Model save
Browse files- README.md +80 -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 +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-gpt_consistent-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-gpt_consistent-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: 0.0327
|
| 21 |
+
- Rewards/chosen: -0.1258
|
| 22 |
+
- Rewards/rejected: -0.4253
|
| 23 |
+
- Rewards/accuracies: 0.7543
|
| 24 |
+
- Rewards/margins: 0.2995
|
| 25 |
+
- Logps/rejected: -289.0509
|
| 26 |
+
- Logps/chosen: -297.6692
|
| 27 |
+
- Logits/rejected: -1.4762
|
| 28 |
+
- Logits/chosen: -1.6863
|
| 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 |
+
| 0.0487 | 0.1147 | 50 | 0.0455 | 0.0300 | -0.0845 | 0.7026 | 0.1145 | -254.9693 | -282.0912 | -2.4928 | -2.5717 |
|
| 66 |
+
| 0.0408 | 0.2294 | 100 | 0.0391 | -0.1071 | -0.3089 | 0.6897 | 0.2018 | -277.4089 | -295.8004 | -1.7416 | -1.8404 |
|
| 67 |
+
| 0.0379 | 0.3440 | 150 | 0.0365 | -0.1406 | -0.4136 | 0.7155 | 0.2730 | -287.8874 | -299.1519 | -1.6357 | -1.7904 |
|
| 68 |
+
| 0.0365 | 0.4587 | 200 | 0.0350 | -0.0650 | -0.3264 | 0.7543 | 0.2614 | -279.1631 | -291.5866 | -1.7552 | -1.9178 |
|
| 69 |
+
| 0.0346 | 0.5734 | 250 | 0.0337 | -0.1319 | -0.4539 | 0.7543 | 0.3220 | -291.9156 | -298.2828 | -1.4871 | -1.7192 |
|
| 70 |
+
| 0.036 | 0.6881 | 300 | 0.0331 | -0.1336 | -0.4291 | 0.75 | 0.2955 | -289.4286 | -298.4504 | -1.4842 | -1.6835 |
|
| 71 |
+
| 0.0359 | 0.8028 | 350 | 0.0327 | -0.1378 | -0.4472 | 0.7586 | 0.3094 | -291.2399 | -298.8666 | -1.4658 | -1.6786 |
|
| 72 |
+
| 0.0351 | 0.9174 | 400 | 0.0327 | -0.1258 | -0.4253 | 0.7543 | 0.2995 | -289.0509 | -297.6692 | -1.4762 | -1.6863 |
|
| 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": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.0384325077500912,
|
| 5 |
+
"train_runtime": 11659.9634,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.782,
|
| 8 |
+
"train_steps_per_second": 0.037
|
| 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:de11b87922b4b44c8cca13b4fd1512d1d8080e08297e93d938d39bd61aa84066
|
| 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 sha256:57aaa71d01065cadbe8c8c60ab7a99a32cc956d53ba264fbbe7c076dfe1f5471
|
| 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:dcfba798eaadf6d113a0642a66150398a3a3378d41bce7656fb7e6e2eff751f1
|
| 3 |
size 4540516344
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.0384325077500912,
|
| 5 |
+
"train_runtime": 11659.9634,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.782,
|
| 8 |
+
"train_steps_per_second": 0.037
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,815 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 50,
|
| 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": 3.1641646757229265,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.6194896697998047,
|
| 16 |
+
"logits/rejected": -2.552656650543213,
|
| 17 |
+
"logps/chosen": -265.41778564453125,
|
| 18 |
+
"logps/rejected": -236.15280151367188,
|
| 19 |
+
"loss": 0.0585,
|
| 20 |
+
"rewards/accuracies": 0.39375001192092896,
|
| 21 |
+
"rewards/chosen": 0.00028036616276949644,
|
| 22 |
+
"rewards/margins": 0.0005085940356366336,
|
| 23 |
+
"rewards/rejected": -0.00022822784376330674,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045871559633027525,
|
| 28 |
+
"grad_norm": 2.9506602825308903,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.6579580307006836,
|
| 31 |
+
"logits/rejected": -2.575996160507202,
|
| 32 |
+
"logps/chosen": -298.8293151855469,
|
| 33 |
+
"logps/rejected": -274.3349304199219,
|
| 34 |
+
"loss": 0.059,
|
| 35 |
+
"rewards/accuracies": 0.643750011920929,
|
| 36 |
+
"rewards/chosen": -0.0004406004154589027,
|
| 37 |
+
"rewards/margins": 0.0018091611564159393,
|
| 38 |
+
"rewards/rejected": -0.0022497617173939943,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06880733944954129,
|
| 43 |
+
"grad_norm": 2.7648586784794067,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.6760480403900146,
|
| 46 |
+
"logits/rejected": -2.6023683547973633,
|
| 47 |
+
"logps/chosen": -290.41986083984375,
|
| 48 |
+
"logps/rejected": -234.40579223632812,
|
| 49 |
+
"loss": 0.0571,
|
| 50 |
+
"rewards/accuracies": 0.699999988079071,
|
| 51 |
+
"rewards/chosen": 0.009129038080573082,
|
| 52 |
+
"rewards/margins": 0.01336810551583767,
