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
- runs/Dec10_16-40-16_n124-104-101/events.out.tfevents.1733850033.n124-104-101.588222.0 +2 -2
- train_results.json +9 -0
- trainer_state.json +826 -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
|
| 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
|
| 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 an unknown dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.5031
|
| 21 |
+
- Rewards/chosen: -0.8365
|
| 22 |
+
- Rewards/rejected: -1.7881
|
| 23 |
+
- Rewards/accuracies: 0.7852
|
| 24 |
+
- Rewards/margins: 0.9517
|
| 25 |
+
- Logps/rejected: -441.4968
|
| 26 |
+
- Logps/chosen: -346.2211
|
| 27 |
+
- Logits/rejected: 1.6106
|
| 28 |
+
- Logits/chosen: 0.7959
|
| 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: 42
|
| 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 |
+
| 0.5709 | 0.2092 | 100 | 0.5765 | -0.3415 | -0.8321 | 0.7305 | 0.4906 | -345.8958 | -296.7220 | -1.0885 | -1.2465 |
|
| 66 |
+
| 0.5427 | 0.4184 | 200 | 0.5256 | -0.7352 | -1.5375 | 0.7695 | 0.8022 | -416.4311 | -336.0986 | 0.9059 | 0.1815 |
|
| 67 |
+
| 0.4892 | 0.6276 | 300 | 0.5082 | -0.8910 | -1.8210 | 0.7695 | 0.9300 | -444.7822 | -351.6719 | 1.3892 | 0.5828 |
|
| 68 |
+
| 0.5037 | 0.8368 | 400 | 0.5031 | -0.8365 | -1.7881 | 0.7852 | 0.9517 | -441.4968 | -346.2211 | 1.6106 | 0.7959 |
|
| 69 |
+
|
| 70 |
+
|
| 71 |
+
### Framework versions
|
| 72 |
+
|
| 73 |
+
- Transformers 4.44.0
|
| 74 |
+
- Pytorch 2.4.0
|
| 75 |
+
- Datasets 3.1.0
|
| 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": 0.5395244165444474,
|
| 5 |
+
"train_runtime": 4445.7731,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 13.751,
|
| 8 |
+
"train_steps_per_second": 0.108
|
| 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"
|
| 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 sha256:0979c0f1f3c4896df5330a2d5c3b64225ba8a6b6b380c36056ef0f7ec372afe3
|
| 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 sha256:383002490eb2b5af350e9f776c27c9b0a54ad89e7620d829add4f498581960bd
|
| 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 sha256:77458028ddcda56dbb7ead56719d829e1f0a1b87dc214ece421ae26ed7f7fa0f
|
| 3 |
size 4540516344
|
runs/Dec10_16-40-16_n124-104-101/events.out.tfevents.1733850033.n124-104-101.588222.0
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a7724901a79aa8d1d3992c9ff05aca2fc3036543dd37bab7949df80a18fd456a
|
| 3 |
+
size 42430
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5395244165444474,
|
| 5 |
+
"train_runtime": 4445.7731,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 13.751,
|
| 8 |
+
"train_steps_per_second": 0.108
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,826 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 478,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0020920502092050207,
|
| 13 |
+
"grad_norm": 9.085185309257627,
|
| 14 |
+
"learning_rate": 1.0416666666666666e-08,
|
| 15 |
+
"logits/chosen": -2.7662220001220703,
|
| 16 |
+
"logits/rejected": -2.7178127765655518,
|
| 17 |
+
"logps/chosen": -269.6776123046875,
|
| 18 |
+
"logps/rejected": -360.6510314941406,
|
| 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": 0.02092050209205021,
|
| 28 |
+
"grad_norm": 9.1082652647099,
|
| 29 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 30 |
+
"logits/chosen": -2.5920939445495605,
|
| 31 |
+
"logits/rejected": -2.5626003742218018,
|
| 32 |
+
"logps/chosen": -264.7879638671875,
|
| 33 |
+
"logps/rejected": -251.56271362304688,
|
| 34 |
+
"loss": 0.6931,
|
| 35 |
+
"rewards/accuracies": 0.4652777910232544,
|
| 36 |
+
"rewards/chosen": -0.00015384703874588013,
|
| 37 |
+
"rewards/margins": 0.00021662651852238923,
|
| 38 |
+
"rewards/rejected": -0.00037047354271635413,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.04184100418410042,
|
| 43 |
+
"grad_norm": 8.403069183005119,
|
| 44 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 45 |
+
"logits/chosen": -2.6531224250793457,
|
| 46 |
+
"logits/rejected": -2.6051571369171143,
|
| 47 |
+
"logps/chosen": -281.4217224121094,
|
| 48 |
+
"logps/rejected": -296.8309020996094,
|
| 49 |
+
"loss": 0.6923,
|
| 50 |
+
"rewards/accuracies": 0.5,
|
| 51 |
+
"rewards/chosen": 0.0006401228019967675,
|
| 52 |
+
"rewards/margins": 0.0008313875878229737,
|
| 53 |
+
