Model save
Browse files- README.md +76 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- model-00001-of-00003.safetensors +1 -1
- model-00002-of-00003.safetensors +1 -1
- model-00003-of-00003.safetensors +1 -1
- train_results.json +9 -0
- trainer_state.json +751 -0
README.md
ADDED
|
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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-gpt_consistent-3
|
| 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-gpt_consistent-3
|
| 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: 0.4898
|
| 21 |
+
- Rewards/chosen: -0.9477
|
| 22 |
+
- Rewards/rejected: -1.9210
|
| 23 |
+
- Rewards/accuracies: 0.7543
|
| 24 |
+
- Rewards/margins: 0.9733
|
| 25 |
+
- Logps/rejected: -438.6234
|
| 26 |
+
- Logps/chosen: -379.8649
|
| 27 |
+
- Logits/rejected: 2.2425
|
| 28 |
+
- Logits/chosen: 0.9259
|
| 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 |
+
| 0.5718 | 0.2294 | 100 | 0.5525 | -0.3811 | -0.9418 | 0.6940 | 0.5607 | -340.7015 | -323.2031 | -1.1398 | -1.3666 |
|
| 66 |
+
| 0.5281 | 0.4587 | 200 | 0.5102 | -0.6195 | -1.4237 | 0.7457 | 0.8042 | -388.8913 | -347.0378 | 1.0061 | -0.1205 |
|
| 67 |
+
| 0.5213 | 0.6881 | 300 | 0.4939 | -0.9485 | -1.8771 | 0.7543 | 0.9286 | -434.2354 | -379.9385 | 1.7281 | 0.3985 |
|
| 68 |
+
| 0.5037 | 0.9174 | 400 | 0.4898 | -0.9477 | -1.9210 | 0.7543 | 0.9733 | -438.6234 | -379.8649 | 2.2425 | 0.9259 |
|
| 69 |
+
|
| 70 |
+
|
| 71 |
+
### Framework versions
|
| 72 |
+
|
| 73 |
+
- Transformers 4.44.0.dev0
|
| 74 |
+
- Pytorch 2.1.2
|
| 75 |
+
- Datasets 2.20.0
|
| 76 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5438270043889317,
|
| 5 |
+
"train_runtime": 11297.0501,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.936,
|
| 8 |
+
"train_steps_per_second": 0.039
|
| 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 |
+
}
|
model-00001-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4943162336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a3a726c6a3c49fec7a295242da7a31d337a4de81756246752cb13b6ea92e7846
|
| 3 |
size 4943162336
|
model-00002-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4999819336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:01894c62d6b50ea98366120002d4151a8e56e69c8c314ddbd48b5d1e5b6a2b87
|
| 3 |
size 4999819336
|
model-00003-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4540516344
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:287b2cb3d2a8b4f688ee110e0518c085d515038d8c8c6abfbde313f3c2389381
|
| 3 |
size 4540516344
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5438270043889317,
|
| 5 |
+
"train_runtime": 11297.0501,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.936,
|
| 8 |
+
"train_steps_per_second": 0.039
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,751 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 436,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.022935779816513763,
|
| 13 |
+
"grad_norm": 9.13641688324014,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.6193928718566895,
|
| 16 |
+
"logits/rejected": -2.552712917327881,
|
| 17 |
+
"logps/chosen": -265.43743896484375,
|
| 18 |
+
"logps/rejected": -236.1606903076172,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.4312500059604645,
|
| 21 |
+
"rewards/chosen": 8.363189408555627e-05,
|
| 22 |
+
"rewards/margins": 0.0003909034130629152,
|
| 23 |
+
"rewards/rejected": -0.00030727163539268076,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045871559633027525,
|
| 28 |
+
"grad_norm": 8.694428459805497,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.6578612327575684,
|
| 31 |
+
"logits/rejected": -2.575941324234009,
|
| 32 |
+
"logps/chosen": -298.83441162109375,
|
| 33 |
+
"logps/rejected": -274.2864685058594,
|
| 34 |
+
"loss": 0.6921,
|
| 35 |
+
"rewards/accuracies": 0.612500011920929,
|
| 36 |
+
"rewards/chosen": -0.0004915857571177185,
|
| 37 |
+
"rewards/margins": 0.0012736506760120392,
|
| 38 |
+
"rewards/rejected": -0.0017652364913374186,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06880733944954129,
|
| 43 |
+
"grad_norm": 8.216933133006783,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.675276279449463,