|
| 53 |
+
"rewards/rejected": -0.004239068366587162,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09174311926605505,
|
| 58 |
+
"grad_norm": 2.7634574887860825,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.6598825454711914,
|
| 61 |
+
"logits/rejected": -2.610196352005005,
|
| 62 |
+
"logps/chosen": -280.923828125,
|
| 63 |
+
"logps/rejected": -267.7273864746094,
|
| 64 |
+
"loss": 0.0529,
|
| 65 |
+
"rewards/accuracies": 0.6875,
|
| 66 |
+
"rewards/chosen": 0.04245365411043167,
|
| 67 |
+
"rewards/margins": 0.04179743677377701,
|
| 68 |
+
"rewards/rejected": 0.0006562232738360763,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11467889908256881,
|
| 73 |
+
"grad_norm": 2.9044861722207567,
|
| 74 |
+
"learning_rate": 4.997110275491701e-07,
|
| 75 |
+
"logits/chosen": -2.620276927947998,
|
| 76 |
+
"logits/rejected": -2.6136012077331543,
|
| 77 |
+
"logps/chosen": -292.2350158691406,
|
| 78 |
+
"logps/rejected": -302.38018798828125,
|
| 79 |
+
"loss": 0.0487,
|
| 80 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 81 |
+
"rewards/chosen": 0.04016200453042984,
|
| 82 |
+
"rewards/margins": 0.07212186604738235,
|
| 83 |
+
"rewards/rejected": -0.031959857791662216,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11467889908256881,
|
| 88 |
+
"eval_logits/chosen": -2.5716989040374756,
|
| 89 |
+
"eval_logits/rejected": -2.4927897453308105,
|
| 90 |
+
"eval_logps/chosen": -282.0911865234375,
|
| 91 |
+
"eval_logps/rejected": -254.96926879882812,
|
| 92 |
+
"eval_loss": 0.04548575356602669,
|
| 93 |
+
"eval_rewards/accuracies": 0.7025862336158752,
|
| 94 |
+
"eval_rewards/chosen": 0.02998923696577549,
|
| 95 |
+
"eval_rewards/margins": 0.11445755511522293,
|
| 96 |
+
"eval_rewards/rejected": -0.08446833491325378,
|
| 97 |
+
"eval_runtime": 95.7446,
|
| 98 |
+
"eval_samples_per_second": 18.988,
|
| 99 |
+
"eval_steps_per_second": 0.303,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13761467889908258,
|
| 104 |
+
"grad_norm": 3.2642316994426857,
|
| 105 |
+
"learning_rate": 4.979475034558115e-07,
|
| 106 |
+
"logits/chosen": -2.5661163330078125,
|
| 107 |
+
"logits/rejected": -2.50827956199646,
|
| 108 |
+
"logps/chosen": -289.7759704589844,
|
| 109 |
+
"logps/rejected": -267.379638671875,
|
| 110 |
+
"loss": 0.0449,
|
| 111 |
+
"rewards/accuracies": 0.706250011920929,
|
| 112 |
+
"rewards/chosen": -0.0029130387119948864,
|
| 113 |
+
"rewards/margins": 0.09531786292791367,
|
| 114 |
+
"rewards/rejected": -0.09823091328144073,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16055045871559634,
|
| 119 |
+
"grad_norm": 3.167167289379583,
|
| 120 |
+
"learning_rate": 4.945923025551788e-07,
|
| 121 |
+
"logits/chosen": -2.460839033126831,
|
| 122 |
+
"logits/rejected": -2.4139201641082764,
|
| 123 |
+
"logps/chosen": -324.13690185546875,
|
| 124 |
+
"logps/rejected": -276.5367126464844,
|
| 125 |
+
"loss": 0.0446,
|
| 126 |
+
"rewards/accuracies": 0.706250011920929,
|
| 127 |
+
"rewards/chosen": 0.034340955317020416,
|
| 128 |
+
"rewards/margins": 0.1856319159269333,
|
| 129 |
+
"rewards/rejected": -0.15129096806049347,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.1834862385321101,
|
| 134 |
+
"grad_norm": 3.00053019606861,
|
| 135 |
+
"learning_rate": 4.896669632591651e-07,
|
| 136 |
+
"logits/chosen": -2.356163740158081,
|
| 137 |
+
"logits/rejected": -2.2467029094696045,
|
| 138 |
+
"logps/chosen": -287.4019470214844,
|
| 139 |
+
"logps/rejected": -272.3543701171875,
|
| 140 |
+
"loss": 0.0416,
|
| 141 |
+
"rewards/accuracies": 0.71875,
|
| 142 |
+
"rewards/chosen": -0.015520763583481312,
|
| 143 |
+
"rewards/margins": 0.18146219849586487,
|
| 144 |
+
"rewards/rejected": -0.1969829499721527,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.20642201834862386,
|
| 149 |
+
"grad_norm": 3.936347429433164,
|
| 150 |
+
"learning_rate": 4.832031033425662e-07,
|
| 151 |
+
"logits/chosen": -2.0837090015411377,
|
| 152 |
+
"logits/rejected": -2.0328478813171387,
|
| 153 |
+
"logps/chosen": -287.70654296875,
|
| 154 |
+
"logps/rejected": -268.2559509277344,
|
| 155 |
+
"loss": 0.0405,
|
| 156 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 157 |
+
"rewards/chosen": -0.001131159020587802,
|
| 158 |
+
"rewards/margins": 0.2134155035018921,
|
| 159 |