"rewards/rejected": -0.00019126484403386712,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.06276150627615062,
|
| 58 |
+
"grad_norm": 8.308611367087158,
|
| 59 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 60 |
+
"logits/chosen": -2.6641345024108887,
|
| 61 |
+
"logits/rejected": -2.591866970062256,
|
| 62 |
+
"logps/chosen": -299.44696044921875,
|
| 63 |
+
"logps/rejected": -263.68865966796875,
|
| 64 |
+
"loss": 0.6886,
|
| 65 |
+
"rewards/accuracies": 0.625,
|
| 66 |
+
"rewards/chosen": 0.012143267318606377,
|
| 67 |
+
"rewards/margins": 0.009724361822009087,
|
| 68 |
+
"rewards/rejected": 0.0024189057294279337,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.08368200836820083,
|
| 73 |
+
"grad_norm": 8.158936380689568,
|
| 74 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 75 |
+
"logits/chosen": -2.556858539581299,
|
| 76 |
+
"logits/rejected": -2.521756172180176,
|
| 77 |
+
"logps/chosen": -263.13677978515625,
|
| 78 |
+
"logps/rejected": -245.13369750976562,
|
| 79 |
+
"loss": 0.6778,
|
| 80 |
+
"rewards/accuracies": 0.71875,
|
| 81 |
+
"rewards/chosen": 0.039376698434352875,
|
| 82 |
+
"rewards/margins": 0.04692060127854347,
|
| 83 |
+
"rewards/rejected": -0.007543901912868023,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.10460251046025104,
|
| 88 |
+
"grad_norm": 8.442774298195335,
|
| 89 |
+
"learning_rate": 4.999733114418725e-07,
|
| 90 |
+
"logits/chosen": -2.4983766078948975,
|
| 91 |
+
"logits/rejected": -2.458930730819702,
|
| 92 |
+
"logps/chosen": -279.6597595214844,
|
| 93 |
+
"logps/rejected": -297.4883117675781,
|
| 94 |
+
"loss": 0.6615,
|
| 95 |
+
"rewards/accuracies": 0.612500011920929,
|
| 96 |
+
"rewards/chosen": 0.0010005368385463953,
|
| 97 |
+
"rewards/margins": 0.048170365393161774,
|
| 98 |
+
"rewards/rejected": -0.04716982692480087,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.12552301255230125,
|
| 103 |
+
"grad_norm": 11.320329514836741,
|
| 104 |
+
"learning_rate": 4.990398100856366e-07,
|
| 105 |
+
"logits/chosen": -2.514209270477295,
|
| 106 |
+
"logits/rejected": -2.4425783157348633,
|
| 107 |
+
"logps/chosen": -273.6862487792969,
|
| 108 |
+
"logps/rejected": -312.9122314453125,
|
| 109 |
+
"loss": 0.6389,
|
| 110 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 111 |
+
"rewards/chosen": -0.020198455080389977,
|
| 112 |
+
"rewards/margins": 0.12713830173015594,
|
| 113 |
+
"rewards/rejected": -0.14733675122261047,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.14644351464435146,
|
| 118 |
+
"grad_norm": 10.658450428160506,
|
| 119 |
+
"learning_rate": 4.967775735898179e-07,
|
| 120 |
+
"logits/chosen": -2.3650529384613037,
|
| 121 |
+
"logits/rejected": -2.3627445697784424,
|
| 122 |
+
"logps/chosen": -270.68896484375,
|
| 123 |
+
"logps/rejected": -283.2526550292969,
|
| 124 |
+
"loss": 0.6157,
|
| 125 |
+
"rewards/accuracies": 0.762499988079071,
|
| 126 |
+
"rewards/chosen": -0.05055179074406624,
|
| 127 |
+
"rewards/margins": 0.21437402069568634,
|
| 128 |
+
"rewards/rejected": -0.26492583751678467,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.16736401673640167,
|
| 133 |
+
"grad_norm": 18.344270948626654,
|
| 134 |
+
"learning_rate": 4.931986719649298e-07,
|
| 135 |
+
"logits/chosen": -2.5614020824432373,
|
| 136 |
+
"logits/rejected": -2.4784748554229736,
|
| 137 |
+
"logps/chosen": -336.35430908203125,
|
| 138 |
+
"logps/rejected": -303.9096374511719,
|
| 139 |
+
"loss": 0.6025,
|
| 140 |
+
"rewards/accuracies": 0.6875,
|
| 141 |
+
"rewards/chosen": -0.1354917585849762,
|
| 142 |
+
"rewards/margins": 0.26337119936943054,
|
| 143 |
+
"rewards/rejected": -0.39886292815208435,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.18828451882845187,
|
| 148 |
+
"grad_norm": 19.12478751671118,
|
| 149 |
+
"learning_rate": 4.883222001996351e-07,
|
| 150 |
+
"logits/chosen": -2.134124755859375,
|
| 151 |
+
"logits/rejected": -2.0814406871795654,
|
| 152 |
+
"logps/chosen": -305.53082275390625,
|
| 153 |
+
"logps/rejected": -345.31109619140625,
|
| 154 |
+
"loss": 0.5741,
|
| 155 |
+
"rewards/accuracies": 0.737500011920929,
|
| 156 |
+
"rewards/chosen": -0.4640178680419922,
|
| 157 |
+
"rewards/margins": 0.46487826108932495,
|
| 158 |
+
"rewards/rejected": -0.9288961291313171,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.20920502092050208,
|
| 163 |
+