|
| 46 |
+
"logits/rejected": -2.601776599884033,
|
| 47 |
+
"logps/chosen": -290.555908203125,
|
| 48 |
+
"logps/rejected": -234.41598510742188,
|
| 49 |
+
"loss": 0.6878,
|
| 50 |
+
"rewards/accuracies": 0.706250011920929,
|
| 51 |
+
"rewards/chosen": 0.007768664509057999,
|
| 52 |
+
"rewards/margins": 0.0121098468080163,
|
| 53 |
+
"rewards/rejected": -0.0043411822989583015,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09174311926605505,
|
| 58 |
+
"grad_norm": 8.711915321379939,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.660461902618408,
|
| 61 |
+
"logits/rejected": -2.6112020015716553,
|
| 62 |
+
"logps/chosen": -281.26617431640625,
|
| 63 |
+
"logps/rejected": -267.74810791015625,
|
| 64 |
+
"loss": 0.6757,
|
| 65 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 66 |
+
"rewards/chosen": 0.039030514657497406,
|
| 67 |
+
"rewards/margins": 0.03858140856027603,
|
| 68 |
+
"rewards/rejected": 0.0004491090658120811,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11467889908256881,
|
| 73 |
+
"grad_norm": 9.886855722455959,
|
| 74 |
+
"learning_rate": 4.997110275491701e-07,
|
| 75 |
+
"logits/chosen": -2.6204681396484375,
|
| 76 |
+
"logits/rejected": -2.6130080223083496,
|
| 77 |
+
"logps/chosen": -294.60052490234375,
|
| 78 |
+
"logps/rejected": -304.6225280761719,
|
| 79 |
+
"loss": 0.6583,
|
| 80 |
+
"rewards/accuracies": 0.706250011920929,
|
| 81 |
+
"rewards/chosen": 0.016506755724549294,
|
| 82 |
+
"rewards/margins": 0.07089035958051682,
|
| 83 |
+
"rewards/rejected": -0.05438361316919327,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.13761467889908258,
|
| 88 |
+
"grad_norm": 11.187614568198285,
|
| 89 |
+
"learning_rate": 4.979475034558115e-07,
|
| 90 |
+
"logits/chosen": -2.575127124786377,
|
| 91 |
+
"logits/rejected": -2.5189614295959473,
|
| 92 |
+
"logps/chosen": -294.8846130371094,
|
| 93 |
+
"logps/rejected": -274.777587890625,
|
| 94 |
+
"loss": 0.6358,
|
| 95 |
+
"rewards/accuracies": 0.675000011920929,
|
| 96 |
+
"rewards/chosen": -0.05399893596768379,
|
| 97 |
+
"rewards/margins": 0.11821160465478897,
|
| 98 |
+
"rewards/rejected": -0.17221052944660187,
|
| 99 |
+
"step": 60
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.16055045871559634,
|
| 103 |
+
"grad_norm": 12.12486219819406,
|
| 104 |
+
"learning_rate": 4.945923025551788e-07,
|
| 105 |
+
"logits/chosen": -2.4791793823242188,
|
| 106 |
+
"logits/rejected": -2.43456768989563,
|
| 107 |
+
"logps/chosen": -341.71612548828125,
|
| 108 |
+
"logps/rejected": -300.43731689453125,
|
| 109 |
+
"loss": 0.6121,
|
| 110 |
+
"rewards/accuracies": 0.699999988079071,
|
| 111 |
+
"rewards/chosen": -0.1414516121149063,
|
| 112 |
+
"rewards/margins": 0.24884521961212158,
|
| 113 |
+
"rewards/rejected": -0.3902968466281891,
|
| 114 |
+
"step": 70
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.1834862385321101,
|
| 118 |
+
"grad_norm": 12.125843765946055,
|
| 119 |
+
"learning_rate": 4.896669632591651e-07,
|
| 120 |
+
"logits/chosen": -2.398482084274292,
|
| 121 |
+
"logits/rejected": -2.2879607677459717,
|
| 122 |
+
"logps/chosen": -305.4925537109375,
|
| 123 |
+
"logps/rejected": -298.49224853515625,
|
| 124 |
+
"loss": 0.6049,
|
| 125 |
+
"rewards/accuracies": 0.71875,
|
| 126 |
+
"rewards/chosen": -0.1964271366596222,
|
| 127 |
+
"rewards/margins": 0.2619345188140869,
|
| 128 |
+
"rewards/rejected": -0.4583616256713867,
|
| 129 |
+
"step": 80
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.20642201834862386,
|
| 133 |
+
"grad_norm": 14.181798116852052,
|
| 134 |
+
"learning_rate": 4.832031033425662e-07,
|
| 135 |
+
"logits/chosen": -2.314242124557495,
|
| 136 |
+
"logits/rejected": -2.2462363243103027,
|
| 137 |
+
"logps/chosen": -307.9869079589844,
|
| 138 |
+
"logps/rejected": -303.9811706542969,
|
| 139 |
+
"loss": 0.5963,
|
| 140 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 141 |
+
"rewards/chosen": -0.20393487811088562,
|
| 142 |
+
"rewards/margins": 0.3678644895553589,
|
| 143 |
+
"rewards/rejected": -0.5717993974685669,
|
| 144 |
+
"step": 90
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.22935779816513763,
|
| 148 |
+
"grad_norm": 21.310278276106065,