+
"rewards/rejected": -0.21454665064811707,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22935779816513763,
|
| 164 |
+
"grad_norm": 3.652555898125068,
|
| 165 |
+
"learning_rate": 4.752422169756047e-07,
|
| 166 |
+
"logits/chosen": -1.5697656869888306,
|
| 167 |
+
"logits/rejected": -1.561859130859375,
|
| 168 |
+
"logps/chosen": -289.6307678222656,
|
| 169 |
+
"logps/rejected": -307.66363525390625,
|
| 170 |
+
"loss": 0.0408,
|
| 171 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 172 |
+
"rewards/chosen": -0.17206048965454102,
|
| 173 |
+
"rewards/margins": 0.18529178202152252,
|
| 174 |
+
"rewards/rejected": -0.35735225677490234,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22935779816513763,
|
| 179 |
+
"eval_logits/chosen": -1.8404057025909424,
|
| 180 |
+
"eval_logits/rejected": -1.7416090965270996,
|
| 181 |
+
"eval_logps/chosen": -295.8004455566406,
|
| 182 |
+
"eval_logps/rejected": -277.408935546875,
|
| 183 |
+
"eval_loss": 0.03911930322647095,
|
| 184 |
+
"eval_rewards/accuracies": 0.6896551847457886,
|
| 185 |
+
"eval_rewards/chosen": -0.10710364580154419,
|
| 186 |
+
"eval_rewards/margins": 0.201760932803154,
|
| 187 |
+
"eval_rewards/rejected": -0.308864563703537,
|
| 188 |
+
"eval_runtime": 95.267,
|
| 189 |
+
"eval_samples_per_second": 19.083,
|
| 190 |
+
"eval_steps_per_second": 0.304,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25229357798165136,
|
| 195 |
+
"grad_norm": 3.509676092042844,
|
| 196 |
+
"learning_rate": 4.658354083558188e-07,
|
| 197 |
+
"logits/chosen": -1.9262816905975342,
|
| 198 |
+
"logits/rejected": -1.80355966091156,
|
| 199 |
+
"logps/chosen": -279.3715515136719,
|
| 200 |
+
"logps/rejected": -263.81549072265625,
|
| 201 |
+
"loss": 0.0402,
|
| 202 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 203 |
+
"rewards/chosen": -0.14183419942855835,
|
| 204 |
+
"rewards/margins": 0.15898862481117249,
|
| 205 |
+
"rewards/rejected": -0.30082279443740845,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.27522935779816515,
|
| 210 |
+
"grad_norm": 3.984377752564919,
|
| 211 |
+
"learning_rate": 4.550430636492389e-07,
|
| 212 |
+
"logits/chosen": -1.8143861293792725,
|
| 213 |
+
"logits/rejected": -1.7607357501983643,
|
| 214 |
+
"logps/chosen": -295.43243408203125,
|
| 215 |
+
"logps/rejected": -289.3270568847656,
|
| 216 |
+
"loss": 0.0395,
|
| 217 |
+
"rewards/accuracies": 0.675000011920929,
|
| 218 |
+
"rewards/chosen": -0.19613118469715118,
|
| 219 |
+
"rewards/margins": 0.18250016868114471,
|
| 220 |
+
"rewards/rejected": -0.3786313831806183,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.2981651376146789,
|
| 225 |
+
"grad_norm": 3.895538046894766,
|
| 226 |
+
"learning_rate": 4.429344633468004e-07,
|
| 227 |
+
"logits/chosen": -1.742222785949707,
|
| 228 |
+
"logits/rejected": -1.6753323078155518,
|
| 229 |
+
"logps/chosen": -268.340087890625,
|
| 230 |
+
"logps/rejected": -275.31500244140625,
|
| 231 |
+
"loss": 0.0396,
|
| 232 |
+
"rewards/accuracies": 0.6875,
|
| 233 |
+
"rewards/chosen": -0.12159997224807739,
|
| 234 |
+
"rewards/margins": 0.20542438328266144,
|
| 235 |
+
"rewards/rejected": -0.32702434062957764,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.3211009174311927,
|
| 240 |
+
"grad_norm": 4.968671690125824,
|
| 241 |
+
"learning_rate": 4.2958733752443187e-07,
|
| 242 |
+
"logits/chosen": -1.6684496402740479,
|
| 243 |
+
"logits/rejected": -1.5686622858047485,
|
| 244 |
+
"logps/chosen": -286.41973876953125,
|
| 245 |
+
"logps/rejected": -254.42691040039062,
|
| 246 |
+
"loss": 0.0381,
|
| 247 |
+
"rewards/accuracies": 0.71875,
|
| 248 |
+
"rewards/chosen": -0.10234608501195908,
|
| 249 |
+
"rewards/margins": 0.19895747303962708,
|
| 250 |
+
"rewards/rejected": -0.30130359530448914,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.3440366972477064,
|
| 255 |
+
"grad_norm": 3.5584001205811377,
|
| 256 |
+
"learning_rate": 4.150873668617898e-07,
|
| 257 |
+
"logits/chosen": -1.703791856765747,
|
| 258 |
+
"logits/rejected": -1.5906720161437988,
|
| 259 |
+
"logps/chosen": -280.5138854980469,
|
| 260 |
+
"logps/rejected": -269.3014831542969,
|
| 261 |
+
"loss": 0.0379,
|
| 262 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 263 |
+
"rewards/chosen": -0.13970300555229187,
|
| 264 |
+