"grad_norm": 17.356342239295923,
|
| 164 |
+
"learning_rate": 4.821741763807186e-07,
|
| 165 |
+
"logits/chosen": -1.3313448429107666,
|
| 166 |
+
"logits/rejected": -1.1910110712051392,
|
| 167 |
+
"logps/chosen": -308.7679138183594,
|
| 168 |
+
"logps/rejected": -361.08013916015625,
|
| 169 |
+
"loss": 0.5709,
|
| 170 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 171 |
+
"rewards/chosen": -0.4839816987514496,
|
| 172 |
+
"rewards/margins": 0.5560156106948853,
|
| 173 |
+
"rewards/rejected": -1.0399973392486572,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.20920502092050208,
|
| 178 |
+
"eval_logits/chosen": -1.2465178966522217,
|
| 179 |
+
"eval_logits/rejected": -1.0884699821472168,
|
| 180 |
+
"eval_logps/chosen": -296.7220153808594,
|
| 181 |
+
"eval_logps/rejected": -345.89581298828125,
|
| 182 |
+
"eval_loss": 0.5765112042427063,
|
| 183 |
+
"eval_rewards/accuracies": 0.73046875,
|
| 184 |
+
"eval_rewards/chosen": -0.34146958589553833,
|
| 185 |
+
"eval_rewards/margins": 0.49064701795578003,
|
| 186 |
+
"eval_rewards/rejected": -0.8321166634559631,
|
| 187 |
+
"eval_runtime": 40.3807,
|
| 188 |
+
"eval_samples_per_second": 49.529,
|
| 189 |
+
"eval_steps_per_second": 0.792,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.2301255230125523,
|
| 194 |
+
"grad_norm": 17.47373189781474,
|
| 195 |
+
"learning_rate": 4.747874028753375e-07,
|
| 196 |
+
"logits/chosen": -0.9215749502182007,
|
| 197 |
+
"logits/rejected": -0.43307366967201233,
|
| 198 |
+
"logps/chosen": -360.1551818847656,
|
| 199 |
+
"logps/rejected": -366.24737548828125,
|
| 200 |
+
"loss": 0.5682,
|
| 201 |
+
"rewards/accuracies": 0.706250011920929,
|
| 202 |
+
"rewards/chosen": -0.5557556748390198,
|
| 203 |
+
"rewards/margins": 0.528251588344574,
|
| 204 |
+
"rewards/rejected": -1.0840072631835938,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.2510460251046025,
|
| 209 |
+
"grad_norm": 20.093025180061602,
|
| 210 |
+
"learning_rate": 4.662012913161997e-07,
|
| 211 |
+
"logits/chosen": -0.3439692556858063,
|
| 212 |
+
"logits/rejected": -0.04505181312561035,
|
| 213 |
+
"logps/chosen": -336.732177734375,
|
| 214 |
+
"logps/rejected": -372.484619140625,
|
| 215 |
+
"loss": 0.5504,
|
| 216 |
+
"rewards/accuracies": 0.75,
|
| 217 |
+
"rewards/chosen": -0.5553595423698425,
|
| 218 |
+
"rewards/margins": 0.5710638761520386,
|
| 219 |
+
"rewards/rejected": -1.1264234781265259,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.2719665271966527,
|
| 224 |
+
"grad_norm": 23.028382893640803,
|
| 225 |
+
"learning_rate": 4.5646165232345103e-07,
|
| 226 |
+
"logits/chosen": 0.32941126823425293,
|
| 227 |
+
"logits/rejected": 0.7188884019851685,
|
| 228 |
+
"logps/chosen": -367.50909423828125,
|
| 229 |
+
"logps/rejected": -419.0060119628906,
|
| 230 |
+
"loss": 0.535,
|
| 231 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 232 |
+
"rewards/chosen": -0.8557965159416199,
|
| 233 |
+
"rewards/margins": 0.6497102379798889,
|
| 234 |
+
"rewards/rejected": -1.5055067539215088,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.2928870292887029,
|
| 239 |
+
"grad_norm": 18.677447832810206,
|
| 240 |
+
"learning_rate": 4.456204510851956e-07,
|
| 241 |
+
"logits/chosen": 0.21762561798095703,
|
| 242 |
+
"logits/rejected": 0.6349714398384094,
|
| 243 |
+
"logps/chosen": -381.5950927734375,
|
| 244 |
+
"logps/rejected": -431.7596740722656,
|
| 245 |
+
"loss": 0.534,
|
| 246 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 247 |
+
"rewards/chosen": -0.6563121676445007,
|
| 248 |
+
"rewards/margins": 0.6548308730125427,
|
| 249 |
+
"rewards/rejected": -1.311143159866333,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.3138075313807531,
|
| 254 |
+
"grad_norm": 27.137772429027354,
|
| 255 |
+
"learning_rate": 4.337355301007335e-07,
|
| 256 |
+
"logits/chosen": 0.3337453007698059,
|
| 257 |
+
"logits/rejected": 0.7608418464660645,
|
| 258 |
+
"logps/chosen": -367.8205261230469,
|
| 259 |
+
"logps/rejected": -415.28045654296875,
|
| 260 |
+
"loss": 0.5402,
|
| 261 |
+
"rewards/accuracies": 0.699999988079071,
|
| 262 |
+
"rewards/chosen": -0.8457827568054199,
|
| 263 |
+
"rewards/margins": 0.5450310707092285,
|
| 264 |
+
"rewards/rejected": -1.3908138275146484,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.33472803347280333,
|
| 269 |
+
"grad_norm": 26.991586965942954,