|
| 149 |
+
"learning_rate": 4.752422169756047e-07,
|
| 150 |
+
"logits/chosen": -1.5569207668304443,
|
| 151 |
+
"logits/rejected": -1.490770936012268,
|
| 152 |
+
"logps/chosen": -306.3553161621094,
|
| 153 |
+
"logps/rejected": -343.92852783203125,
|
| 154 |
+
"loss": 0.5718,
|
| 155 |
+
"rewards/accuracies": 0.675000011920929,
|
| 156 |
+
"rewards/chosen": -0.3393058478832245,
|
| 157 |
+
"rewards/margins": 0.3806948661804199,
|
| 158 |
+
"rewards/rejected": -0.7200007438659668,
|
| 159 |
+
"step": 100
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.22935779816513763,
|
| 163 |
+
"eval_logits/chosen": -1.366591215133667,
|
| 164 |
+
"eval_logits/rejected": -1.1398286819458008,
|
| 165 |
+
"eval_logps/chosen": -323.2030944824219,
|
| 166 |
+
"eval_logps/rejected": -340.7015075683594,
|
| 167 |
+
"eval_loss": 0.5524524450302124,
|
| 168 |
+
"eval_rewards/accuracies": 0.693965494632721,
|
| 169 |
+
"eval_rewards/chosen": -0.3811298906803131,
|
| 170 |
+
"eval_rewards/margins": 0.5606611967086792,
|
| 171 |
+
"eval_rewards/rejected": -0.9417910575866699,
|
| 172 |
+
"eval_runtime": 95.9209,
|
| 173 |
+
"eval_samples_per_second": 18.953,
|
| 174 |
+
"eval_steps_per_second": 0.302,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.25229357798165136,
|
| 179 |
+
"grad_norm": 22.6242102385037,
|
| 180 |
+
"learning_rate": 4.658354083558188e-07,
|
| 181 |
+
"logits/chosen": -1.3675466775894165,
|
| 182 |
+
"logits/rejected": -1.107060432434082,
|
| 183 |
+
"logps/chosen": -325.51593017578125,
|
| 184 |
+
"logps/rejected": -339.6572265625,
|
| 185 |
+
"loss": 0.5522,
|
| 186 |
+
"rewards/accuracies": 0.706250011920929,
|
| 187 |
+
"rewards/chosen": -0.6032779216766357,
|
| 188 |
+
"rewards/margins": 0.45596203207969666,
|
| 189 |
+
"rewards/rejected": -1.0592399835586548,
|
| 190 |
+
"step": 110
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.27522935779816515,
|
| 194 |
+
"grad_norm": 26.734564490123315,
|
| 195 |
+
"learning_rate": 4.550430636492389e-07,
|
| 196 |
+
"logits/chosen": -1.174917459487915,
|
| 197 |
+
"logits/rejected": -1.0646188259124756,
|
| 198 |
+
"logps/chosen": -337.2206726074219,
|
| 199 |
+
"logps/rejected": -367.90960693359375,
|
| 200 |
+
"loss": 0.5593,
|
| 201 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 202 |
+
"rewards/chosen": -0.6140136122703552,
|
| 203 |
+
"rewards/margins": 0.5504432320594788,
|
| 204 |
+
"rewards/rejected": -1.1644567251205444,
|
| 205 |
+
"step": 120
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.2981651376146789,
|
| 209 |
+
"grad_norm": 25.5727518504957,
|
| 210 |
+
"learning_rate": 4.429344633468004e-07,
|
| 211 |
+
"logits/chosen": -0.7371357083320618,
|
| 212 |
+
"logits/rejected": -0.4288739562034607,
|
| 213 |
+
"logps/chosen": -328.1370544433594,
|
| 214 |
+
"logps/rejected": -368.2788391113281,
|
| 215 |
+
"loss": 0.5539,
|
| 216 |
+
"rewards/accuracies": 0.71875,
|
| 217 |
+
"rewards/chosen": -0.7195693850517273,
|
| 218 |
+
"rewards/margins": 0.5370928645133972,
|
| 219 |
+
"rewards/rejected": -1.256662130355835,
|
| 220 |
+
"step": 130
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.3211009174311927,
|
| 224 |
+
"grad_norm": 27.718146660570127,
|
| 225 |
+
"learning_rate": 4.2958733752443187e-07,
|
| 226 |
+
"logits/chosen": -0.9468735456466675,
|
| 227 |
+
"logits/rejected": -0.6918075680732727,
|
| 228 |
+
"logps/chosen": -324.7828063964844,
|
| 229 |
+
"logps/rejected": -322.2585144042969,
|
| 230 |
+
"loss": 0.5515,
|
| 231 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 232 |
+
"rewards/chosen": -0.48597660660743713,
|
| 233 |
+
"rewards/margins": 0.4936434328556061,
|
| 234 |
+
"rewards/rejected": -0.9796198606491089,
|
| 235 |
+
"step": 140
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.3440366972477064,
|
| 239 |
+
"grad_norm": 27.968029844080455,
|
| 240 |
+
"learning_rate": 4.150873668617898e-07,
|
| 241 |
+
"logits/chosen": -0.30871185660362244,
|
| 242 |
+
"logits/rejected": 0.24544291198253632,
|
| 243 |
+
"logps/chosen": -354.21392822265625,
|
| 244 |
+
"logps/rejected": -385.76666259765625,
|
| 245 |
+
"loss": 0.5389,
|
| 246 |
+
"rewards/accuracies": 0.706250011920929,
