531 |
+
"logps/rejected": -1335.3438720703125,
|
| 532 |
+
"loss": 0.1899,
|
| 533 |
+
"rewards/accuracies": 0.871874988079071,
|
| 534 |
+
"rewards/chosen": -3.29571533203125,
|
| 535 |
+
"rewards/margins": 7.18838357925415,
|
| 536 |
+
"rewards/rejected": -10.484098434448242,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.6743940990516333,
|
| 541 |
+
"grad_norm": 14.390167758723463,
|
| 542 |
+
"learning_rate": 1.4462400118323796e-07,
|
| 543 |
+
"logits/chosen": -2.0499463081359863,
|
| 544 |
+
"logits/rejected": -1.637145757675171,
|
| 545 |
+
"logps/chosen": -694.0391235351562,
|
| 546 |
+
"logps/rejected": -1580.717041015625,
|
| 547 |
+
"loss": 0.1836,
|
| 548 |
+
"rewards/accuracies": 0.8968750238418579,
|
| 549 |
+
"rewards/chosen": -4.01133394241333,
|
| 550 |
+
"rewards/margins": 8.607083320617676,
|
| 551 |
+
"rewards/rejected": -12.618416786193848,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.6954689146469969,
|
| 556 |
+
"grad_norm": 24.67534044273584,
|
| 557 |
+
"learning_rate": 1.2820678900980092e-07,
|
| 558 |
+
"logits/chosen": -2.1196751594543457,
|
| 559 |
+
"logits/rejected": -1.6566823720932007,
|
| 560 |
+
"logps/chosen": -678.884033203125,
|
| 561 |
+
"logps/rejected": -1541.774169921875,
|
| 562 |
+
"loss": 0.215,
|
| 563 |
+
"rewards/accuracies": 0.8656250238418579,
|
| 564 |
+
"rewards/chosen": -3.847275495529175,
|
| 565 |
+
"rewards/margins": 8.554319381713867,
|
| 566 |
+
"rewards/rejected": -12.401594161987305,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.7165437302423604,
|
| 571 |
+
"grad_norm": 12.333676271606883,
|
| 572 |
+
"learning_rate": 1.1245165123118359e-07,
|
| 573 |
+
"logits/chosen": -2.0177483558654785,
|
| 574 |
+
"logits/rejected": -1.5283257961273193,
|
| 575 |
+
"logps/chosen": -682.48486328125,
|
| 576 |
+
"logps/rejected": -1445.4771728515625,
|
| 577 |
+
"loss": 0.1979,
|
| 578 |
+
"rewards/accuracies": 0.878125011920929,
|
| 579 |
+
"rewards/chosen": -3.8393378257751465,
|
| 580 |
+
"rewards/margins": 7.487583160400391,
|
| 581 |
+
"rewards/rejected": -11.326921463012695,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.7376185458377239,
|
| 586 |
+
"grad_norm": 19.10139426185179,
|
| 587 |
+
"learning_rate": 9.744423361717321e-08,
|
| 588 |
+
"logits/chosen": -1.8916494846343994,
|
| 589 |
+
"logits/rejected": -1.4087257385253906,
|
| 590 |
+
"logps/chosen": -687.3905639648438,
|
| 591 |
+
"logps/rejected": -1476.8948974609375,
|
| 592 |
+
"loss": 0.1986,
|
| 593 |
+
"rewards/accuracies": 0.887499988079071,
|
| 594 |
+
"rewards/chosen": -3.84537935256958,
|
| 595 |
+
"rewards/margins": 7.789330959320068,
|
| 596 |
+
"rewards/rejected": -11.634710311889648,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.7586933614330874,
|
| 601 |
+
"grad_norm": 18.80792334093226,
|
| 602 |
+
"learning_rate": 8.32661172908373e-08,
|
| 603 |
+
"logits/chosen": -1.90652334690094,