|
| 270 |
+
"learning_rate": 4.2087030056579986e-07,
|
| 271 |
+
"logits/chosen": 0.3724555969238281,
|
| 272 |
+
"logits/rejected": 1.1069540977478027,
|
| 273 |
+
"logps/chosen": -358.0096130371094,
|
| 274 |
+
"logps/rejected": -428.70361328125,
|
| 275 |
+
"loss": 0.5324,
|
| 276 |
+
"rewards/accuracies": 0.737500011920929,
|
| 277 |
+
"rewards/chosen": -0.8497047424316406,
|
| 278 |
+
"rewards/margins": 0.8634451031684875,
|
| 279 |
+
"rewards/rejected": -1.7131500244140625,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.35564853556485354,
|
| 284 |
+
"grad_norm": 25.592749003897193,
|
| 285 |
+
"learning_rate": 4.070934040463998e-07,
|
| 286 |
+
"logits/chosen": 0.5049823522567749,
|
| 287 |
+
"logits/rejected": 0.9404481649398804,
|
| 288 |
+
"logps/chosen": -348.8735046386719,
|
| 289 |
+
"logps/rejected": -392.8125915527344,
|
| 290 |
+
"loss": 0.5363,
|
| 291 |
+
"rewards/accuracies": 0.699999988079071,
|
| 292 |
+
"rewards/chosen": -0.9956283569335938,
|
| 293 |
+
"rewards/margins": 0.620826244354248,
|
| 294 |
+
"rewards/rejected": -1.6164544820785522,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.37656903765690375,
|
| 299 |
+
"grad_norm": 22.372745978564954,
|
| 300 |
+
"learning_rate": 3.9247834624635404e-07,
|
| 301 |
+
"logits/chosen": 0.6127285361289978,
|
| 302 |
+
"logits/rejected": 1.2147839069366455,
|
| 303 |
+
"logps/chosen": -340.34906005859375,
|
| 304 |
+
"logps/rejected": -381.84661865234375,
|
| 305 |
+
"loss": 0.5064,
|
| 306 |
+
"rewards/accuracies": 0.75,
|
| 307 |
+
"rewards/chosen": -0.8403685688972473,
|
| 308 |
+
"rewards/margins": 0.7025831937789917,
|
| 309 |
+
"rewards/rejected": -1.5429518222808838,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.39748953974895396,
|
| 314 |
+
"grad_norm": 29.66314474064595,
|
| 315 |
+
"learning_rate": 3.7710310482256523e-07,
|
| 316 |
+
"logits/chosen": 0.3294413983821869,
|
| 317 |
+
"logits/rejected": 0.9074909090995789,
|
| 318 |
+
"logps/chosen": -347.7164611816406,
|
| 319 |
+
"logps/rejected": -403.9298400878906,
|
| 320 |
+
"loss": 0.5209,
|
| 321 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 322 |
+
"rewards/chosen": -0.7645334601402283,
|
| 323 |
+
"rewards/margins": 0.67705237865448,
|
| 324 |
+
"rewards/rejected": -1.441585898399353,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.41841004184100417,
|
| 329 |
+
"grad_norm": 58.26403433971581,
|
| 330 |
+
"learning_rate": 3.610497133404795e-07,
|
| 331 |
+
"logits/chosen": 0.549667477607727,
|
| 332 |
+
"logits/rejected": 1.1599102020263672,
|
| 333 |
+
"logps/chosen": -346.0694580078125,
|
| 334 |
+
"logps/rejected": -415.4115295410156,
|
| 335 |
+
"loss": 0.5427,
|
| 336 |
+
"rewards/accuracies": 0.75,
|
| 337 |
+
"rewards/chosen": -0.8615196943283081,
|
| 338 |
+
"rewards/margins": 0.7718194723129272,
|
| 339 |
+
"rewards/rejected": -1.6333389282226562,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.41841004184100417,
|
| 344 |
+
"eval_logits/chosen": 0.18153175711631775,
|
| 345 |
+
"eval_logits/rejected": 0.9058605432510376,
|
| 346 |
+
"eval_logps/chosen": -336.09857177734375,
|
| 347 |
+
"eval_logps/rejected": -416.4310607910156,
|
| 348 |
+
"eval_loss": 0.5255736708641052,
|
| 349 |
+
"eval_rewards/accuracies": 0.76953125,
|
| 350 |
+
"eval_rewards/chosen": -0.7352346181869507,
|
| 351 |
+
"eval_rewards/margins": 0.8022347688674927,
|
| 352 |
+
"eval_rewards/rejected": -1.5374693870544434,
|
| 353 |
+
"eval_runtime": 40.3301,
|
| 354 |
+
"eval_samples_per_second": 49.591,
|
| 355 |
+
"eval_steps_per_second": 0.793,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.4393305439330544,
|
| 360 |
+
"grad_norm": 23.975315777665053,
|
| 361 |
+
"learning_rate": 3.4440382358952115e-07,
|
| 362 |
+
"logits/chosen": 0.23435361683368683,
|
| 363 |
+
"logits/rejected": 1.2575860023498535,
|
| 364 |
+
"logps/chosen": -365.595703125,
|
| 365 |
+
"logps/rejected": -398.31182861328125,
|
| 366 |
+
"loss": 0.5374,
|
| 367 |
+
"rewards/accuracies": 0.731249988079071,
|
| 368 |
+
"rewards/chosen": -0.759240448474884,
|
| 369 |
+
"rewards/margins": 0.711215615272522,
|
| 370 |
+
"rewards/rejected": -1.4704558849334717,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.4602510460251046,
|
| 375 |
+
"grad_norm": 33.35111797134312,
|
| 376 |
+
