Model save
Browse files- README.md +80 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +815 -0
README.md
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
model-index:
|
| 9 |
+
- name: zephyr-7b-dpo-full-prometheus-reward-scale-1-rpo
|
| 10 |
+
results: []
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 14 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 15 |
+
|
| 16 |
+
# zephyr-7b-dpo-full-prometheus-reward-scale-1-rpo
|
| 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: 0.0240
|
| 21 |
+
- Rewards/chosen: -0.1305
|
| 22 |
+
- Rewards/rejected: -0.3804
|
| 23 |
+
- Rewards/accuracies: 0.7328
|
| 24 |
+
- Rewards/margins: 0.2499
|
| 25 |
+
- Logps/rejected: -286.3204
|
| 26 |
+
- Logps/chosen: -273.0143
|
| 27 |
+
- Logits/rejected: -2.4857
|
| 28 |
+
- Logits/chosen: -2.5477
|
| 29 |
+
|
| 30 |
+
## Model description
|
| 31 |
+
|
| 32 |
+
More information needed
|
| 33 |
+
|
| 34 |
+
## Intended uses & limitations
|
| 35 |
+
|
| 36 |
+
More information needed
|
| 37 |
+
|
| 38 |
+
## Training and evaluation data
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Training procedure
|
| 43 |
+
|
| 44 |
+
### Training hyperparameters
|
| 45 |
+
|
| 46 |
+
The following hyperparameters were used during training:
|
| 47 |
+
- learning_rate: 5e-07
|
| 48 |
+
- train_batch_size: 8
|
| 49 |
+
- eval_batch_size: 8
|
| 50 |
+
- seed: 55
|
| 51 |
+
- distributed_type: multi-GPU
|
| 52 |
+
- num_devices: 8
|
| 53 |
+
- gradient_accumulation_steps: 2
|
| 54 |
+
- total_train_batch_size: 128
|
| 55 |
+
- total_eval_batch_size: 64
|
| 56 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 57 |
+
- lr_scheduler_type: cosine
|
| 58 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 59 |
+
- num_epochs: 1
|
| 60 |
+
|
| 61 |
+
### Training results
|
| 62 |
+
|
| 63 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 64 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 65 |
+
| 0.0371 | 0.1143 | 50 | 0.0330 | 0.0162 | -0.0844 | 0.6638 | 0.1006 | -256.7177 | -258.3393 | -2.5728 | -2.6169 |
|
| 66 |
+
| 0.0326 | 0.2286 | 100 | 0.0284 | -0.1604 | -0.3762 | 0.7069 | 0.2158 | -285.9020 | -276.0061 | -2.2908 | -2.3469 |
|
| 67 |
+
| 0.0276 | 0.3429 | 150 | 0.0261 | -0.1426 | -0.3607 | 0.7198 | 0.2181 | -284.3463 | -274.2200 | -2.3538 | -2.3992 |
|
| 68 |
+
| 0.0257 | 0.4571 | 200 | 0.0255 | -0.1250 | -0.3457 | 0.7371 | 0.2206 | -282.8442 | -272.4646 | -2.4669 | -2.5054 |
|
| 69 |
+
| 0.0259 | 0.5714 | 250 | 0.0249 | -0.1421 | -0.3761 | 0.7457 | 0.2340 | -285.8867 | -274.1743 | -2.5651 | -2.6093 |
|
| 70 |
+
| 0.0244 | 0.6857 | 300 | 0.0244 | -0.1115 | -0.3601 | 0.7328 | 0.2486 | -284.2852 | -271.1066 | -2.4931 | -2.5556 |
|
| 71 |
+
| 0.0232 | 0.8 | 350 | 0.0241 | -0.1203 | -0.3689 | 0.7328 | 0.2486 | -285.1674 | -271.9947 | -2.4940 | -2.5542 |
|
| 72 |
+
| 0.0253 | 0.9143 | 400 | 0.0240 | -0.1305 | -0.3804 | 0.7328 | 0.2499 | -286.3204 | -273.0143 | -2.4857 | -2.5477 |
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
### Framework versions
|
| 76 |
+
|
| 77 |
+
- Transformers 4.44.0.dev0
|
| 78 |
+
- Pytorch 2.1.2
|
| 79 |
+
- Datasets 2.20.0
|
| 80 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9988571428571429,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.02869435827710263,
|
| 5 |
+
"train_runtime": 11449.0759,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.891,
|
| 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": 0.9988571428571429,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.02869435827710263,
|
| 5 |
+
"train_runtime": 11449.0759,
|
| 6 |
+
"train_samples": 55999,
|
| 7 |
+
"train_samples_per_second": 4.891,
|
| 8 |
+
"train_steps_per_second": 0.038
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,815 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9988571428571429,
|
| 5 |
+
"eval_steps": 50,
|
| 6 |
+
"global_step": 437,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.022857142857142857,
|
| 13 |
+
"grad_norm": 2.9285875213659387,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.6540167331695557,
|
| 16 |
+
"logits/rejected": -2.6346607208251953,
|
| 17 |
+
"logps/chosen": -287.88873291015625,
|
| 18 |
+
"logps/rejected": -266.501953125,
|
| 19 |
+
"loss": 0.0456,
|
| 20 |
+
"rewards/accuracies": 0.4000000059604645,
|
| 21 |
+
"rewards/chosen": 0.0002630561648402363,
|
| 22 |
+
"rewards/margins": 0.000395975133869797,
|
| 23 |
+
"rewards/rejected": -0.00013291896902956069,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045714285714285714,
|
| 28 |