|
| 247 |
+
"rewards/chosen": -0.8767034411430359,
|
| 248 |
+
"rewards/margins": 0.6490874290466309,
|
| 249 |
+
"rewards/rejected": -1.5257909297943115,
|
| 250 |
+
"step": 150
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.3669724770642202,
|
| 254 |
+
"grad_norm": 28.129228197799993,
|
| 255 |
+
"learning_rate": 3.9952763262280397e-07,
|
| 256 |
+
"logits/chosen": -0.545559287071228,
|
| 257 |
+
"logits/rejected": -0.1624789535999298,
|
| 258 |
+
"logps/chosen": -366.7821350097656,
|
| 259 |
+
"logps/rejected": -417.90545654296875,
|
| 260 |
+
"loss": 0.5313,
|
| 261 |
+
"rewards/accuracies": 0.699999988079071,
|
| 262 |
+
"rewards/chosen": -0.7023102045059204,
|
| 263 |
+
"rewards/margins": 0.6170581579208374,
|
| 264 |
+
"rewards/rejected": -1.3193682432174683,
|
| 265 |
+
"step": 160
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.38990825688073394,
|
| 269 |
+
"grad_norm": 22.14745919356143,
|
| 270 |
+
"learning_rate": 3.8300801912883414e-07,
|
| 271 |
+
"logits/chosen": 0.30408772826194763,
|
| 272 |
+
"logits/rejected": 0.8390473127365112,
|
| 273 |
+
"logps/chosen": -362.6156311035156,
|
| 274 |
+
"logps/rejected": -431.7569274902344,
|
| 275 |
+
"loss": 0.5007,
|
| 276 |
+
"rewards/accuracies": 0.71875,
|
| 277 |
+
"rewards/chosen": -0.9101476669311523,
|
| 278 |
+
"rewards/margins": 0.7067325711250305,
|
| 279 |
+
"rewards/rejected": -1.616880178451538,
|
| 280 |
+
"step": 170
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.41284403669724773,
|
| 284 |
+
"grad_norm": 21.278511576979266,
|
| 285 |
+
"learning_rate": 3.6563457256020884e-07,
|
| 286 |
+
"logits/chosen": 0.29116564989089966,
|
| 287 |
+
"logits/rejected": 1.1298694610595703,
|
| 288 |
+
"logps/chosen": -386.0079650878906,
|
| 289 |
+
"logps/rejected": -382.1528625488281,
|
| 290 |
+
"loss": 0.5349,
|
| 291 |
+
"rewards/accuracies": 0.699999988079071,
|
| 292 |
+
"rewards/chosen": -0.8265001177787781,
|
| 293 |
+
"rewards/margins": 0.6100791692733765,
|
| 294 |
+
"rewards/rejected": -1.4365794658660889,
|
| 295 |
+
"step": 180
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.43577981651376146,
|
| 299 |
+
"grad_norm": 27.634146578624538,
|
| 300 |
+
"learning_rate": 3.475188202022617e-07,
|
| 301 |
+
"logits/chosen": 0.34451404213905334,
|
| 302 |
+
"logits/rejected": 0.971908688545227,
|
| 303 |
+
"logps/chosen": -325.24951171875,
|
| 304 |
+
"logps/rejected": -408.52886962890625,
|
| 305 |
+
"loss": 0.5132,
|
| 306 |
+
"rewards/accuracies": 0.768750011920929,
|
| 307 |
+
"rewards/chosen": -0.6713501811027527,
|
| 308 |
+
"rewards/margins": 0.7704941034317017,
|
| 309 |
+
"rewards/rejected": -1.4418442249298096,
|
| 310 |
+
"step": 190
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.45871559633027525,
|
| 314 |
+
"grad_norm": 23.689525185525035,
|
| 315 |
+
"learning_rate": 3.287770545059052e-07,
|
| 316 |
+
"logits/chosen": 0.3892287611961365,
|
| 317 |
+
"logits/rejected": 1.2381625175476074,
|
| 318 |
+
"logps/chosen": -350.74658203125,
|
| 319 |
+
"logps/rejected": -392.03265380859375,
|
| 320 |
+
"loss": 0.5281,
|
| 321 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 322 |
+
"rewards/chosen": -0.7246497869491577,
|
| 323 |
+
"rewards/margins": 0.7571171522140503,
|
| 324 |
+
"rewards/rejected": -1.4817668199539185,
|
| 325 |
+
"step": 200
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.45871559633027525,
|
| 329 |
+
"eval_logits/chosen": -0.12046875804662704,
|
| 330 |
+
"eval_logits/rejected": 1.0060540437698364,
|
| 331 |
+
"eval_logps/chosen": -347.0378112792969,
|
| 332 |
+
"eval_logps/rejected": -388.8913269042969,
|
| 333 |
+
"eval_loss": 0.5102300047874451,
|
| 334 |
+
"eval_rewards/accuracies": 0.7456896305084229,
|
| 335 |
+
"eval_rewards/chosen": -0.6194772124290466,
|
| 336 |
+
"eval_rewards/margins": 0.804211437702179,
|
| 337 |
+
"eval_rewards/rejected": -1.4236886501312256,
|
| 338 |
+
"eval_runtime": 96.4808,
|
| 339 |
+
"eval_samples_per_second": 18.843,
|
| 340 |
+
"eval_steps_per_second": 0.301,
|
| 341 |
+
"step": 200
|
| 342 |
+
},
|
| 343 |
+
{
|
| 344 |
+
"epoch": 0.481651376146789,
|
| 345 |
+
"grad_norm": 29.46326810590551,