|
| 604 |
+
"logits/rejected": -1.4300750494003296,
|
| 605 |
+
"logps/chosen": -688.5945434570312,
|
| 606 |
+
"logps/rejected": -1452.357177734375,
|
| 607 |
+
"loss": 0.195,
|
| 608 |
+
"rewards/accuracies": 0.8812500238418579,
|
| 609 |
+
"rewards/chosen": -4.015127182006836,
|
| 610 |
+
"rewards/margins": 7.490170955657959,
|
| 611 |
+
"rewards/rejected": -11.505297660827637,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.779768177028451,
|
| 616 |
+
"grad_norm": 16.997599298870874,
|
| 617 |
+
"learning_rate": 6.999437524918569e-08,
|
| 618 |
+
"logits/chosen": -1.9973640441894531,
|
| 619 |
+
"logits/rejected": -1.5239393711090088,
|
| 620 |
+
"logps/chosen": -659.0184936523438,
|
| 621 |
+
"logps/rejected": -1305.682861328125,
|
| 622 |
+
"loss": 0.1912,
|
| 623 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 624 |
+
"rewards/chosen": -3.621332883834839,
|
| 625 |
+
"rewards/margins": 6.495469570159912,
|
| 626 |
+
"rewards/rejected": -10.116803169250488,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.8008429926238145,
|
| 631 |
+
"grad_norm": 14.084078614585282,
|
| 632 |
+
"learning_rate": 5.770115339024484e-08,
|
| 633 |
+
"logits/chosen": -1.9838542938232422,
|
| 634 |
+
"logits/rejected": -1.5452864170074463,
|
| 635 |
+
"logps/chosen": -661.1948852539062,
|
| 636 |
+
"logps/rejected": -1432.19140625,
|
| 637 |
+
"loss": 0.1824,
|
| 638 |
+
"rewards/accuracies": 0.893750011920929,
|
| 639 |
+
"rewards/chosen": -3.7382915019989014,
|
| 640 |
+
"rewards/margins": 7.500711917877197,
|
| 641 |
+
"rewards/rejected": -11.239004135131836,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.821917808219178,
|
| 646 |
+
"grad_norm": 30.61074285987269,
|
| 647 |
+
"learning_rate": 4.645327832410648e-08,
|
| 648 |
+
"logits/chosen": -1.9398887157440186,
|
| 649 |
+
"logits/rejected": -1.4298299551010132,
|
| 650 |
+
"logps/chosen": -683.1571655273438,
|
| 651 |
+
"logps/rejected": -1461.3238525390625,
|
| 652 |
+
"loss": 0.1854,
|
| 653 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 654 |
+
"rewards/chosen": -4.023918628692627,
|
| 655 |
+
"rewards/margins": 7.673205375671387,
|
| 656 |
+
"rewards/rejected": -11.697123527526855,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.8429926238145417,
|
| 661 |
+
"grad_norm": 22.843164623474657,
|
| 662 |
+
"learning_rate": 3.6311894099908145e-08,
|
| 663 |
+
"logits/chosen": -1.9181022644042969,
|
| 664 |
+
"logits/rejected": -1.380041480064392,
|
| 665 |
+
"logps/chosen": -674.0886840820312,
|
| 666 |
+
"logps/rejected": -1500.5560302734375,
|
| 667 |
+
"loss": 0.1992,
|
| 668 |
+
"rewards/accuracies": 0.862500011920929,
|
| 669 |
+
"rewards/chosen": -3.9246249198913574,
|
| 670 |
+
"rewards/margins": 8.195623397827148,
|
| 671 |
+
"rewards/rejected": -12.120248794555664,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.8429926238145417,
|
| 676 |
+