"rewards/margins": 0.22143635153770447,
|
| 265 |
+
"rewards/rejected": -0.36113935708999634,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.3440366972477064,
|
| 270 |
+
"eval_logits/chosen": -1.7903783321380615,
|
| 271 |
+
"eval_logits/rejected": -1.635672688484192,
|
| 272 |
+
"eval_logps/chosen": -299.1519470214844,
|
| 273 |
+
"eval_logps/rejected": -287.88739013671875,
|
| 274 |
+
"eval_loss": 0.03649381175637245,
|
| 275 |
+
"eval_rewards/accuracies": 0.7155172228813171,
|
| 276 |
+
"eval_rewards/chosen": -0.14061833918094635,
|
| 277 |
+
"eval_rewards/margins": 0.27303099632263184,
|
| 278 |
+
"eval_rewards/rejected": -0.413649320602417,
|
| 279 |
+
"eval_runtime": 94.6259,
|
| 280 |
+
"eval_samples_per_second": 19.212,
|
| 281 |
+
"eval_steps_per_second": 0.306,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3669724770642202,
|
| 286 |
+
"grad_norm": 5.2149351876346905,
|
| 287 |
+
"learning_rate": 3.9952763262280397e-07,
|
| 288 |
+
"logits/chosen": -1.6572643518447876,
|
| 289 |
+
"logits/rejected": -1.5417451858520508,
|
| 290 |
+
"logps/chosen": -309.5204162597656,
|
| 291 |
+
"logps/rejected": -324.3457946777344,
|
| 292 |
+
"loss": 0.0359,
|
| 293 |
+
"rewards/accuracies": 0.731249988079071,
|
| 294 |
+
"rewards/chosen": -0.12969312071800232,
|
| 295 |
+
"rewards/margins": 0.254078209400177,
|
| 296 |
+
"rewards/rejected": -0.3837713599205017,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38990825688073394,
|
| 301 |
+
"grad_norm": 2.9551901080690395,
|
| 302 |
+
"learning_rate": 3.8300801912883414e-07,
|
| 303 |
+
"logits/chosen": -1.6375774145126343,
|
| 304 |
+
"logits/rejected": -1.5148413181304932,
|
| 305 |
+
"logps/chosen": -277.994873046875,
|
| 306 |
+
"logps/rejected": -301.9854431152344,
|
| 307 |
+
"loss": 0.0339,
|
| 308 |
+
"rewards/accuracies": 0.731249988079071,
|
| 309 |
+
"rewards/chosen": -0.06393978744745255,
|
| 310 |
+
"rewards/margins": 0.2552258372306824,
|
| 311 |
+
"rewards/rejected": -0.3191656172275543,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.41284403669724773,
|
| 316 |
+
"grad_norm": 3.6761291533851423,
|
| 317 |
+
"learning_rate": 3.6563457256020884e-07,
|
| 318 |
+
"logits/chosen": -1.527038335800171,
|
| 319 |
+
"logits/rejected": -1.2771427631378174,
|
| 320 |
+
"logps/chosen": -323.6756896972656,
|
| 321 |
+
"logps/rejected": -284.1974792480469,
|
| 322 |
+
"loss": 0.0365,
|
| 323 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 324 |
+
"rewards/chosen": -0.20317766070365906,
|
| 325 |
+
"rewards/margins": 0.2538478374481201,
|
| 326 |
+
"rewards/rejected": -0.4570254683494568,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.43577981651376146,
|
| 331 |
+
"grad_norm": 5.675419312165092,
|
| 332 |
+
"learning_rate": 3.475188202022617e-07,
|
| 333 |
+
"logits/chosen": -1.6745436191558838,
|
| 334 |
+
"logits/rejected": -1.604264259338379,
|
| 335 |
+
"logps/chosen": -268.5342102050781,
|
| 336 |
+
"logps/rejected": -302.51898193359375,
|
| 337 |
+
"loss": 0.0368,
|
| 338 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 339 |
+
"rewards/chosen": -0.10419674217700958,
|
| 340 |
+
"rewards/margins": 0.27754858136177063,
|
| 341 |
+
"rewards/rejected": -0.38174527883529663,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45871559633027525,
|
| 346 |
+
"grad_norm": 3.4589111733943283,
|
| 347 |
+
"learning_rate": 3.287770545059052e-07,
|
| 348 |
+
"logits/chosen": -1.8974136114120483,
|
| 349 |
+
"logits/rejected": -1.7577918767929077,
|
| 350 |
+
"logps/chosen": -291.4000549316406,
|
| 351 |
+
"logps/rejected": -281.4195861816406,
|
| 352 |
+
"loss": 0.0365,
|
| 353 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 354 |
+
"rewards/chosen": -0.1311841905117035,
|
| 355 |
+
"rewards/margins": 0.24445155262947083,
|
| 356 |
+
"rewards/rejected": -0.3756357431411743,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45871559633027525,
|
| 361 |
+
"eval_logits/chosen": -1.9178062677383423,
|
| 362 |
+
"eval_logits/rejected": -1.7552225589752197,
|
| 363 |
+
"eval_logps/chosen": -291.5865783691406,
|
| 364 |
+
"eval_logps/rejected": -279.1630859375,
|
| 365 |
+
"eval_loss": 0.03504867106676102,
|
| 366 |
+
"eval_rewards/accuracies": 0.7543103694915771,
|
| 367 |
+
"eval_rewards/chosen": -0.06496486067771912,