+
"grad_norm": 2.6703796663983237,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.6553053855895996,
|
| 31 |
+
"logits/rejected": -2.62392258644104,
|
| 32 |
+
"logps/chosen": -271.43328857421875,
|
| 33 |
+
"logps/rejected": -260.1735534667969,
|
| 34 |
+
"loss": 0.0465,
|
| 35 |
+
"rewards/accuracies": 0.550000011920929,
|
| 36 |
+
"rewards/chosen": 0.0027677875477820635,
|
| 37 |
+
"rewards/margins": 0.001332368585281074,
|
| 38 |
+
"rewards/rejected": 0.0014354187296703458,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06857142857142857,
|
| 43 |
+
"grad_norm": 2.4626318533205516,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.65043568611145,
|
| 46 |
+
"logits/rejected": -2.5593643188476562,
|
| 47 |
+
"logps/chosen": -271.85040283203125,
|
| 48 |
+
"logps/rejected": -254.72055053710938,
|
| 49 |
+
"loss": 0.0437,
|
| 50 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 51 |
+
"rewards/chosen": 0.020383944734930992,
|
| 52 |
+
"rewards/margins": 0.012471409514546394,
|
| 53 |
+
"rewards/rejected": 0.007912535220384598,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09142857142857143,
|
| 58 |
+
"grad_norm": 2.1219802043829175,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.6147427558898926,
|
| 61 |
+
"logits/rejected": -2.6001462936401367,
|
| 62 |
+
"logps/chosen": -261.89312744140625,
|
| 63 |
+
"logps/rejected": -255.43844604492188,
|
| 64 |
+
"loss": 0.0409,
|
| 65 |
+
"rewards/accuracies": 0.699999988079071,
|
| 66 |
+
"rewards/chosen": 0.03881374001502991,
|
| 67 |
+
"rewards/margins": 0.03865280747413635,
|
| 68 |
+
"rewards/rejected": 0.0001609304890735075,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11428571428571428,
|
| 73 |
+
"grad_norm": 2.648446772286267,
|
| 74 |
+
"learning_rate": 4.997124959943201e-07,
|
| 75 |
+
"logits/chosen": -2.5977413654327393,
|
| 76 |
+
"logits/rejected": -2.530109405517578,
|
| 77 |
+
"logps/chosen": -287.03973388671875,
|
| 78 |
+
"logps/rejected": -261.2123718261719,
|
| 79 |
+
"loss": 0.0371,
|
| 80 |
+
"rewards/accuracies": 0.768750011920929,
|
| 81 |
+
"rewards/chosen": 0.024105368182063103,
|
| 82 |
+
"rewards/margins": 0.09380187839269638,
|
| 83 |
+
"rewards/rejected": -0.06969650834798813,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11428571428571428,
|
| 88 |
+
"eval_logits/chosen": -2.6169474124908447,
|
| 89 |
+
"eval_logits/rejected": -2.5727975368499756,
|
| 90 |
+
"eval_logps/chosen": -258.3393249511719,
|
| 91 |
+
"eval_logps/rejected": -256.71771240234375,
|
| 92 |
+
"eval_loss": 0.032985787838697433,
|
| 93 |
+
"eval_rewards/accuracies": 0.6637930870056152,
|
| 94 |
+
"eval_rewards/chosen": 0.01621999777853489,
|
| 95 |
+
"eval_rewards/margins": 0.10062520951032639,
|
| 96 |
+
"eval_rewards/rejected": -0.08440519869327545,
|
| 97 |
+
"eval_runtime": 90.7209,
|
| 98 |
+
"eval_samples_per_second": 20.183,
|
| 99 |
+
"eval_steps_per_second": 0.32,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13714285714285715,
|
| 104 |
+
"grad_norm": 4.351801812595937,
|
| 105 |
+
"learning_rate": 4.979579212164186e-07,
|
| 106 |
+
"logits/chosen": -2.5900657176971436,
|
| 107 |
+
"logits/rejected": -2.5341601371765137,
|
| 108 |
+
"logps/chosen": -294.68328857421875,
|
| 109 |
+
"logps/rejected": -255.0167999267578,
|
| 110 |
+
"loss": 0.0338,
|
| 111 |
+
"rewards/accuracies": 0.71875,
|
| 112 |
+
"rewards/chosen": 0.048027873039245605,
|
| 113 |
+
"rewards/margins": 0.13361422717571259,
|
| 114 |
+
"rewards/rejected": -0.08558637648820877,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16,
|
| 119 |
+
"grad_norm": 3.0187175488356703,
|
| 120 |
+
"learning_rate": 4.946196886175515e-07,
|
| 121 |
+
"logits/chosen": -2.539175510406494,
|
| 122 |
+
"logits/rejected": -2.4328207969665527,
|
| 123 |
+
"logps/chosen": -299.51702880859375,
|
| 124 |
+
"logps/rejected": -263.10400390625,
|
| 125 |
+
"loss": 0.0324,
|
| 126 |
+
"rewards/accuracies": 0.706250011920929,
|
| 127 |
+
"rewards/chosen": 0.03838000446557999,
|
| 128 |
+
"rewards/margins": 0.1820719689130783,
|
| 129 |
+
"rewards/rejected": -0.14369197189807892,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.18285714285714286,
|
| 134 |
+
"grad_norm": 2.829278868602528,
|
| 135 |
+
"learning_rate": 4.897191188239667e-07,
|
| 136 |
+
"logits/chosen": -2.503023147583008,
|
| 137 |
+
"logits/rejected": -2.4128475189208984,
|
| 138 |
+
"logps/chosen": -271.768310546875,
|
| 139 |
+
"logps/rejected": -266.82537841796875,
|
| 140 |
+
"loss": 0.0321,
|
| 141 |
+
"rewards/accuracies": 0.6875,