"eval_logits/chosen": -2.0689847469329834,
|
| 677 |
+
"eval_logits/rejected": -1.6459288597106934,
|
| 678 |
+
"eval_logps/chosen": -781.930419921875,
|
| 679 |
+
"eval_logps/rejected": -1544.589111328125,
|
| 680 |
+
"eval_loss": 0.20072707533836365,
|
| 681 |
+
"eval_rewards/accuracies": 0.8650000095367432,
|
| 682 |
+
"eval_rewards/chosen": -4.217607021331787,
|
| 683 |
+
"eval_rewards/margins": 7.540782928466797,
|
| 684 |
+
"eval_rewards/rejected": -11.75838851928711,
|
| 685 |
+
"eval_runtime": 293.2964,
|
| 686 |
+
"eval_samples_per_second": 21.794,
|
| 687 |
+
"eval_steps_per_second": 0.341,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.8640674394099052,
|
| 692 |
+
"grad_norm": 19.85885717669904,
|
| 693 |
+
"learning_rate": 2.7332129823519568e-08,
|
| 694 |
+
"logits/chosen": -1.881338357925415,
|
| 695 |
+
"logits/rejected": -1.340071439743042,
|
| 696 |
+
"logps/chosen": -689.0841064453125,
|
| 697 |
+
"logps/rejected": -1438.508544921875,
|
| 698 |
+
"loss": 0.1926,
|
| 699 |
+
"rewards/accuracies": 0.862500011920929,
|
| 700 |
+
"rewards/chosen": -3.9749252796173096,
|
| 701 |
+
"rewards/margins": 7.4057297706604,
|
| 702 |
+
"rewards/rejected": -11.380653381347656,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.8851422550052687,
|
| 707 |
+
"grad_norm": 17.487805159196274,
|
| 708 |
+
"learning_rate": 1.956279997278043e-08,
|
| 709 |
+
"logits/chosen": -1.9210704565048218,
|
| 710 |
+
"logits/rejected": -1.4240577220916748,
|
| 711 |
+
"logps/chosen": -656.7258911132812,
|
| 712 |
+
"logps/rejected": -1409.5032958984375,
|
| 713 |
+
"loss": 0.1887,
|
| 714 |
+
"rewards/accuracies": 0.909375011920929,
|
| 715 |
+
"rewards/chosen": -3.5839180946350098,
|
| 716 |
+
"rewards/margins": 7.417031288146973,
|
| 717 |
+
"rewards/rejected": -11.00094985961914,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.9062170706006323,
|
| 722 |
+
"grad_norm": 20.395649107934336,
|
| 723 |
+
"learning_rate": 1.3046139039394e-08,
|
| 724 |
+
"logits/chosen": -1.998051643371582,
|
| 725 |
+
"logits/rejected": -1.4957393407821655,
|
| 726 |
+
"logps/chosen": -653.7247314453125,
|
| 727 |
+
"logps/rejected": -1424.0172119140625,
|
| 728 |
+
"loss": 0.1918,
|
| 729 |
+
"rewards/accuracies": 0.840624988079071,
|
| 730 |
+
"rewards/chosen": -3.6267688274383545,
|
| 731 |
+
"rewards/margins": 7.642977714538574,
|
| 732 |
+
"rewards/rejected": -11.269746780395508,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9272918861959958,
|
| 737 |
+
"grad_norm": 29.88901784574838,
|
| 738 |
+
"learning_rate": 7.817571939976286e-09,
|
| 739 |
+
"logits/chosen": -1.935821294784546,
|
| 740 |
+
"logits/rejected": -1.4820038080215454,
|
| 741 |
+
"logps/chosen": -660.3615112304688,
|
| 742 |
+
"logps/rejected": -1442.939697265625,
|
| 743 |
+
"loss": 0.1977,
|
| 744 |
+
"rewards/accuracies": 0.890625,
|
| 745 |
+
"rewards/chosen": -3.6211764812469482,
|
| 746 |