|
| 346 |
+
"learning_rate": 3.0952958655864954e-07,
|
| 347 |
+
"logits/chosen": 0.42962178587913513,
|
| 348 |
+
"logits/rejected": 1.0683366060256958,
|
| 349 |
+
"logps/chosen": -348.16741943359375,
|
| 350 |
+
"logps/rejected": -402.72344970703125,
|
| 351 |
+
"loss": 0.5081,
|
| 352 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 353 |
+
"rewards/chosen": -0.7486302256584167,
|
| 354 |
+
"rewards/margins": 0.7351824045181274,
|
| 355 |
+
"rewards/rejected": -1.4838125705718994,
|
| 356 |
+
"step": 210
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.5045871559633027,
|
| 360 |
+
"grad_norm": 25.818080936093143,
|
| 361 |
+
"learning_rate": 2.898999737583448e-07,
|
| 362 |
+
"logits/chosen": 1.011344075202942,
|
| 363 |
+
"logits/rejected": 2.023766279220581,
|
| 364 |
+
"logps/chosen": -427.4793395996094,
|
| 365 |
+
"logps/rejected": -478.3731994628906,
|
| 366 |
+
"loss": 0.4988,
|
| 367 |
+
"rewards/accuracies": 0.8125,
|
| 368 |
+
"rewards/chosen": -1.0479902029037476,
|
| 369 |
+
"rewards/margins": 0.8860396146774292,
|
| 370 |
+
"rewards/rejected": -1.9340295791625977,
|
| 371 |
+
"step": 220
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.5275229357798165,
|
| 375 |
+
"grad_norm": 22.47549460336781,
|
| 376 |
+
"learning_rate": 2.7001422664752333e-07,
|
| 377 |
+
"logits/chosen": 0.21876761317253113,
|
| 378 |
+
"logits/rejected": 0.814139723777771,
|
| 379 |
+
"logps/chosen": -328.03802490234375,
|
| 380 |
+
"logps/rejected": -390.96160888671875,
|
| 381 |
+
"loss": 0.5217,
|
| 382 |
+
"rewards/accuracies": 0.737500011920929,
|
| 383 |
+
"rewards/chosen": -0.600945234298706,
|
| 384 |
+
"rewards/margins": 0.6524718403816223,
|
| 385 |
+
"rewards/rejected": -1.2534170150756836,
|
| 386 |
+
"step": 230
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.5504587155963303,
|
| 390 |
+
"grad_norm": 21.171293378773015,
|
| 391 |
+
"learning_rate": 2.5e-07,
|
| 392 |
+
"logits/chosen": -0.24384799599647522,
|
| 393 |
+
"logits/rejected": 0.6541222333908081,
|
| 394 |
+
"logps/chosen": -352.31732177734375,
|
| 395 |
+
"logps/rejected": -387.68975830078125,
|
| 396 |
+
"loss": 0.5355,
|
| 397 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 398 |
+
"rewards/chosen": -0.6478549838066101,
|
| 399 |
+
"rewards/margins": 0.6155428886413574,
|
| 400 |
+
"rewards/rejected": -1.2633978128433228,
|
| 401 |
+
"step": 240
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.573394495412844,
|
| 405 |
+
"grad_norm": 20.854041628569117,
|
| 406 |
+
"learning_rate": 2.2998577335247667e-07,
|
| 407 |
+
"logits/chosen": 0.6312379837036133,
|
| 408 |
+
"logits/rejected": 1.7261543273925781,
|
| 409 |
+
"logps/chosen": -404.70721435546875,
|
| 410 |
+
"logps/rejected": -441.50775146484375,
|
| 411 |
+
"loss": 0.5067,
|
| 412 |
+
"rewards/accuracies": 0.768750011920929,
|
| 413 |
+
"rewards/chosen": -1.0091339349746704,
|
| 414 |
+
"rewards/margins": 0.801504909992218,
|
| 415 |
+
"rewards/rejected": -1.8106390237808228,
|
| 416 |
+
"step": 250
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5963302752293578,
|
| 420 |
+
"grad_norm": 22.834810900622486,
|
| 421 |
+
"learning_rate": 2.1010002624165524e-07,
|
| 422 |
+
"logits/chosen": 1.0223064422607422,
|
| 423 |
+
"logits/rejected": 1.8627128601074219,
|
| 424 |
+
"logps/chosen": -392.29034423828125,
|
| 425 |
+
"logps/rejected": -494.05767822265625,
|
| 426 |
+
"loss": 0.5001,
|
| 427 |
+
"rewards/accuracies": 0.793749988079071,
|
| 428 |
+
"rewards/chosen": -1.215757966041565,
|
| 429 |
+
"rewards/margins": 0.9808815121650696,
|
| 430 |
+
"rewards/rejected": -2.1966395378112793,
|
| 431 |
+
"step": 260
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.6192660550458715,
|
| 435 |
+
"grad_norm": 24.01182646236212,
|
| 436 |
+
"learning_rate": 1.9047041344135043e-07,
|
| 437 |
+
"logits/chosen": 0.8764435052871704,
|
| 438 |
+
"logits/rejected": 1.520572304725647,
|
| 439 |
+
"logps/chosen": -372.30413818359375,
|
| 440 |
+
"logps/rejected": -447.957763671875,
|
| 441 |
+
"loss": 0.4994,
|
| 442 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 443 |
+
"rewards/chosen": -1.0624181032180786,
|
| 444 |
+