"learning_rate": 3.272542485937368e-07,
|
| 377 |
+
"logits/chosen": 0.5184121131896973,
|
| 378 |
+
"logits/rejected": 1.4770405292510986,
|
| 379 |
+
"logps/chosen": -347.60296630859375,
|
| 380 |
+
"logps/rejected": -407.5792541503906,
|
| 381 |
+
"loss": 0.5233,
|
| 382 |
+
"rewards/accuracies": 0.78125,
|
| 383 |
+
"rewards/chosen": -0.8224745988845825,
|
| 384 |
+
"rewards/margins": 0.7966977953910828,
|
| 385 |
+
"rewards/rejected": -1.61917245388031,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.4811715481171548,
|
| 390 |
+
"grad_norm": 27.065916482787443,
|
| 391 |
+
"learning_rate": 3.096924887558854e-07,
|
| 392 |
+
"logits/chosen": 0.49935898184776306,
|
| 393 |
+
"logits/rejected": 1.3911898136138916,
|
| 394 |
+
"logps/chosen": -331.6248474121094,
|
| 395 |
+
"logps/rejected": -421.27374267578125,
|
| 396 |
+
"loss": 0.536,
|
| 397 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 398 |
+
"rewards/chosen": -0.8267892003059387,
|
| 399 |
+
"rewards/margins": 0.9082108736038208,
|
| 400 |
+
"rewards/rejected": -1.7350002527236938,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.502092050209205,
|
| 405 |
+
"grad_norm": 22.856393344097995,
|
| 406 |
+
"learning_rate": 2.9181224366319943e-07,
|
| 407 |
+
"logits/chosen": 0.2771614193916321,
|
| 408 |
+
"logits/rejected": 0.9905085563659668,
|
| 409 |
+
"logps/chosen": -338.65069580078125,
|
| 410 |
+
"logps/rejected": -394.545654296875,
|
| 411 |
+
"loss": 0.5024,
|
| 412 |
+
"rewards/accuracies": 0.78125,
|
| 413 |
+
"rewards/chosen": -0.7491301894187927,
|
| 414 |
+
"rewards/margins": 0.7059187293052673,
|
| 415 |
+
"rewards/rejected": -1.4550487995147705,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5230125523012552,
|
| 420 |
+
"grad_norm": 26.034393008774543,
|
| 421 |
+
"learning_rate": 2.7370891215954565e-07,
|
| 422 |
+
"logits/chosen": 0.8687924146652222,
|
| 423 |
+
"logits/rejected": 1.8204635381698608,
|
| 424 |
+
"logps/chosen": -394.45635986328125,
|
| 425 |
+
"logps/rejected": -441.63739013671875,
|
| 426 |
+
"loss": 0.4997,
|
| 427 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 428 |
+
"rewards/chosen": -0.9626559019088745,
|
| 429 |
+
"rewards/margins": 0.858391284942627,
|
| 430 |
+
"rewards/rejected": -1.821047067642212,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.5439330543933054,
|
| 435 |
+
"grad_norm": 32.298336541514864,
|
| 436 |
+
"learning_rate": 2.55479083351317e-07,
|
| 437 |
+
"logits/chosen": 0.8348854184150696,
|
| 438 |
+
"logits/rejected": 1.8038705587387085,
|
| 439 |
+
"logps/chosen": -395.5223693847656,
|
| 440 |
+
"logps/rejected": -442.0123596191406,
|
| 441 |
+
"loss": 0.5091,
|
| 442 |
+
"rewards/accuracies": 0.737500011920929,
|
| 443 |
+
"rewards/chosen": -1.0121302604675293,
|
| 444 |
+
"rewards/margins": 0.8687250018119812,
|
| 445 |
+
"rewards/rejected": -1.8808553218841553,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.5648535564853556,
|
| 450 |
+
"grad_norm": 23.028145780758834,
|
| 451 |
+
"learning_rate": 2.3722002126275822e-07,
|
| 452 |
+
"logits/chosen": 0.5836859941482544,
|
| 453 |
+
"logits/rejected": 1.3243392705917358,
|
| 454 |
+
"logps/chosen": -365.3858337402344,
|
| 455 |
+
"logps/rejected": -417.47735595703125,
|
| 456 |
+
"loss": 0.5181,
|
| 457 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 458 |
+
"rewards/chosen": -0.8838497996330261,
|
| 459 |
+
"rewards/margins": 0.7144038081169128,
|
| 460 |
+
"rewards/rejected": -1.5982534885406494,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.5857740585774058,
|
| 465 |
+
"grad_norm": 24.69890342328338,
|
| 466 |
+
"learning_rate": 2.19029145890313e-07,
|
| 467 |
+
"logits/chosen": 0.8636754155158997,
|
| 468 |
+
"logits/rejected": 1.8474575281143188,
|
| 469 |
+
"logps/chosen": -362.3246154785156,
|
| 470 |
+
"logps/rejected": -422.9788513183594,
|
| 471 |
+
"loss": 0.5317,
|
| 472 |
+
"rewards/accuracies": 0.71875,
|
| 473 |
+
"rewards/chosen": -1.0066673755645752,
|
| 474 |
+
"rewards/margins": 0.8012599945068359,
|
| 475 |
+
"rewards/rejected": -1.8079273700714111,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.606694560669456,
|
| 480 |
+
"grad_norm": 23.579877718409463,
|
| 481 |
+
"learning_rate": 2.0100351342479216e-07,
|
| 482 |
+
"logits/chosen": 0.9099249839782715,
|
| 483 |
+
"logits/rejected": 1.71127188205719,
|