|
| 142 |
+
"rewards/chosen": -0.04572376608848572,
|
| 143 |
+
"rewards/margins": 0.15518848598003387,
|
| 144 |
+
"rewards/rejected": -0.2009122371673584,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.2057142857142857,
|
| 149 |
+
"grad_norm": 2.659426058179347,
|
| 150 |
+
"learning_rate": 4.832875107981763e-07,
|
| 151 |
+
"logits/chosen": -2.122408390045166,
|
| 152 |
+
"logits/rejected": -2.066514730453491,
|
| 153 |
+
"logps/chosen": -249.37106323242188,
|
| 154 |
+
"logps/rejected": -279.49212646484375,
|
| 155 |
+
"loss": 0.0311,
|
| 156 |
+
"rewards/accuracies": 0.762499988079071,
|
| 157 |
+
"rewards/chosen": -0.07777222990989685,
|
| 158 |
+
"rewards/margins": 0.199189230799675,
|
| 159 |
+
"rewards/rejected": -0.27696144580841064,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22857142857142856,
|
| 164 |
+
"grad_norm": 3.4560236437659873,
|
| 165 |
+
"learning_rate": 4.753659419387223e-07,
|
| 166 |
+
"logits/chosen": -2.338256359100342,
|
| 167 |
+
"logits/rejected": -2.29487943649292,
|
| 168 |
+
"logps/chosen": -266.89556884765625,
|
| 169 |
+
"logps/rejected": -254.134521484375,
|
| 170 |
+
"loss": 0.0326,
|
| 171 |
+
"rewards/accuracies": 0.75,
|
| 172 |
+
"rewards/chosen": -0.056456904858350754,
|
| 173 |
+
"rewards/margins": 0.17859290540218353,
|
| 174 |
+
"rewards/rejected": -0.23504984378814697,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22857142857142856,
|
| 179 |
+
"eval_logits/chosen": -2.3468899726867676,
|
| 180 |
+
"eval_logits/rejected": -2.2908225059509277,
|
| 181 |
+
"eval_logps/chosen": -276.0060729980469,
|
| 182 |
+
"eval_logps/rejected": -285.9020080566406,
|
| 183 |
+
"eval_loss": 0.02837423048913479,
|
| 184 |
+
"eval_rewards/accuracies": 0.7068965435028076,
|
| 185 |
+
"eval_rewards/chosen": -0.16044749319553375,
|
| 186 |
+
"eval_rewards/margins": 0.21580053865909576,
|
| 187 |
+
"eval_rewards/rejected": -0.3762480318546295,
|
| 188 |
+
"eval_runtime": 90.6197,
|
| 189 |
+
"eval_samples_per_second": 20.205,
|
| 190 |
+
"eval_steps_per_second": 0.32,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25142857142857145,
|
| 195 |
+
"grad_norm": 3.986741503860454,
|
| 196 |
+
"learning_rate": 4.660050057270191e-07,
|
| 197 |
+
"logits/chosen": -2.414931058883667,
|
| 198 |
+
"logits/rejected": -2.3160946369171143,
|
| 199 |
+
"logps/chosen": -314.08258056640625,
|
| 200 |
+
"logps/rejected": -264.74102783203125,
|
| 201 |
+
"loss": 0.0274,
|
| 202 |
+
"rewards/accuracies": 0.768750011920929,
|
| 203 |
+
"rewards/chosen": -0.07821311056613922,
|
| 204 |
+
"rewards/margins": 0.22488529980182648,
|
| 205 |
+
"rewards/rejected": -0.3030984103679657,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.2742857142857143,
|
| 210 |
+
"grad_norm": 3.0310867108282418,
|
| 211 |
+
"learning_rate": 4.5526448859687144e-07,
|
| 212 |
+
"logits/chosen": -2.4880690574645996,
|
| 213 |
+
"logits/rejected": -2.4331300258636475,
|
| 214 |
+
"logps/chosen": -281.57952880859375,
|
| 215 |
+
"logps/rejected": -279.35137939453125,
|
| 216 |
+
"loss": 0.0292,
|
| 217 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 218 |
+
"rewards/chosen": -0.08748000860214233,
|
| 219 |
+
"rewards/margins": 0.20896463096141815,
|
| 220 |
+
"rewards/rejected": -0.2964446544647217,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.29714285714285715,
|
| 225 |
+
"grad_norm": 2.686160441272964,
|
| 226 |
+
"learning_rate": 4.432129880904388e-07,
|
| 227 |
+
"logits/chosen": -2.505074977874756,
|
| 228 |
+
"logits/rejected": -2.4108517169952393,
|
| 229 |
+
"logps/chosen": -281.3280944824219,
|
| 230 |
+
"logps/rejected": -264.3760986328125,
|
| 231 |
+
"loss": 0.0271,
|
| 232 |
+
"rewards/accuracies": 0.625,
|
| 233 |
+
"rewards/chosen": -0.07846375554800034,
|
| 234 |
+
"rewards/margins": 0.17637231945991516,
|
| 235 |
+
"rewards/rejected": -0.2548360228538513,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.32,
|
| 240 |
+
"grad_norm": 2.7176410561935582,
|
| 241 |
+
"learning_rate": 4.299274747394055e-07,
|
| 242 |
+
"logits/chosen": -2.41599702835083,
|
| 243 |
+
"logits/rejected": -2.3766045570373535,
|
| 244 |
+
"logps/chosen": -328.3199157714844,
|
| 245 |
+
"logps/rejected": -292.5992736816406,
|
| 246 |
+
"loss": 0.0269,
|
| 247 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 248 |
+
"rewards/chosen": -0.10486980527639389,
|
| 249 |
+
"rewards/margins": 0.21833749115467072,
|
| 250 |
+
"rewards/rejected": -0.323207288980484,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.34285714285714286,