|
| 368 |
+
"eval_rewards/margins": 0.2614416182041168,
|
| 369 |
+
"eval_rewards/rejected": -0.32640647888183594,
|
| 370 |
+
"eval_runtime": 95.5023,
|
| 371 |
+
"eval_samples_per_second": 19.036,
|
| 372 |
+
"eval_steps_per_second": 0.304,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.481651376146789,
|
| 377 |
+
"grad_norm": 2.8163994777137717,
|
| 378 |
+
"learning_rate": 3.0952958655864954e-07,
|
| 379 |
+
"logits/chosen": -1.794647455215454,
|
| 380 |
+
"logits/rejected": -1.7260096073150635,
|
| 381 |
+
"logps/chosen": -280.9212951660156,
|
| 382 |
+
"logps/rejected": -287.47674560546875,
|
| 383 |
+
"loss": 0.0349,
|
| 384 |
+
"rewards/accuracies": 0.762499988079071,
|
| 385 |
+
"rewards/chosen": -0.07616890966892242,
|
| 386 |
+
"rewards/margins": 0.2551766633987427,
|
| 387 |
+
"rewards/rejected": -0.3313455581665039,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5045871559633027,
|
| 392 |
+
"grad_norm": 3.602353089915511,
|
| 393 |
+
"learning_rate": 2.898999737583448e-07,
|
| 394 |
+
"logits/chosen": -1.672721266746521,
|
| 395 |
+
"logits/rejected": -1.4848198890686035,
|
| 396 |
+
"logps/chosen": -344.14825439453125,
|
| 397 |
+
"logps/rejected": -337.38470458984375,
|
| 398 |
+
"loss": 0.0336,
|
| 399 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 400 |
+
"rewards/chosen": -0.2146792709827423,
|
| 401 |
+
"rewards/margins": 0.3094654977321625,
|
| 402 |
+
"rewards/rejected": -0.5241447687149048,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5275229357798165,
|
| 407 |
+
"grad_norm": 3.2287234258499633,
|
| 408 |
+
"learning_rate": 2.7001422664752333e-07,
|
| 409 |
+
"logits/chosen": -1.4607375860214233,
|
| 410 |
+
"logits/rejected": -1.3907452821731567,
|
| 411 |
+
"logps/chosen": -288.92095947265625,
|
| 412 |
+
"logps/rejected": -309.4497985839844,
|
| 413 |
+
"loss": 0.0354,
|
| 414 |
+
"rewards/accuracies": 0.731249988079071,
|
| 415 |
+
"rewards/chosen": -0.20977428555488586,
|
| 416 |
+
"rewards/margins": 0.22852472960948944,
|
| 417 |
+
"rewards/rejected": -0.4382990002632141,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5504587155963303,
|
| 422 |
+
"grad_norm": 3.200697411595364,
|
| 423 |
+
"learning_rate": 2.5e-07,
|
| 424 |
+
"logits/chosen": -1.6722291707992554,
|
| 425 |
+
"logits/rejected": -1.496265172958374,
|
| 426 |
+
"logps/chosen": -304.0669250488281,
|
| 427 |
+
"logps/rejected": -301.3495788574219,
|
| 428 |
+
"loss": 0.0352,
|
| 429 |
+
"rewards/accuracies": 0.75,
|
| 430 |
+
"rewards/chosen": -0.16535088419914246,
|
| 431 |
+
"rewards/margins": 0.23464509844779968,
|
| 432 |
+
"rewards/rejected": -0.39999598264694214,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.573394495412844,
|
| 437 |
+
"grad_norm": 3.068087094061713,
|
| 438 |
+
"learning_rate": 2.2998577335247667e-07,
|
| 439 |
+
"logits/chosen": -1.8118059635162354,
|
| 440 |
+
"logits/rejected": -1.58005952835083,
|
| 441 |
+
"logps/chosen": -316.8459777832031,
|
| 442 |
+
"logps/rejected": -301.0585632324219,
|
| 443 |
+
"loss": 0.0346,
|
| 444 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 445 |
+
"rewards/chosen": -0.1305214762687683,
|
| 446 |
+
"rewards/margins": 0.27562573552131653,
|
| 447 |
+
"rewards/rejected": -0.40614718198776245,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.573394495412844,
|
| 452 |
+
"eval_logits/chosen": -1.719161868095398,
|
| 453 |
+
"eval_logits/rejected": -1.4870593547821045,
|
| 454 |
+
"eval_logps/chosen": -298.28277587890625,
|
| 455 |
+
"eval_logps/rejected": -291.9156494140625,
|
| 456 |
+
"eval_loss": 0.033715467900037766,
|
| 457 |
+
"eval_rewards/accuracies": 0.7543103694915771,
|
| 458 |
+
"eval_rewards/chosen": -0.1319267898797989,
|
| 459 |
+
"eval_rewards/margins": 0.3220053017139435,
|
| 460 |
+
"eval_rewards/rejected": -0.4539320766925812,
|
| 461 |
+
"eval_runtime": 94.8489,
|
| 462 |
+
"eval_samples_per_second": 19.167,
|
| 463 |
+
"eval_steps_per_second": 0.306,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5963302752293578,
|
| 468 |
+
"grad_norm": 2.973513185196363,
|
| 469 |
+
"learning_rate": 2.1010002624165524e-07,
|
| 470 |
+
"logits/chosen": -1.625998854637146,
|
| 471 |
+
"logits/rejected": -1.528102159500122,
|
| 472 |
+
"logps/chosen": -289.2758483886719,
|
| 473 |
+