"rewards/margins": 0.8986889123916626,
|
| 445 |
+
"rewards/rejected": -1.9611070156097412,
|
| 446 |
+
"step": 270
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.6422018348623854,
|
| 450 |
+
"grad_norm": 27.05306764931255,
|
| 451 |
+
"learning_rate": 1.7122294549409482e-07,
|
| 452 |
+
"logits/chosen": 0.5640047788619995,
|
| 453 |
+
"logits/rejected": 1.584166169166565,
|
| 454 |
+
"logps/chosen": -370.31890869140625,
|
| 455 |
+
"logps/rejected": -454.9625549316406,
|
| 456 |
+
"loss": 0.5238,
|
| 457 |
+
"rewards/accuracies": 0.78125,
|
| 458 |
+
"rewards/chosen": -0.9732006192207336,
|
| 459 |
+
"rewards/margins": 0.9384667277336121,
|
| 460 |
+
"rewards/rejected": -1.9116674661636353,
|
| 461 |
+
"step": 280
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.6651376146788991,
|
| 465 |
+
"grad_norm": 26.88772523041706,
|
| 466 |
+
"learning_rate": 1.524811797977383e-07,
|
| 467 |
+
"logits/chosen": 0.8203527331352234,
|
| 468 |
+
"logits/rejected": 1.9352061748504639,
|
| 469 |
+
"logps/chosen": -392.50665283203125,
|
| 470 |
+
"logps/rejected": -447.2642517089844,
|
| 471 |
+
"loss": 0.5027,
|
| 472 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 473 |
+
"rewards/chosen": -1.0477584600448608,
|
| 474 |
+
"rewards/margins": 0.8857167363166809,
|
| 475 |
+
"rewards/rejected": -1.9334752559661865,
|
| 476 |
+
"step": 290
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.6880733944954128,
|
| 480 |
+
"grad_norm": 23.287905403576715,
|
| 481 |
+
"learning_rate": 1.3436542743979125e-07,
|
| 482 |
+
"logits/chosen": 0.5250275731086731,
|
| 483 |
+
"logits/rejected": 1.2394458055496216,
|
| 484 |
+
"logps/chosen": -408.4005432128906,
|
| 485 |
+
"logps/rejected": -430.564453125,
|
| 486 |
+
"loss": 0.5213,
|
| 487 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 488 |
+
"rewards/chosen": -1.0245119333267212,
|
| 489 |
+
"rewards/margins": 0.6925337314605713,
|
| 490 |
+
"rewards/rejected": -1.717045545578003,
|
| 491 |
+
"step": 300
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6880733944954128,
|
| 495 |
+
"eval_logits/chosen": 0.3984619081020355,
|
| 496 |
+
"eval_logits/rejected": 1.7280592918395996,
|
| 497 |
+
"eval_logps/chosen": -379.9385070800781,
|
| 498 |
+
"eval_logps/rejected": -434.2354431152344,
|
| 499 |
+
"eval_loss": 0.4939006567001343,
|
| 500 |
+
"eval_rewards/accuracies": 0.7543103694915771,
|
| 501 |
+
"eval_rewards/chosen": -0.9484842419624329,
|
| 502 |
+
"eval_rewards/margins": 0.9286458492279053,
|
| 503 |
+
"eval_rewards/rejected": -1.877130150794983,
|
| 504 |
+
"eval_runtime": 93.6728,
|
| 505 |
+
"eval_samples_per_second": 19.408,
|
| 506 |
+
"eval_steps_per_second": 0.31,
|
| 507 |
+
"step": 300
|
| 508 |
+
},
|
| 509 |
+
{
|
| 510 |
+
"epoch": 0.7110091743119266,
|
| 511 |
+
"grad_norm": 27.450541462100524,
|
| 512 |
+
"learning_rate": 1.1699198087116588e-07,
|
| 513 |
+
"logits/chosen": 0.4360331594944,
|
| 514 |
+
"logits/rejected": 1.4486221075057983,
|
| 515 |
+
"logps/chosen": -380.5530700683594,
|
| 516 |
+
"logps/rejected": -445.94012451171875,
|
| 517 |
+
"loss": 0.5239,
|
| 518 |
+
"rewards/accuracies": 0.75,
|
| 519 |
+
"rewards/chosen": -1.0258790254592896,
|
| 520 |
+
"rewards/margins": 0.7690252065658569,
|
| 521 |
+
"rewards/rejected": -1.794904351234436,
|
| 522 |
+
"step": 310
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.7339449541284404,
|
| 526 |
+
"grad_norm": 28.27778056239839,
|
| 527 |
+
"learning_rate": 1.00472367377196e-07,
|
| 528 |
+
"logits/chosen": 0.6628949642181396,
|
| 529 |
+
"logits/rejected": 2.0110418796539307,
|
| 530 |
+
"logps/chosen": -364.7244873046875,
|
| 531 |
+
"logps/rejected": -434.63568115234375,
|
| 532 |
+
"loss": 0.5013,
|
| 533 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 534 |
+
"rewards/chosen": -0.9049250483512878,
|
| 535 |
+
"rewards/margins": 1.0965297222137451,
|
| 536 |
+
"rewards/rejected": -2.0014548301696777,
|
| 537 |
+
"step": 320
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.7568807339449541,
|
| 541 |
+
"grad_norm": 29.667411994858202,
|
| 542 |
+
"learning_rate": 8.49126331382102e-08,
|
| 543 |
+
"logits/chosen": 0.9564372897148132,