|
| 255 |
+
"grad_norm": 2.624843592752447,
|
| 256 |
+
"learning_rate": 4.1549280046953653e-07,
|
| 257 |
+
"logits/chosen": -2.360866069793701,
|
| 258 |
+
"logits/rejected": -2.3315412998199463,
|
| 259 |
+
"logps/chosen": -309.66278076171875,
|
| 260 |
+
"logps/rejected": -299.77325439453125,
|
| 261 |
+
"loss": 0.0276,
|
| 262 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 263 |
+
"rewards/chosen": -0.13608722388744354,
|
| 264 |
+
"rewards/margins": 0.20585337281227112,
|
| 265 |
+
"rewards/rejected": -0.34194058179855347,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.34285714285714286,
|
| 270 |
+
"eval_logits/chosen": -2.3992269039154053,
|
| 271 |
+
"eval_logits/rejected": -2.353783369064331,
|
| 272 |
+
"eval_logps/chosen": -274.219970703125,
|
| 273 |
+
"eval_logps/rejected": -284.3463439941406,
|
| 274 |
+
"eval_loss": 0.026143357157707214,
|
| 275 |
+
"eval_rewards/accuracies": 0.7198275923728943,
|
| 276 |
+
"eval_rewards/chosen": -0.1425865888595581,
|
| 277 |
+
"eval_rewards/margins": 0.21810482442378998,
|
| 278 |
+
"eval_rewards/rejected": -0.3606913983821869,
|
| 279 |
+
"eval_runtime": 91.9073,
|
| 280 |
+
"eval_samples_per_second": 19.922,
|
| 281 |
+
"eval_steps_per_second": 0.316,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3657142857142857,
|
| 286 |
+
"grad_norm": 5.477307060879475,
|
| 287 |
+
"learning_rate": 4.000011566683401e-07,
|
| 288 |
+
"logits/chosen": -2.4539685249328613,
|
| 289 |
+
"logits/rejected": -2.354241371154785,
|
| 290 |
+
"logps/chosen": -321.659423828125,
|
| 291 |
+
"logps/rejected": -301.2185974121094,
|
| 292 |
+
"loss": 0.0272,
|
| 293 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 294 |
+
"rewards/chosen": -0.12505923211574554,
|
| 295 |
+
"rewards/margins": 0.18981832265853882,
|
| 296 |
+
"rewards/rejected": -0.31487756967544556,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38857142857142857,
|
| 301 |
+
"grad_norm": 3.291959162082193,
|
| 302 |
+
"learning_rate": 3.8355148537705047e-07,
|
| 303 |
+
"logits/chosen": -2.4768872261047363,
|
| 304 |
+
"logits/rejected": -2.4600863456726074,
|
| 305 |
+
"logps/chosen": -300.3105773925781,
|
| 306 |
+
"logps/rejected": -303.3681335449219,
|
| 307 |
+
"loss": 0.0265,
|
| 308 |
+
"rewards/accuracies": 0.78125,
|
| 309 |
+
"rewards/chosen": -0.06613761186599731,
|
| 310 |
+
"rewards/margins": 0.22872081398963928,
|
| 311 |
+
"rewards/rejected": -0.2948583960533142,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.4114285714285714,
|
| 316 |
+
"grad_norm": 3.4863952801217724,
|
| 317 |
+
"learning_rate": 3.662488473675315e-07,
|
| 318 |
+
"logits/chosen": -2.496547222137451,
|
| 319 |
+
"logits/rejected": -2.414828062057495,
|
| 320 |
+
"logps/chosen": -312.77105712890625,
|
| 321 |
+
"logps/rejected": -283.5912170410156,
|
| 322 |
+
"loss": 0.0277,
|
| 323 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 324 |
+
"rewards/chosen": -0.1033080667257309,
|
| 325 |
+
"rewards/margins": 0.268028199672699,
|
| 326 |
+
"rewards/rejected": -0.37133628129959106,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.4342857142857143,
|
| 331 |
+
"grad_norm": 2.6342821858302345,
|
| 332 |
+
"learning_rate": 3.48203751140067e-07,
|
| 333 |
+
"logits/chosen": -2.5662384033203125,
|
| 334 |
+
"logits/rejected": -2.477693796157837,
|
| 335 |
+
"logps/chosen": -274.94903564453125,
|
| 336 |
+
"logps/rejected": -275.15032958984375,
|
| 337 |
+
"loss": 0.0267,
|
| 338 |
+
"rewards/accuracies": 0.800000011920929,
|
| 339 |
+
"rewards/chosen": -0.14113394916057587,
|
| 340 |
+
"rewards/margins": 0.26104897260665894,
|
| 341 |
+
"rewards/rejected": -0.4021829068660736,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45714285714285713,
|
| 346 |
+
"grad_norm": 2.36382999827445,
|
| 347 |
+
"learning_rate": 3.2953144712759537e-07,
|
| 348 |
+
"logits/chosen": -2.5955493450164795,
|
| 349 |
+
"logits/rejected": -2.541722059249878,
|
| 350 |
+
"logps/chosen": -258.66326904296875,
|
| 351 |
+
"logps/rejected": -270.30609130859375,
|
| 352 |
+
"loss": 0.0257,
|
| 353 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 354 |
+
"rewards/chosen": -0.08326815813779831,
|
| 355 |
+
"rewards/margins": 0.23202641308307648,
|
| 356 |
+
"rewards/rejected": -0.3152945637702942,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45714285714285713,
|
| 361 |
+
"eval_logits/chosen": -2.505389451980591,
|
| 362 |
+
"eval_logits/rejected": -2.466869831085205,
|
| 363 |
+
"eval_logps/chosen": -272.464599609375,
|