"logps/rejected": -324.4759216308594,
|
| 474 |
+
"loss": 0.0343,
|
| 475 |
+
"rewards/accuracies": 0.800000011920929,
|
| 476 |
+
"rewards/chosen": -0.18561328947544098,
|
| 477 |
+
"rewards/margins": 0.3152090609073639,
|
| 478 |
+
"rewards/rejected": -0.5008223652839661,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6192660550458715,
|
| 483 |
+
"grad_norm": 3.3992650920748635,
|
| 484 |
+
"learning_rate": 1.9047041344135043e-07,
|
| 485 |
+
"logits/chosen": -1.468390703201294,
|
| 486 |
+
"logits/rejected": -1.4237253665924072,
|
| 487 |
+
"logps/chosen": -287.03558349609375,
|
| 488 |
+
"logps/rejected": -301.84613037109375,
|
| 489 |
+
"loss": 0.0339,
|
| 490 |
+
"rewards/accuracies": 0.8125,
|
| 491 |
+
"rewards/chosen": -0.209732323884964,
|
| 492 |
+
"rewards/margins": 0.2902582287788391,
|
| 493 |
+
"rewards/rejected": -0.4999905526638031,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.6422018348623854,
|
| 498 |
+
"grad_norm": 3.2930402827157863,
|
| 499 |
+
"learning_rate": 1.7122294549409482e-07,
|
| 500 |
+
"logits/chosen": -1.6038357019424438,
|
| 501 |
+
"logits/rejected": -1.4040790796279907,
|
| 502 |
+
"logps/chosen": -287.1599426269531,
|
| 503 |
+
"logps/rejected": -311.6736755371094,
|
| 504 |
+
"loss": 0.0349,
|
| 505 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 506 |
+
"rewards/chosen": -0.14161059260368347,
|
| 507 |
+
"rewards/margins": 0.3371681869029999,
|
| 508 |
+
"rewards/rejected": -0.4787788391113281,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6651376146788991,
|
| 513 |
+
"grad_norm": 3.539437987763887,
|
| 514 |
+
"learning_rate": 1.524811797977383e-07,
|
| 515 |
+
"logits/chosen": -1.6025069952011108,
|
| 516 |
+
"logits/rejected": -1.4434142112731934,
|
| 517 |
+
"logps/chosen": -300.49578857421875,
|
| 518 |
+
"logps/rejected": -295.3720397949219,
|
| 519 |
+
"loss": 0.0334,
|
| 520 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 521 |
+
"rewards/chosen": -0.12764978408813477,
|
| 522 |
+
"rewards/margins": 0.2869029641151428,
|
| 523 |
+
"rewards/rejected": -0.4145527780056,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6880733944954128,
|
| 528 |
+
"grad_norm": 3.11392798700287,
|
| 529 |
+
"learning_rate": 1.3436542743979125e-07,
|
| 530 |
+
"logits/chosen": -1.6770799160003662,
|
| 531 |
+
"logits/rejected": -1.6160328388214111,
|
| 532 |
+
"logps/chosen": -319.42236328125,
|
| 533 |
+
"logps/rejected": -295.0827331542969,
|
| 534 |
+
"loss": 0.036,
|
| 535 |
+
"rewards/accuracies": 0.762499988079071,
|
| 536 |
+
"rewards/chosen": -0.1347302943468094,
|
| 537 |
+
"rewards/margins": 0.22749857604503632,
|
| 538 |
+
"rewards/rejected": -0.3622289299964905,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6880733944954128,
|
| 543 |
+
"eval_logits/chosen": -1.6834876537322998,
|
| 544 |
+
"eval_logits/rejected": -1.4841814041137695,
|
| 545 |
+
"eval_logps/chosen": -298.45037841796875,
|
| 546 |
+
"eval_logps/rejected": -289.4286193847656,
|
| 547 |
+
"eval_loss": 0.03310078755021095,
|
| 548 |
+
"eval_rewards/accuracies": 0.75,
|
| 549 |
+
"eval_rewards/chosen": -0.13360273838043213,
|
| 550 |
+
"eval_rewards/margins": 0.29545897245407104,
|
| 551 |
+
"eval_rewards/rejected": -0.4290617108345032,
|
| 552 |
+
"eval_runtime": 94.9394,
|
| 553 |
+
"eval_samples_per_second": 19.149,
|
| 554 |
+
"eval_steps_per_second": 0.305,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7110091743119266,
|
| 559 |
+
"grad_norm": 5.19818595642257,
|
| 560 |
+
"learning_rate": 1.1699198087116588e-07,
|
| 561 |
+
"logits/chosen": -1.741039514541626,
|
| 562 |
+
"logits/rejected": -1.6022865772247314,
|
| 563 |
+
"logps/chosen": -294.25433349609375,
|
| 564 |
+
"logps/rejected": -306.18817138671875,
|
| 565 |
+
"loss": 0.0357,
|
| 566 |
+
"rewards/accuracies": 0.71875,
|
| 567 |
+
"rewards/chosen": -0.16289182007312775,
|
| 568 |
+
"rewards/margins": 0.23449257016181946,
|
| 569 |
+
"rewards/rejected": -0.397384375333786,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7339449541284404,
|
| 574 |
+
"grad_norm": 3.4771448226315957,
|
| 575 |
+
"learning_rate": 1.00472367377196e-07,
|
| 576 |
+
"logits/chosen": -1.6833099126815796,
|
| 577 |
+
"logits/rejected": -1.5138328075408936,
|
| 578 |
+
"logps/chosen": -290.2877197265625,
|
| 579 |