| 484 |
+
"logps/chosen": -365.8359375,
|
| 485 |
+
"logps/rejected": -429.7777404785156,
|
| 486 |
+
"loss": 0.5117,
|
| 487 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 488 |
+
"rewards/chosen": -1.1478241682052612,
|
| 489 |
+
"rewards/margins": 0.7301602363586426,
|
| 490 |
+
"rewards/rejected": -1.8779844045639038,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6276150627615062,
|
| 495 |
+
"grad_norm": 28.794362121773794,
|
| 496 |
+
"learning_rate": 1.8323929841460178e-07,
|
| 497 |
+
"logits/chosen": 0.5150594115257263,
|
| 498 |
+
"logits/rejected": 1.7005367279052734,
|
| 499 |
+
"logps/chosen": -400.82708740234375,
|
| 500 |
+
"logps/rejected": -440.55181884765625,
|
| 501 |
+
"loss": 0.4892,
|
| 502 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 503 |
+
"rewards/chosen": -0.9979560971260071,
|
| 504 |
+
"rewards/margins": 0.8070657849311829,
|
| 505 |
+
"rewards/rejected": -1.80502188205719,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.6276150627615062,
|
| 510 |
+
"eval_logits/chosen": 0.5828189253807068,
|
| 511 |
+
"eval_logits/rejected": 1.3891677856445312,
|
| 512 |
+
"eval_logps/chosen": -351.67193603515625,
|
| 513 |
+
"eval_logps/rejected": -444.7821960449219,
|
| 514 |
+
"eval_loss": 0.5082466006278992,
|
| 515 |
+
"eval_rewards/accuracies": 0.76953125,
|
| 516 |
+
"eval_rewards/chosen": -0.8909686207771301,
|
| 517 |
+
"eval_rewards/margins": 0.9300119280815125,
|
| 518 |
+
"eval_rewards/rejected": -1.820980429649353,
|
| 519 |
+
"eval_runtime": 40.346,
|
| 520 |
+
"eval_samples_per_second": 49.571,
|
| 521 |
+
"eval_steps_per_second": 0.793,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.6485355648535565,
|
| 526 |
+
"grad_norm": 33.12731092406144,
|
| 527 |
+
"learning_rate": 1.6583128063291573e-07,
|
| 528 |
+
"logits/chosen": 0.7290142774581909,
|
| 529 |
+
"logits/rejected": 1.4886902570724487,
|
| 530 |
+
"logps/chosen": -397.9615478515625,
|
| 531 |
+
"logps/rejected": -434.9144592285156,
|
| 532 |
+
"loss": 0.4935,
|
| 533 |
+
"rewards/accuracies": 0.75,
|
| 534 |
+
"rewards/chosen": -1.0262950658798218,
|
| 535 |
+
"rewards/margins": 0.7377363443374634,
|
| 536 |
+
"rewards/rejected": -1.764031171798706,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.6694560669456067,
|
| 541 |
+
"grad_norm": 33.809250573756664,
|
| 542 |
+
"learning_rate": 1.488723393865766e-07,
|
| 543 |
+
"logits/chosen": 0.8037419319152832,
|
| 544 |
+
"logits/rejected": 1.8283920288085938,
|
| 545 |
+
"logps/chosen": -393.73876953125,
|
| 546 |
+
"logps/rejected": -423.55279541015625,
|
| 547 |
+
"loss": 0.4925,
|
| 548 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 549 |
+
"rewards/chosen": -1.0004233121871948,
|
| 550 |
+
"rewards/margins": 0.7867400050163269,
|
| 551 |
+
"rewards/rejected": -1.7871633768081665,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.6903765690376569,
|
| 556 |
+
"grad_norm": 29.294631844662618,
|
| 557 |
+
"learning_rate": 1.3245295796480788e-07,
|
| 558 |
+
"logits/chosen": 0.5138927698135376,
|
| 559 |
+
"logits/rejected": 1.4465028047561646,
|
| 560 |
+
"logps/chosen": -355.19622802734375,
|
| 561 |
+
"logps/rejected": -424.10052490234375,
|
| 562 |
+
"loss": 0.4984,
|
| 563 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 564 |
+
"rewards/chosen": -0.9297013282775879,
|
| 565 |
+
"rewards/margins": 0.7162156701087952,
|
| 566 |
+
"rewards/rejected": -1.6459171772003174,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.7112970711297071,
|
| 571 |
+
"grad_norm": 31.084849867717274,
|
| 572 |
+
"learning_rate": 1.1666074087171627e-07,
|
| 573 |
+
"logits/chosen": 0.8492459058761597,
|
| 574 |
+
"logits/rejected": 1.7086740732192993,
|
| 575 |
+
"logps/chosen": -376.3741455078125,
|
| 576 |
+
"logps/rejected": -468.2669982910156,
|
| 577 |
+
"loss": 0.5005,
|
| 578 |
+
"rewards/accuracies": 0.78125,
|
| 579 |
+
"rewards/chosen": -0.8633196949958801,
|
| 580 |
+
"rewards/margins": 1.0182921886444092,
|
| 581 |
+
"rewards/rejected": -1.8816118240356445,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.7322175732217573,
|
| 586 |
+
"grad_norm": 24.550583619455697,
|
| 587 |
+
"learning_rate": 1.0157994641835734e-07,
|
| 588 |
+
"logits/chosen": 0.9272977113723755,
|
| 589 |
+
"logits/rejected": 1.8347976207733154,
|
| 590 |
+
"logps/chosen": -351.08538818359375,