| 364 |
+
"eval_logps/rejected": -282.8442077636719,
|
| 365 |
+
"eval_loss": 0.025467805564403534,
|
| 366 |
+
"eval_rewards/accuracies": 0.7370689511299133,
|
| 367 |
+
"eval_rewards/chosen": -0.12503290176391602,
|
| 368 |
+
"eval_rewards/margins": 0.22063730657100677,
|
| 369 |
+
"eval_rewards/rejected": -0.3456701934337616,
|
| 370 |
+
"eval_runtime": 92.5527,
|
| 371 |
+
"eval_samples_per_second": 19.783,
|
| 372 |
+
"eval_steps_per_second": 0.313,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.48,
|
| 377 |
+
"grad_norm": 2.537677375344279,
|
| 378 |
+
"learning_rate": 3.103511916141658e-07,
|
| 379 |
+
"logits/chosen": -2.515646457672119,
|
| 380 |
+
"logits/rejected": -2.4341044425964355,
|
| 381 |
+
"logps/chosen": -296.09417724609375,
|
| 382 |
+
"logps/rejected": -296.98992919921875,
|
| 383 |
+
"loss": 0.0255,
|
| 384 |
+
"rewards/accuracies": 0.762499988079071,
|
| 385 |
+
"rewards/chosen": -0.10250511020421982,
|
| 386 |
+
"rewards/margins": 0.24679327011108398,
|
| 387 |
+
"rewards/rejected": -0.3492983877658844,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5028571428571429,
|
| 392 |
+
"grad_norm": 3.017540832091563,
|
| 393 |
+
"learning_rate": 2.9078548506882117e-07,
|
| 394 |
+
"logits/chosen": -2.455840826034546,
|
| 395 |
+
"logits/rejected": -2.4036200046539307,
|
| 396 |
+
"logps/chosen": -314.2608642578125,
|
| 397 |
+
"logps/rejected": -324.8536071777344,
|
| 398 |
+
"loss": 0.0268,
|
| 399 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 400 |
+
"rewards/chosen": -0.09878429025411606,
|
| 401 |
+
"rewards/margins": 0.24836881458759308,
|
| 402 |
+
"rewards/rejected": -0.3471531271934509,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5257142857142857,
|
| 407 |
+
"grad_norm": 2.5806806439798318,
|
| 408 |
+
"learning_rate": 2.709592897595191e-07,
|
| 409 |
+
"logits/chosen": -2.5932087898254395,
|
| 410 |
+
"logits/rejected": -2.5489110946655273,
|
| 411 |
+
"logps/chosen": -291.34063720703125,
|
| 412 |
+
"logps/rejected": -303.5892028808594,
|
| 413 |
+
"loss": 0.027,
|
| 414 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 415 |
+
"rewards/chosen": -0.06546103954315186,
|
| 416 |
+
"rewards/margins": 0.18855178356170654,
|
| 417 |
+
"rewards/rejected": -0.2540128231048584,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5485714285714286,
|
| 422 |
+
"grad_norm": 2.6133927348580595,
|
| 423 |
+
"learning_rate": 2.509992316440332e-07,
|
| 424 |
+
"logits/chosen": -2.605285167694092,
|
| 425 |
+
"logits/rejected": -2.5287115573883057,
|
| 426 |
+
"logps/chosen": -316.57171630859375,
|
| 427 |
+
"logps/rejected": -295.0510559082031,
|
| 428 |
+
"loss": 0.0264,
|
| 429 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 430 |
+
"rewards/chosen": -0.08883465826511383,
|
| 431 |
+
"rewards/margins": 0.24537017941474915,
|
| 432 |
+
"rewards/rejected": -0.3342048227787018,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.5714285714285714,
|
| 437 |
+
"grad_norm": 2.534725769830134,
|
| 438 |
+
"learning_rate": 2.3103279163519918e-07,
|
| 439 |
+
"logits/chosen": -2.5661091804504395,
|
| 440 |
+
"logits/rejected": -2.5358757972717285,
|
| 441 |
+
"logps/chosen": -294.138427734375,
|
| 442 |
+
"logps/rejected": -314.68145751953125,
|
| 443 |
+
"loss": 0.0259,
|
| 444 |
+
"rewards/accuracies": 0.675000011920929,
|
| 445 |
+
"rewards/chosen": -0.16875380277633667,
|
| 446 |
+
"rewards/margins": 0.16810715198516846,
|
| 447 |
+
"rewards/rejected": -0.3368609547615051,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.5714285714285714,
|
| 452 |
+
"eval_logits/chosen": -2.609321117401123,
|
| 453 |
+
"eval_logits/rejected": -2.5650839805603027,
|
| 454 |
+
"eval_logps/chosen": -274.17425537109375,
|
| 455 |
+
"eval_logps/rejected": -285.8867492675781,
|
| 456 |
+
"eval_loss": 0.02494961954653263,
|
| 457 |
+
"eval_rewards/accuracies": 0.7456896305084229,
|
| 458 |
+
"eval_rewards/chosen": -0.14212976396083832,
|
| 459 |
+
"eval_rewards/margins": 0.23396606743335724,
|
| 460 |
+
"eval_rewards/rejected": -0.37609583139419556,
|
| 461 |
+
"eval_runtime": 92.1346,
|
| 462 |
+
"eval_samples_per_second": 19.873,
|
| 463 |
+
"eval_steps_per_second": 0.315,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5942857142857143,
|
| 468 |
+
"grad_norm": 2.414707825956405,
|
| 469 |
+
"learning_rate": 2.1118749140573358e-07,
|
| 470 |
+
"logits/chosen": -2.580400228500366,
|
| 471 |
+
"logits/rejected": -2.5031521320343018,
|
| 472 |
+
"logps/chosen": -283.9717712402344,
|
| 473 |
+