+
"logps/rejected": -281.9583435058594,
|
| 580 |
+
"loss": 0.0353,
|
| 581 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 582 |
+
"rewards/chosen": -0.16055694222450256,
|
| 583 |
+
"rewards/margins": 0.3141244947910309,
|
| 584 |
+
"rewards/rejected": -0.47468137741088867,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7568807339449541,
|
| 589 |
+
"grad_norm": 3.5214688030455066,
|
| 590 |
+
"learning_rate": 8.49126331382102e-08,
|
| 591 |
+
"logits/chosen": -1.6057231426239014,
|
| 592 |
+
"logits/rejected": -1.4786289930343628,
|
| 593 |
+
"logps/chosen": -291.7789306640625,
|
| 594 |
+
"logps/rejected": -288.677001953125,
|
| 595 |
+
"loss": 0.0348,
|
| 596 |
+
"rewards/accuracies": 0.706250011920929,
|
| 597 |
+
"rewards/chosen": -0.17256753146648407,
|
| 598 |
+
"rewards/margins": 0.24320188164710999,
|
| 599 |
+
"rewards/rejected": -0.41576939821243286,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7798165137614679,
|
| 604 |
+
"grad_norm": 3.5006792226854495,
|
| 605 |
+
"learning_rate": 7.041266247556812e-08,
|
| 606 |
+
"logits/chosen": -1.7440093755722046,
|
| 607 |
+
"logits/rejected": -1.6453059911727905,
|
| 608 |
+
"logps/chosen": -304.0636901855469,
|
| 609 |
+
"logps/rejected": -299.057861328125,
|
| 610 |
+
"loss": 0.0337,
|
| 611 |
+
"rewards/accuracies": 0.78125,
|
| 612 |
+
"rewards/chosen": -0.16071084141731262,
|
| 613 |
+
"rewards/margins": 0.23002786934375763,
|
| 614 |
+
"rewards/rejected": -0.39073869585990906,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8027522935779816,
|
| 619 |
+
"grad_norm": 3.5127255839066094,
|
| 620 |
+
"learning_rate": 5.706553665319955e-08,
|
| 621 |
+
"logits/chosen": -1.7355537414550781,
|
| 622 |
+
"logits/rejected": -1.5236848592758179,
|
| 623 |
+
"logps/chosen": -292.27850341796875,
|
| 624 |
+
"logps/rejected": -280.9353332519531,
|
| 625 |
+
"loss": 0.0359,
|
| 626 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 627 |
+
"rewards/chosen": -0.17787429690361023,
|
| 628 |
+
"rewards/margins": 0.25486475229263306,
|
| 629 |
+
"rewards/rejected": -0.4327389597892761,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8027522935779816,
|
| 634 |
+
"eval_logits/chosen": -1.6786376237869263,
|
| 635 |
+
"eval_logits/rejected": -1.465800166130066,
|
| 636 |
+
"eval_logps/chosen": -298.86663818359375,
|
| 637 |
+
"eval_logps/rejected": -291.2398681640625,
|
| 638 |
+
"eval_loss": 0.0327293798327446,
|
| 639 |
+
"eval_rewards/accuracies": 0.7586206793785095,
|
| 640 |
+
"eval_rewards/chosen": -0.1377655267715454,
|
| 641 |
+
"eval_rewards/margins": 0.30940860509872437,
|
| 642 |
+
"eval_rewards/rejected": -0.44717416167259216,
|
| 643 |
+
"eval_runtime": 93.7918,
|
| 644 |
+
"eval_samples_per_second": 19.383,
|
| 645 |
+
"eval_steps_per_second": 0.309,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8256880733944955,
|
| 650 |
+
"grad_norm": 2.7935239013392237,
|
| 651 |
+
"learning_rate": 4.4956936350761005e-08,
|
| 652 |
+
"logits/chosen": -1.6336042881011963,
|
| 653 |
+
"logits/rejected": -1.5328700542449951,
|
| 654 |
+
"logps/chosen": -265.51251220703125,
|
| 655 |
+
"logps/rejected": -297.154541015625,
|
| 656 |
+
"loss": 0.0328,
|
| 657 |
+
"rewards/accuracies": 0.737500011920929,
|
| 658 |
+
"rewards/chosen": -0.1319248378276825,
|
| 659 |
+
"rewards/margins": 0.2542495131492615,
|
| 660 |
+
"rewards/rejected": -0.38617435097694397,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8486238532110092,
|
| 665 |
+
"grad_norm": 3.203845878056458,
|
| 666 |
+
"learning_rate": 3.416459164418123e-08,
|
| 667 |
+
"logits/chosen": -1.746138572692871,
|
| 668 |
+
"logits/rejected": -1.5625007152557373,
|
| 669 |
+
"logps/chosen": -313.5556640625,
|
| 670 |
+
"logps/rejected": -302.9591369628906,
|
| 671 |
+
"loss": 0.0326,
|
| 672 |
+
"rewards/accuracies": 0.737500011920929,
|
| 673 |
+
"rewards/chosen": -0.13971158862113953,
|
| 674 |
+
"rewards/margins": 0.2641337513923645,
|
| 675 |
+
"rewards/rejected": -0.4038453996181488,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8715596330275229,
|
| 680 |
+
"grad_norm": 3.9723478814606876,
|
| 681 |
+
"learning_rate": 2.475778302439524e-08,
|
| 682 |
+
"logits/chosen": -1.700884461402893,
|
| 683 |
+
"logits/rejected": -1.4876903295516968,
|
| 684 |
+