|
| 591 |
+
"logps/rejected": -420.5586853027344,
|
| 592 |
+
"loss": 0.4773,
|
| 593 |
+
"rewards/accuracies": 0.762499988079071,
|
| 594 |
+
"rewards/chosen": -0.9292421340942383,
|
| 595 |
+
"rewards/margins": 0.8692865371704102,
|
| 596 |
+
"rewards/rejected": -1.7985286712646484,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.7531380753138075,
|
| 601 |
+
"grad_norm": 23.786028174418703,
|
| 602 |
+
"learning_rate": 8.729103716819111e-08,
|
| 603 |
+
"logits/chosen": 0.772596001625061,
|
| 604 |
+
"logits/rejected": 1.983689308166504,
|
| 605 |
+
"logps/chosen": -400.6702575683594,
|
| 606 |
+
"logps/rejected": -450.78521728515625,
|
| 607 |
+
"loss": 0.511,
|
| 608 |
+
"rewards/accuracies": 0.762499988079071,
|
| 609 |
+
"rewards/chosen": -0.9801367521286011,
|
| 610 |
+
"rewards/margins": 0.9152604341506958,
|
| 611 |
+
"rewards/rejected": -1.8953969478607178,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.7740585774058577,
|
| 616 |
+
"grad_norm": 28.031131650473906,
|
| 617 |
+
"learning_rate": 7.387025063449081e-08,
|
| 618 |
+
"logits/chosen": 1.3180190324783325,
|
| 619 |
+
"logits/rejected": 2.0549569129943848,
|
| 620 |
+
"logps/chosen": -387.903076171875,
|
| 621 |
+
"logps/rejected": -421.43560791015625,
|
| 622 |
+
"loss": 0.5062,
|
| 623 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 624 |
+
"rewards/chosen": -1.1257370710372925,
|
| 625 |
+
"rewards/margins": 0.7061508893966675,
|
| 626 |
+
"rewards/rejected": -1.83188796043396,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.7949790794979079,
|
| 631 |
+
"grad_norm": 25.874688989134494,
|
| 632 |
+
"learning_rate": 6.138919252022435e-08,
|
| 633 |
+
"logits/chosen": 1.4991798400878906,
|
| 634 |
+
"logits/rejected": 2.2097418308258057,
|
| 635 |
+
"logps/chosen": -361.59136962890625,
|
| 636 |
+
"logps/rejected": -474.5025939941406,
|
| 637 |
+
"loss": 0.5045,
|
| 638 |
+
"rewards/accuracies": 0.78125,
|
| 639 |
+
"rewards/chosen": -1.2211620807647705,
|
| 640 |
+
"rewards/margins": 0.9370111227035522,
|
| 641 |
+
"rewards/rejected": -2.1581733226776123,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.8158995815899581,
|
| 646 |
+
"grad_norm": 25.013956798098185,
|
| 647 |
+
"learning_rate": 4.991445467064689e-08,
|
| 648 |
+
"logits/chosen": 0.9765421152114868,
|
| 649 |
+
"logits/rejected": 1.5557312965393066,
|
| 650 |
+
"logps/chosen": -405.4990234375,
|
| 651 |
+
"logps/rejected": -466.1669006347656,
|
| 652 |
+
"loss": 0.4919,
|
| 653 |
+
"rewards/accuracies": 0.731249988079071,
|
| 654 |
+
"rewards/chosen": -0.9986322522163391,
|
| 655 |
+
"rewards/margins": 0.8234176635742188,
|
| 656 |
+
"rewards/rejected": -1.822049856185913,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.8368200836820083,
|
| 661 |
+
"grad_norm": 25.325590503420294,
|
| 662 |
+
"learning_rate": 3.9507259776993954e-08,
|
| 663 |
+
"logits/chosen": 1.118195652961731,
|
| 664 |
+
"logits/rejected": 2.1151347160339355,
|
| 665 |
+
"logps/chosen": -369.673095703125,
|
| 666 |
+
"logps/rejected": -455.16949462890625,
|
| 667 |
+
"loss": 0.5037,
|
| 668 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 669 |
+
"rewards/chosen": -0.991197943687439,
|
| 670 |
+
"rewards/margins": 0.8906410336494446,
|
| 671 |
+
"rewards/rejected": -1.8818390369415283,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.8368200836820083,
|
| 676 |
+
"eval_logits/chosen": 0.7959027290344238,
|
| 677 |
+
"eval_logits/rejected": 1.6106007099151611,
|
| 678 |
+
"eval_logps/chosen": -346.2210998535156,
|
| 679 |
+
"eval_logps/rejected": -441.496826171875,
|
| 680 |
+
"eval_loss": 0.5030511617660522,
|
| 681 |
+
"eval_rewards/accuracies": 0.78515625,
|
| 682 |
+
"eval_rewards/chosen": -0.8364605903625488,
|
| 683 |
+
"eval_rewards/margins": 0.9516662955284119,
|
| 684 |
+
"eval_rewards/rejected": -1.788127064704895,
|
| 685 |
+
"eval_runtime": 40.3046,
|
| 686 |
+
"eval_samples_per_second": 49.622,
|
| 687 |
+
"eval_steps_per_second": 0.794,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.8577405857740585,
|
| 692 |
+
"grad_norm": 25.826488801828866,
|
| 693 |
+
"learning_rate": 3.022313472693447e-08,
|
| 694 |
+
"logits/chosen": 0.8946113586425781,
|
| 695 |
+
"logits/rejected": 2.032439708709717,
|
| 696 |
+
"logps/chosen": -391.67340087890625,
|
| 697 |
+