"logps/rejected": -252.01821899414062,
|
| 474 |
+
"loss": 0.0261,
|
| 475 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 476 |
+
"rewards/chosen": -0.0814717561006546,
|
| 477 |
+
"rewards/margins": 0.21402212977409363,
|
| 478 |
+
"rewards/rejected": -0.2954939305782318,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6171428571428571,
|
| 483 |
+
"grad_norm": 2.5939402682824984,
|
| 484 |
+
"learning_rate": 1.9159007893272703e-07,
|
| 485 |
+
"logits/chosen": -2.6335721015930176,
|
| 486 |
+
"logits/rejected": -2.547212600708008,
|
| 487 |
+
"logps/chosen": -297.3409423828125,
|
| 488 |
+
"logps/rejected": -272.5791931152344,
|
| 489 |
+
"loss": 0.026,
|
| 490 |
+
"rewards/accuracies": 0.637499988079071,
|
| 491 |
+
"rewards/chosen": -0.09214051067829132,
|
| 492 |
+
"rewards/margins": 0.1732146292924881,
|
| 493 |
+
"rewards/rejected": -0.26535511016845703,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.64,
|
| 498 |
+
"grad_norm": 2.2469749248074895,
|
| 499 |
+
"learning_rate": 1.7236571898357766e-07,
|
| 500 |
+
"logits/chosen": -2.4656567573547363,
|
| 501 |
+
"logits/rejected": -2.41697359085083,
|
| 502 |
+
"logps/chosen": -263.90411376953125,
|
| 503 |
+
"logps/rejected": -272.6389465332031,
|
| 504 |
+
"loss": 0.0259,
|
| 505 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 506 |
+
"rewards/chosen": -0.1417647898197174,
|
| 507 |
+
"rewards/margins": 0.24923905730247498,
|
| 508 |
+
"rewards/rejected": -0.3910038471221924,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6628571428571428,
|
| 513 |
+
"grad_norm": 2.6187338744329365,
|
| 514 |
+
"learning_rate": 1.5363719371356882e-07,
|
| 515 |
+
"logits/chosen": -2.569354295730591,
|
| 516 |
+
"logits/rejected": -2.459669589996338,
|
| 517 |
+
"logps/chosen": -297.64971923828125,
|
| 518 |
+
"logps/rejected": -284.9115295410156,
|
| 519 |
+
"loss": 0.0263,
|
| 520 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 521 |
+
"rewards/chosen": -0.11722731590270996,
|
| 522 |
+
"rewards/margins": 0.20035290718078613,
|
| 523 |
+
"rewards/rejected": -0.3175801634788513,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6857142857142857,
|
| 528 |
+
"grad_norm": 2.6936960874326923,
|
| 529 |
+
"learning_rate": 1.3552411848071565e-07,
|
| 530 |
+
"logits/chosen": -2.6174628734588623,
|
| 531 |
+
"logits/rejected": -2.498176097869873,
|
| 532 |
+
"logps/chosen": -322.8263854980469,
|
| 533 |
+
"logps/rejected": -298.7721252441406,
|
| 534 |
+
"loss": 0.0244,
|
| 535 |
+
"rewards/accuracies": 0.762499988079071,
|
| 536 |
+
"rewards/chosen": -0.09476250410079956,
|
| 537 |
+
"rewards/margins": 0.24882373213768005,
|
| 538 |
+
"rewards/rejected": -0.343586266040802,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6857142857142857,
|
| 543 |
+
"eval_logits/chosen": -2.5555710792541504,
|
| 544 |
+
"eval_logits/rejected": -2.4930553436279297,
|
| 545 |
+
"eval_logps/chosen": -271.10662841796875,
|
| 546 |
+
"eval_logps/rejected": -284.2852478027344,
|
| 547 |
+
"eval_loss": 0.0244239903986454,
|
| 548 |
+
"eval_rewards/accuracies": 0.732758641242981,
|
| 549 |
+
"eval_rewards/chosen": -0.11145297437906265,
|
| 550 |
+
"eval_rewards/margins": 0.24862772226333618,
|
| 551 |
+
"eval_rewards/rejected": -0.3600807189941406,
|
| 552 |
+
"eval_runtime": 91.0911,
|
| 553 |
+
"eval_samples_per_second": 20.101,
|
| 554 |
+
"eval_steps_per_second": 0.318,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7085714285714285,
|
| 559 |
+
"grad_norm": 2.3645242243925644,
|
| 560 |
+
"learning_rate": 1.1814217788631473e-07,
|
| 561 |
+
"logits/chosen": -2.63641095161438,
|
| 562 |
+
"logits/rejected": -2.527548313140869,
|
| 563 |
+
"logps/chosen": -320.08880615234375,
|
| 564 |
+
"logps/rejected": -300.36761474609375,
|
| 565 |
+
"loss": 0.0242,
|
| 566 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 567 |
+
"rewards/chosen": -0.0683048814535141,
|
| 568 |
+
"rewards/margins": 0.2592974901199341,
|
| 569 |
+
"rewards/rejected": -0.3276023864746094,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7314285714285714,
|
| 574 |
+
"grad_norm": 2.7644839761717734,
|
| 575 |
+
"learning_rate": 1.0160238692045331e-07,
|
| 576 |
+
"logits/chosen": -2.511667251586914,
|
| 577 |
+
"logits/rejected": -2.400014638900757,
|
| 578 |
+
"logps/chosen": -307.7832336425781,
|
| 579 |
+
"logps/rejected": -310.0841064453125,
|
| 580 |
+
"loss": 0.0244,
|
| 581 |
+
"rewards/accuracies": 0.793749988079071,
|
| 582 |
+
"rewards/chosen": -0.10665328800678253,
|
| 583 |
+
"rewards/margins": 0.24441497027873993,