+
"rewards/margins": 7.658024787902832,
|
| 747 |
+
"rewards/rejected": -11.279200553894043,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.9483667017913593,
|
| 752 |
+
"grad_norm": 24.434459876362702,
|
| 753 |
+
"learning_rate": 3.905521444318604e-09,
|
| 754 |
+
"logits/chosen": -1.974572777748108,
|
| 755 |
+
"logits/rejected": -1.43704092502594,
|
| 756 |
+
"logps/chosen": -653.2598266601562,
|
| 757 |
+
"logps/rejected": -1464.2042236328125,
|
| 758 |
+
"loss": 0.1991,
|
| 759 |
+
"rewards/accuracies": 0.846875011920929,
|
| 760 |
+
"rewards/chosen": -3.468592405319214,
|
| 761 |
+
"rewards/margins": 8.024352073669434,
|
| 762 |
+
"rewards/rejected": -11.492944717407227,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.9694415173867229,
|
| 767 |
+
"grad_norm": 15.522136682704172,
|
| 768 |
+
"learning_rate": 1.3312536676942377e-09,
|
| 769 |
+
"logits/chosen": -1.9699881076812744,
|
| 770 |
+
"logits/rejected": -1.501794457435608,
|
| 771 |
+
"logps/chosen": -654.7131958007812,
|
| 772 |
+
"logps/rejected": -1468.2657470703125,
|
| 773 |
+
"loss": 0.1974,
|
| 774 |
+
"rewards/accuracies": 0.878125011920929,
|
| 775 |
+
"rewards/chosen": -3.683525800704956,
|
| 776 |
+
"rewards/margins": 7.84524393081665,
|
| 777 |
+
"rewards/rejected": -11.528770446777344,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.9905163329820864,
|
| 782 |
+
"grad_norm": 13.129882834290527,
|
| 783 |
+
"learning_rate": 1.0876246712074322e-10,
|
| 784 |
+
"logits/chosen": -2.0095367431640625,
|
| 785 |
+
"logits/rejected": -1.5517585277557373,
|
| 786 |
+
"logps/chosen": -640.3642578125,
|
| 787 |
+
"logps/rejected": -1447.8231201171875,
|
| 788 |
+
"loss": 0.1857,
|
| 789 |
+
"rewards/accuracies": 0.893750011920929,
|
| 790 |
+
"rewards/chosen": -3.563875675201416,
|
| 791 |
+
"rewards/margins": 7.8468523025512695,
|
| 792 |
+
"rewards/rejected": -11.410726547241211,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.9989462592202318,
|
| 797 |
+
"step": 474,
|
| 798 |
+
"total_flos": 0.0,
|
| 799 |
+
"train_loss": 0.26123517437323235,
|
| 800 |
+
"train_runtime": 15193.8056,
|
| 801 |
+
"train_samples_per_second": 7.992,
|
| 802 |
+
"train_steps_per_second": 0.031
|
| 803 |
+
}
|
| 804 |
+
],
|
| 805 |
+
"logging_steps": 10,
|
| 806 |
+
"max_steps": 474,
|
| 807 |
+
"num_input_tokens_seen": 0,
|
| 808 |
+
"num_train_epochs": 1,
|
| 809 |
+
"save_steps": 100,
|
| 810 |
+
"stateful_callbacks": {
|
| 811 |
+
"TrainerControl": {
|
| 812 |
+
"args": {
|
| 813 |
+
"should_epoch_stop": false,
|
| 814 |
+
"should_evaluate": false,
|
| 815 |
+
"should_log": false,
|
| 816 |
+
"should_save": true,
|
| 817 |
+
"should_training_stop": true
|
| 818 |
+
},
|
| 819 |
+
"attributes": {}
|
| 820 |
+
}
|
| 821 |
+
},
|
| 822 |
+
"total_flos": 0.0,
|
| 823 |
+
"train_batch_size": 8,
|
| 824 |
+
"trial_name": null,
|
| 825 |
+
"trial_params": null
|
| 826 |
+
}
|