|
| 544 |
+
"logits/rejected": 1.7951595783233643,
|
| 545 |
+
"logps/chosen": -377.6117858886719,
|
| 546 |
+
"logps/rejected": -421.7931213378906,
|
| 547 |
+
"loss": 0.5021,
|
| 548 |
+
"rewards/accuracies": 0.706250011920929,
|
| 549 |
+
"rewards/chosen": -1.030896544456482,
|
| 550 |
+
"rewards/margins": 0.7160336375236511,
|
| 551 |
+
"rewards/rejected": -1.7469301223754883,
|
| 552 |
+
"step": 330
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.7798165137614679,
|
| 556 |
+
"grad_norm": 30.895742020543867,
|
| 557 |
+
"learning_rate": 7.041266247556812e-08,
|
| 558 |
+
"logits/chosen": 0.8280539512634277,
|
| 559 |
+
"logits/rejected": 1.6557743549346924,
|
| 560 |
+
"logps/chosen": -388.16351318359375,
|
| 561 |
+
"logps/rejected": -436.76531982421875,
|
| 562 |
+
"loss": 0.4897,
|
| 563 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 564 |
+
"rewards/chosen": -1.001708984375,
|
| 565 |
+
"rewards/margins": 0.7661042809486389,
|
| 566 |
+
"rewards/rejected": -1.7678134441375732,
|
| 567 |
+
"step": 340
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.8027522935779816,
|
| 571 |
+
"grad_norm": 27.764567632511632,
|
| 572 |
+
"learning_rate": 5.706553665319955e-08,
|
| 573 |
+
"logits/chosen": 0.9885643124580383,
|
| 574 |
+
"logits/rejected": 2.0599663257598877,
|
| 575 |
+
"logps/chosen": -376.7607727050781,
|
| 576 |
+
"logps/rejected": -424.14910888671875,
|
| 577 |
+
"loss": 0.5125,
|
| 578 |
+
"rewards/accuracies": 0.800000011920929,
|
| 579 |
+
"rewards/chosen": -1.0226970911026,
|
| 580 |
+
"rewards/margins": 0.8421795964241028,
|
| 581 |
+
"rewards/rejected": -1.8648765087127686,
|
| 582 |
+
"step": 350
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.8256880733944955,
|
| 586 |
+
"grad_norm": 25.930811449352582,
|
| 587 |
+
"learning_rate": 4.4956936350761005e-08,
|
| 588 |
+
"logits/chosen": 0.9005087614059448,
|
| 589 |
+
"logits/rejected": 1.6520808935165405,
|
| 590 |
+
"logps/chosen": -343.540771484375,
|
| 591 |
+
"logps/rejected": -431.3505859375,
|
| 592 |
+
"loss": 0.4866,
|
| 593 |
+
"rewards/accuracies": 0.699999988079071,
|
| 594 |
+
"rewards/chosen": -0.912207305431366,
|
| 595 |
+
"rewards/margins": 0.8159275054931641,
|
| 596 |
+
"rewards/rejected": -1.7281348705291748,
|
| 597 |
+
"step": 360
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.8486238532110092,
|
| 601 |
+
"grad_norm": 26.612138820793557,
|
| 602 |
+
"learning_rate": 3.416459164418123e-08,
|
| 603 |
+
"logits/chosen": 0.6678069829940796,
|
| 604 |
+
"logits/rejected": 1.7626101970672607,
|
| 605 |
+
"logps/chosen": -394.8943176269531,
|
| 606 |
+
"logps/rejected": -445.3042907714844,
|
| 607 |
+
"loss": 0.498,
|
| 608 |
+
"rewards/accuracies": 0.737500011920929,
|
| 609 |
+
"rewards/chosen": -0.9530981183052063,
|
| 610 |
+
"rewards/margins": 0.8741987943649292,
|
| 611 |
+
"rewards/rejected": -1.8272969722747803,
|
| 612 |
+
"step": 370
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.8715596330275229,
|
| 616 |
+
"grad_norm": 33.570889750971375,
|
| 617 |
+
"learning_rate": 2.475778302439524e-08,
|
| 618 |
+
"logits/chosen": 0.7491241693496704,
|
| 619 |
+
"logits/rejected": 2.0040550231933594,
|
| 620 |
+
"logps/chosen": -383.720703125,
|
| 621 |
+
"logps/rejected": -444.7874450683594,
|
| 622 |
+
"loss": 0.5038,
|
| 623 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 624 |
+
"rewards/chosen": -0.8734270930290222,
|
| 625 |
+
"rewards/margins": 0.9978850483894348,
|
| 626 |
+
"rewards/rejected": -1.871312141418457,
|
| 627 |
+
"step": 380
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.8944954128440367,
|
| 631 |
+
"grad_norm": 27.00226950820523,
|
| 632 |
+
"learning_rate": 1.6796896657433805e-08,
|
| 633 |
+
"logits/chosen": 1.115740180015564,
|
| 634 |
+
"logits/rejected": 2.061350107192993,
|
| 635 |
+
"logps/chosen": -346.07574462890625,
|
| 636 |
+
"logps/rejected": -409.1957092285156,
|
| 637 |
+
"loss": 0.5106,
|
| 638 |
+
"rewards/accuracies": 0.75,
|
| 639 |
+
"rewards/chosen": -0.9411827325820923,
|
| 640 |
+
"rewards/margins": 0.8577120900154114,
|
| 641 |
+
"rewards/rejected": -1.7988946437835693,
|
| 642 |
+
"step": 390