|
| 584 |
+
"rewards/rejected": -0.35106825828552246,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7542857142857143,
|
| 589 |
+
"grad_norm": 2.781520358885941,
|
| 590 |
+
"learning_rate": 8.601038193139438e-08,
|
| 591 |
+
"logits/chosen": -2.4999566078186035,
|
| 592 |
+
"logits/rejected": -2.454909563064575,
|
| 593 |
+
"logps/chosen": -257.72882080078125,
|
| 594 |
+
"logps/rejected": -273.76507568359375,
|
| 595 |
+
"loss": 0.0274,
|
| 596 |
+
"rewards/accuracies": 0.706250011920929,
|
| 597 |
+
"rewards/chosen": -0.12040361016988754,
|
| 598 |
+
"rewards/margins": 0.21321117877960205,
|
| 599 |
+
"rewards/rejected": -0.3336147665977478,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7771428571428571,
|
| 604 |
+
"grad_norm": 2.7244436343518386,
|
| 605 |
+
"learning_rate": 7.146574594727572e-08,
|
| 606 |
+
"logits/chosen": -2.5485446453094482,
|
| 607 |
+
"logits/rejected": -2.4302964210510254,
|
| 608 |
+
"logps/chosen": -257.6943664550781,
|
| 609 |
+
"logps/rejected": -247.9900665283203,
|
| 610 |
+
"loss": 0.0256,
|
| 611 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 612 |
+
"rewards/chosen": -0.1169048398733139,
|
| 613 |
+
"rewards/margins": 0.22713470458984375,
|
| 614 |
+
"rewards/rejected": -0.34403952956199646,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8,
|
| 619 |
+
"grad_norm": 2.9548985236957557,
|
| 620 |
+
"learning_rate": 5.8061372659157306e-08,
|
| 621 |
+
"logits/chosen": -2.5826621055603027,
|
| 622 |
+
"logits/rejected": -2.4442431926727295,
|
| 623 |
+
"logps/chosen": -319.64111328125,
|
| 624 |
+
"logps/rejected": -302.02886962890625,
|
| 625 |
+
"loss": 0.0232,
|
| 626 |
+
"rewards/accuracies": 0.762499988079071,
|
| 627 |
+
"rewards/chosen": -0.0946800708770752,
|
| 628 |
+
"rewards/margins": 0.24179792404174805,
|
| 629 |
+
"rewards/rejected": -0.33647799491882324,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8,
|
| 634 |
+
"eval_logits/chosen": -2.5542304515838623,
|
| 635 |
+
"eval_logits/rejected": -2.493983268737793,
|
| 636 |
+
"eval_logps/chosen": -271.9947204589844,
|
| 637 |
+
"eval_logps/rejected": -285.1673889160156,
|
| 638 |
+
"eval_loss": 0.024052616208791733,
|
| 639 |
+
"eval_rewards/accuracies": 0.732758641242981,
|
| 640 |
+
"eval_rewards/chosen": -0.12033439427614212,
|
| 641 |
+
"eval_rewards/margins": 0.24856774508953094,
|
| 642 |
+
"eval_rewards/rejected": -0.36890214681625366,
|
| 643 |
+
"eval_runtime": 92.5532,
|
| 644 |
+
"eval_samples_per_second": 19.783,
|
| 645 |
+
"eval_steps_per_second": 0.313,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8228571428571428,
|
| 650 |
+
"grad_norm": 2.6937215781427297,
|
| 651 |
+
"learning_rate": 4.5882873127531614e-08,
|
| 652 |
+
"logits/chosen": -2.5525920391082764,
|
| 653 |
+
"logits/rejected": -2.4930379390716553,
|
| 654 |
+
"logps/chosen": -303.6790771484375,
|
| 655 |
+
"logps/rejected": -306.0064392089844,
|
| 656 |
+
"loss": 0.0264,
|
| 657 |
+
"rewards/accuracies": 0.75,
|
| 658 |
+
"rewards/chosen": -0.11168299615383148,
|
| 659 |
+
"rewards/margins": 0.2747558355331421,
|
| 660 |
+
"rewards/rejected": -0.38643878698349,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8457142857142858,
|
| 665 |
+
"grad_norm": 3.0478094770904853,
|
| 666 |
+
"learning_rate": 3.500802900154412e-08,
|
| 667 |
+
"logits/chosen": -2.560382604598999,
|
| 668 |
+
"logits/rejected": -2.522139549255371,
|
| 669 |
+
"logps/chosen": -266.638671875,
|
| 670 |
+
"logps/rejected": -294.6788330078125,
|
| 671 |
+
"loss": 0.0246,
|
| 672 |
+
"rewards/accuracies": 0.75,
|
| 673 |
+
"rewards/chosen": -0.10632523149251938,
|
| 674 |
+
"rewards/margins": 0.21934013068675995,
|
| 675 |
+
"rewards/rejected": -0.3256654143333435,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8685714285714285,
|
| 680 |
+
"grad_norm": 2.925024120524796,
|
| 681 |
+
"learning_rate": 2.550629574310309e-08,
|
| 682 |
+
"logits/chosen": -2.484462261199951,
|
| 683 |
+
"logits/rejected": -2.4379611015319824,
|
| 684 |
+
"logps/chosen": -269.59234619140625,
|
| 685 |
+
"logps/rejected": -288.28961181640625,
|
| 686 |
+
"loss": 0.0253,
|
| 687 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 688 |
+
"rewards/chosen": -0.13730758428573608,
|
| 689 |
+
"rewards/margins": 0.16804398596286774,
|
| 690 |
+
"rewards/rejected": -0.305351585149765,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8914285714285715,
|
| 695 |
+
"grad_norm": 2.515719678530588,
|
| 696 |
+
"learning_rate": 1.7438359028687983e-08,
|
| 697 |
+