"logps/rejected": -448.6036682128906,
|
| 698 |
+
"loss": 0.5091,
|
| 699 |
+
"rewards/accuracies": 0.768750011920929,
|
| 700 |
+
"rewards/chosen": -0.9272828102111816,
|
| 701 |
+
"rewards/margins": 0.8706603050231934,
|
| 702 |
+
"rewards/rejected": -1.797943115234375,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.8786610878661087,
|
| 707 |
+
"grad_norm": 24.611045682765628,
|
| 708 |
+
"learning_rate": 2.2111614344599684e-08,
|
| 709 |
+
"logits/chosen": 0.7512882947921753,
|
| 710 |
+
"logits/rejected": 1.5819581747055054,
|
| 711 |
+
"logps/chosen": -391.5676574707031,
|
| 712 |
+
"logps/rejected": -447.89959716796875,
|
| 713 |
+
"loss": 0.4949,
|
| 714 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 715 |
+
"rewards/chosen": -0.9452153444290161,
|
| 716 |
+
"rewards/margins": 0.789088249206543,
|
| 717 |
+
"rewards/rejected": -1.7343034744262695,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.899581589958159,
|
| 722 |
+
"grad_norm": 24.57477424595926,
|
| 723 |
+
"learning_rate": 1.521597710086439e-08,
|
| 724 |
+
"logits/chosen": 1.212896704673767,
|
| 725 |
+
"logits/rejected": 1.9935243129730225,
|
| 726 |
+
"logps/chosen": -380.76373291015625,
|
| 727 |
+
"logps/rejected": -443.2093811035156,
|
| 728 |
+
"loss": 0.4764,
|
| 729 |
+
"rewards/accuracies": 0.78125,
|
| 730 |
+
"rewards/chosen": -1.0168489217758179,
|
| 731 |
+
"rewards/margins": 0.8378502130508423,
|
| 732 |
+
"rewards/rejected": -1.8546991348266602,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9205020920502092,
|
| 737 |
+
"grad_norm": 29.526622672975254,
|
| 738 |
+
"learning_rate": 9.57301420397924e-09,
|
| 739 |
+
"logits/chosen": 0.9320109486579895,
|
| 740 |
+
"logits/rejected": 1.8708562850952148,
|
| 741 |
+
"logps/chosen": -375.6236877441406,
|
| 742 |
+
"logps/rejected": -440.24822998046875,
|
| 743 |
+
"loss": 0.4967,
|
| 744 |
+
"rewards/accuracies": 0.768750011920929,
|
| 745 |
+
"rewards/chosen": -0.9130943417549133,
|
| 746 |
+
"rewards/margins": 0.7984671592712402,
|
| 747 |
+
"rewards/rejected": -1.7115614414215088,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.9414225941422594,
|
| 752 |
+
"grad_norm": 25.409754325281586,
|
| 753 |
+
"learning_rate": 5.212833302556258e-09,
|
| 754 |
+
"logits/chosen": 0.8921103477478027,
|
| 755 |
+
"logits/rejected": 1.6804935932159424,
|
| 756 |
+
"logps/chosen": -392.9017333984375,
|
| 757 |
+
"logps/rejected": -487.7478942871094,
|
| 758 |
+
"loss": 0.4971,
|
| 759 |
+
"rewards/accuracies": 0.706250011920929,
|
| 760 |
+
"rewards/chosen": -0.9917811155319214,
|
| 761 |
+
"rewards/margins": 0.8150280117988586,
|
| 762 |
+
"rewards/rejected": -1.8068091869354248,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.9623430962343096,
|
| 767 |
+
"grad_norm": 27.931385213072396,
|
| 768 |
+
"learning_rate": 2.158697848236607e-09,
|
| 769 |
+
"logits/chosen": 1.0774990320205688,
|
| 770 |
+
"logits/rejected": 1.940333366394043,
|
| 771 |
+
"logps/chosen": -366.88958740234375,
|
| 772 |
+
"logps/rejected": -417.73590087890625,
|
| 773 |
+
"loss": 0.5033,
|
| 774 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 775 |
+
"rewards/chosen": -0.9498540759086609,
|
| 776 |
+
"rewards/margins": 0.801009476184845,
|
| 777 |
+
"rewards/rejected": -1.7508634328842163,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.9832635983263598,
|
| 782 |
+
"grad_norm": 26.97146711470604,
|
| 783 |
+
"learning_rate": 4.269029751107489e-10,
|
| 784 |
+
"logits/chosen": 0.9106823801994324,
|
| 785 |
+
"logits/rejected": 1.9309165477752686,
|
| 786 |
+
"logps/chosen": -368.4771728515625,
|
| 787 |
+
"logps/rejected": -455.9451599121094,
|
| 788 |
+
"loss": 0.4933,
|
| 789 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 790 |
+
"rewards/chosen": -0.8981329202651978,
|
| 791 |
+
"rewards/margins": 0.8376766443252563,
|
| 792 |
+
"rewards/rejected": -1.735809564590454,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 1.0,
|
| 797 |
+
"step": 478,
|
| 798 |
+
"total_flos": 0.0,
|
| 799 |
+
"train_loss": 0.5395244165444474,
|
| 800 |
+
"train_runtime": 4445.7731,
|
| 801 |
+
"train_samples_per_second": 13.751,
|
| 802 |
+
"train_steps_per_second": 0.108
|
| 803 |
+
}
|
| 804 |
+
],
|
| 805 |
+
"logging_steps": 10,
|
| 806 |
+
"max_steps": 478,
|
| 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 |
+
}
|