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.9174311926605505,
|
| 646 |
+
"grad_norm": 29.536262595417668,
|
| 647 |
+
"learning_rate": 1.0333036740834855e-08,
|
| 648 |
+
"logits/chosen": 1.2691802978515625,
|
| 649 |
+
"logits/rejected": 2.047137975692749,
|
| 650 |
+
"logps/chosen": -316.4430236816406,
|
| 651 |
+
"logps/rejected": -404.92962646484375,
|
| 652 |
+
"loss": 0.5037,
|
| 653 |
+
"rewards/accuracies": 0.75,
|
| 654 |
+
"rewards/chosen": -0.91883385181427,
|
| 655 |
+
"rewards/margins": 0.8383063077926636,
|
| 656 |
+
"rewards/rejected": -1.7571399211883545,
|
| 657 |
+
"step": 400
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.9174311926605505,
|
| 661 |
+
"eval_logits/chosen": 0.9259076118469238,
|
| 662 |
+
"eval_logits/rejected": 2.242485761642456,
|
| 663 |
+
"eval_logps/chosen": -379.86492919921875,
|
| 664 |
+
"eval_logps/rejected": -438.6234436035156,
|
| 665 |
+
"eval_loss": 0.48977744579315186,
|
| 666 |
+
"eval_rewards/accuracies": 0.7543103694915771,
|
| 667 |
+
"eval_rewards/chosen": -0.9477482438087463,
|
| 668 |
+
"eval_rewards/margins": 0.9732612371444702,
|
| 669 |
+
"eval_rewards/rejected": -1.9210097789764404,
|
| 670 |
+
"eval_runtime": 94.7635,
|
| 671 |
+
"eval_samples_per_second": 19.185,
|
| 672 |
+
"eval_steps_per_second": 0.306,
|
| 673 |
+
"step": 400
|
| 674 |
+
},
|
| 675 |
+
{
|
| 676 |
+
"epoch": 0.9403669724770642,
|
| 677 |
+
"grad_norm": 34.10288961478374,
|
| 678 |
+
"learning_rate": 5.4076974448211685e-09,
|
| 679 |
+
"logits/chosen": 1.2957347631454468,
|
| 680 |
+
"logits/rejected": 2.590200901031494,
|
| 681 |
+
"logps/chosen": -376.28472900390625,
|
| 682 |
+
"logps/rejected": -439.2782287597656,
|
| 683 |
+
"loss": 0.4987,
|
| 684 |
+
"rewards/accuracies": 0.75,
|
| 685 |
+
"rewards/chosen": -1.053328275680542,
|
| 686 |
+
"rewards/margins": 0.9589872360229492,
|
| 687 |
+
"rewards/rejected": -2.012315273284912,
|
| 688 |
+
"step": 410
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.963302752293578,
|
| 692 |
+
"grad_norm": 33.702853354259936,
|
| 693 |
+
"learning_rate": 2.052496544188487e-09,
|
| 694 |
+
"logits/chosen": 1.3565906286239624,
|
| 695 |
+
"logits/rejected": 2.3881638050079346,
|
| 696 |
+
"logps/chosen": -364.3350830078125,
|
| 697 |
+
"logps/rejected": -450.5853576660156,
|
| 698 |
+
"loss": 0.4968,
|
| 699 |
+
"rewards/accuracies": 0.75,
|
| 700 |
+
"rewards/chosen": -1.0489609241485596,
|
| 701 |
+
"rewards/margins": 0.9981186985969543,
|
| 702 |
+
"rewards/rejected": -2.047079563140869,
|
| 703 |
+
"step": 420
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.9862385321100917,
|
| 707 |
+
"grad_norm": 31.154782672945057,
|
| 708 |
+
"learning_rate": 2.889724508297886e-10,
|
| 709 |
+
"logits/chosen": 0.9770921468734741,
|
| 710 |
+
"logits/rejected": 2.4192652702331543,
|
| 711 |
+
"logps/chosen": -402.85589599609375,
|
| 712 |
+
"logps/rejected": -433.6226501464844,
|
| 713 |
+
"loss": 0.4895,
|
| 714 |
+
"rewards/accuracies": 0.737500011920929,
|
| 715 |
+
"rewards/chosen": -0.9968045949935913,
|
| 716 |
+
"rewards/margins": 0.8990565538406372,
|
| 717 |
+
"rewards/rejected": -1.895861268043518,
|
| 718 |
+
"step": 430
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 1.0,
|
| 722 |
+
"step": 436,
|
| 723 |
+
"total_flos": 0.0,
|
| 724 |
+
"train_loss": 0.5438270043889317,
|
| 725 |
+
"train_runtime": 11297.0501,
|
| 726 |
+
"train_samples_per_second": 4.936,
|
| 727 |
+
"train_steps_per_second": 0.039
|
| 728 |
+
}
|
| 729 |
+
],
|
| 730 |
+
"logging_steps": 10,
|
| 731 |
+
"max_steps": 436,
|
| 732 |
+
"num_input_tokens_seen": 0,
|
| 733 |
+
"num_train_epochs": 1,
|
| 734 |
+
"save_steps": 100,
|
| 735 |
+
"stateful_callbacks": {
|
| 736 |
+
"TrainerControl": {
|
| 737 |
+
"args": {
|
| 738 |
+
"should_epoch_stop": false,
|
| 739 |
+
"should_evaluate": false,
|
| 740 |
+
"should_log": false,
|
| 741 |
+
"should_save": true,
|
| 742 |
+
"should_training_stop": true
|
| 743 |
+
},
|
| 744 |
+
"attributes": {}
|
| 745 |
+
}
|
| 746 |
+
},
|
| 747 |
+
"total_flos": 0.0,
|
| 748 |
+
"train_batch_size": 8,
|
| 749 |
+
"trial_name": null,
|
| 750 |
+
"trial_params": null
|
| 751 |
+
}
|