"logits/chosen": -2.5585129261016846,
|
| 698 |
+
"logits/rejected": -2.4909610748291016,
|
| 699 |
+
"logps/chosen": -277.55438232421875,
|
| 700 |
+
"logps/rejected": -285.38214111328125,
|
| 701 |
+
"loss": 0.0251,
|
| 702 |
+
"rewards/accuracies": 0.668749988079071,
|
| 703 |
+
"rewards/chosen": -0.10983601957559586,
|
| 704 |
+
"rewards/margins": 0.23024682700634003,
|
| 705 |
+
"rewards/rejected": -0.3400828242301941,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9142857142857143,
|
| 710 |
+
"grad_norm": 2.819162826588092,
|
| 711 |
+
"learning_rate": 1.0855747162029361e-08,
|
| 712 |
+
"logits/chosen": -2.538931131362915,
|
| 713 |
+
"logits/rejected": -2.500331401824951,
|
| 714 |
+
"logps/chosen": -285.94342041015625,
|
| 715 |
+
"logps/rejected": -301.77130126953125,
|
| 716 |
+
"loss": 0.0253,
|
| 717 |
+
"rewards/accuracies": 0.737500011920929,
|
| 718 |
+
"rewards/chosen": -0.11810322850942612,
|
| 719 |
+
"rewards/margins": 0.22658781707286835,
|
| 720 |
+
"rewards/rejected": -0.34469106793403625,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9142857142857143,
|
| 725 |
+
"eval_logits/chosen": -2.5477468967437744,
|
| 726 |
+
"eval_logits/rejected": -2.485684633255005,
|
| 727 |
+
"eval_logps/chosen": -273.0142517089844,
|
| 728 |
+
"eval_logps/rejected": -286.3204040527344,
|
| 729 |
+
"eval_loss": 0.024040229618549347,
|
| 730 |
+
"eval_rewards/accuracies": 0.732758641242981,
|
| 731 |
+
"eval_rewards/chosen": -0.13052918016910553,
|
| 732 |
+
"eval_rewards/margins": 0.2499030977487564,
|
| 733 |
+
"eval_rewards/rejected": -0.38043221831321716,
|
| 734 |
+
"eval_runtime": 91.0229,
|
| 735 |
+
"eval_samples_per_second": 20.116,
|
| 736 |
+
"eval_steps_per_second": 0.319,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9371428571428572,
|
| 741 |
+
"grad_norm": 3.0051422552626748,
|
| 742 |
+
"learning_rate": 5.8005019731033615e-09,
|
| 743 |
+
"logits/chosen": -2.547819137573242,
|
| 744 |
+
"logits/rejected": -2.4846596717834473,
|
| 745 |
+
"logps/chosen": -271.83538818359375,
|
| 746 |
+
"logps/rejected": -289.2705078125,
|
| 747 |
+
"loss": 0.0246,
|
| 748 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 749 |
+
"rewards/chosen": -0.11531362682580948,
|
| 750 |
+
"rewards/margins": 0.20074260234832764,
|
| 751 |
+
"rewards/rejected": -0.3160562515258789,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.96,
|
| 756 |
+
"grad_norm": 2.650270081198499,
|
| 757 |
+
"learning_rate": 2.3049103053431886e-09,
|
| 758 |
+
"logits/chosen": -2.6320762634277344,
|
| 759 |
+
"logits/rejected": -2.5170094966888428,
|
| 760 |
+
"logps/chosen": -321.63592529296875,
|
| 761 |
+
"logps/rejected": -333.3614501953125,
|
| 762 |
+
"loss": 0.0246,
|
| 763 |
+
"rewards/accuracies": 0.768750011920929,
|
| 764 |
+
"rewards/chosen": -0.1096896082162857,
|
| 765 |
+
"rewards/margins": 0.2813798785209656,
|
| 766 |
+
"rewards/rejected": -0.3910694718360901,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9828571428571429,
|
| 771 |
+
"grad_norm": 2.8479881145572183,
|
| 772 |
+
"learning_rate": 3.9129780600541397e-10,
|
| 773 |
+
"logits/chosen": -2.5273232460021973,
|
| 774 |
+
"logits/rejected": -2.4938735961914062,
|
| 775 |
+
"logps/chosen": -295.13446044921875,
|
| 776 |
+
"logps/rejected": -292.31854248046875,
|
| 777 |
+
"loss": 0.025,
|
| 778 |
+
"rewards/accuracies": 0.768750011920929,
|
| 779 |
+
"rewards/chosen": -0.09540516138076782,
|
| 780 |
+
"rewards/margins": 0.24863004684448242,
|
| 781 |
+
"rewards/rejected": -0.34403520822525024,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 0.9988571428571429,
|
| 786 |
+
"step": 437,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.02869435827710263,
|
| 789 |
+
"train_runtime": 11449.0759,
|
| 790 |
+
"train_samples_per_second": 4.891,
|
| 791 |
+
"train_steps_per_second": 0.038
|
| 792 |
+
}
|
| 793 |
+
],
|
| 794 |
+
"logging_steps": 10,
|
| 795 |
+
"max_steps": 437,
|
| 796 |
+
"num_input_tokens_seen": 0,
|
| 797 |
+
"num_train_epochs": 1,
|
| 798 |
+
"save_steps": 100,
|
| 799 |
+
"stateful_callbacks": {
|
| 800 |
+
"TrainerControl": {
|
| 801 |
+
"args": {
|
| 802 |
+
"should_epoch_stop": false,
|
| 803 |
+
"should_evaluate": false,
|
| 804 |
+
"should_log": false,
|
| 805 |
+
"should_save": true,
|
| 806 |
+
"should_training_stop": true
|
| 807 |
+
},
|
| 808 |
+
"attributes": {}
|
| 809 |
+
}
|
| 810 |
+
},
|
| 811 |
+
"total_flos": 0.0,
|
| 812 |
+
"train_batch_size": 8,
|
| 813 |
+
"trial_name": null,
|
| 814 |
+
"trial_params": null
|
| 815 |
+
}
|