"logps/chosen": -308.3489074707031,
|
| 685 |
+
"logps/rejected": -299.65850830078125,
|
| 686 |
+
"loss": 0.0342,
|
| 687 |
+
"rewards/accuracies": 0.78125,
|
| 688 |
+
"rewards/chosen": -0.11970899254083633,
|
| 689 |
+
"rewards/margins": 0.3003136217594147,
|
| 690 |
+
"rewards/rejected": -0.4200226366519928,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8944954128440367,
|
| 695 |
+
"grad_norm": 3.433718178917216,
|
| 696 |
+
"learning_rate": 1.6796896657433805e-08,
|
| 697 |
+
"logits/chosen": -1.6044851541519165,
|
| 698 |
+
"logits/rejected": -1.4463411569595337,
|
| 699 |
+
"logps/chosen": -265.7919006347656,
|
| 700 |
+
"logps/rejected": -267.8514099121094,
|
| 701 |
+
"loss": 0.0356,
|
| 702 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 703 |
+
"rewards/chosen": -0.13834409415721893,
|
| 704 |
+
"rewards/margins": 0.2471076250076294,
|
| 705 |
+
"rewards/rejected": -0.3854517340660095,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9174311926605505,
|
| 710 |
+
"grad_norm": 3.420165449234039,
|
| 711 |
+
"learning_rate": 1.0333036740834855e-08,
|
| 712 |
+
"logits/chosen": -1.5561296939849854,
|
| 713 |
+
"logits/rejected": -1.4712246656417847,
|
| 714 |
+
"logps/chosen": -236.0593719482422,
|
| 715 |
+
"logps/rejected": -266.1527404785156,
|
| 716 |
+
"loss": 0.0351,
|
| 717 |
+
"rewards/accuracies": 0.731249988079071,
|
| 718 |
+
"rewards/chosen": -0.11499743163585663,
|
| 719 |
+
"rewards/margins": 0.2543741464614868,
|
| 720 |
+
"rewards/rejected": -0.36937159299850464,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9174311926605505,
|
| 725 |
+
"eval_logits/chosen": -1.6862857341766357,
|
| 726 |
+
"eval_logits/rejected": -1.476194977760315,
|
| 727 |
+
"eval_logps/chosen": -297.669189453125,
|
| 728 |
+
"eval_logps/rejected": -289.0508728027344,
|
| 729 |
+
"eval_loss": 0.0327373668551445,
|
| 730 |
+
"eval_rewards/accuracies": 0.7543103694915771,
|
| 731 |
+
"eval_rewards/chosen": -0.12579074501991272,
|
| 732 |
+
"eval_rewards/margins": 0.2994934022426605,
|
| 733 |
+
"eval_rewards/rejected": -0.42528414726257324,
|
| 734 |
+
"eval_runtime": 95.429,
|
| 735 |
+
"eval_samples_per_second": 19.051,
|
| 736 |
+
"eval_steps_per_second": 0.304,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9403669724770642,
|
| 741 |
+
"grad_norm": 3.5258604798563833,
|
| 742 |
+
"learning_rate": 5.4076974448211685e-09,
|
| 743 |
+
"logits/chosen": -1.5972687005996704,
|
| 744 |
+
"logits/rejected": -1.4251827001571655,
|
| 745 |
+
"logps/chosen": -284.83367919921875,
|
| 746 |
+
"logps/rejected": -281.9726257324219,
|
| 747 |
+
"loss": 0.0339,
|
| 748 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 749 |
+
"rewards/chosen": -0.13881805539131165,
|
| 750 |
+
"rewards/margins": 0.30044105648994446,
|
| 751 |
+
"rewards/rejected": -0.4392591118812561,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.963302752293578,
|
| 756 |
+
"grad_norm": 5.008109192532265,
|
| 757 |
+
"learning_rate": 2.052496544188487e-09,
|
| 758 |
+
"logits/chosen": -1.5873286724090576,
|
| 759 |
+
"logits/rejected": -1.4294344186782837,
|
| 760 |
+
"logps/chosen": -275.04388427734375,
|
| 761 |
+
"logps/rejected": -290.7890319824219,
|
| 762 |
+
"loss": 0.0341,
|
| 763 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 764 |
+
"rewards/chosen": -0.1560489982366562,
|
| 765 |
+
"rewards/margins": 0.293067991733551,
|
| 766 |
+
"rewards/rejected": -0.4491170048713684,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9862385321100917,
|
| 771 |
+
"grad_norm": 3.385882880558537,
|
| 772 |
+
"learning_rate": 2.889724508297886e-10,
|
| 773 |
+
"logits/chosen": -1.6612422466278076,
|
| 774 |
+
"logits/rejected": -1.3801202774047852,
|
| 775 |
+
"logps/chosen": -316.4678039550781,
|
| 776 |
+
"logps/rejected": -282.8833923339844,
|
| 777 |
+
"loss": 0.0343,
|
| 778 |
+
"rewards/accuracies": 0.768750011920929,
|
| 779 |
+
"rewards/chosen": -0.13292355835437775,
|
| 780 |
+
"rewards/margins": 0.25554487109184265,
|
| 781 |
+
"rewards/rejected": -0.3884683847427368,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 1.0,
|
| 786 |
+
"step": 436,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.0384325077500912,
|
| 789 |
+
"train_runtime": 11659.9634,
|
| 790 |
+
"train_samples_per_second": 4.782,
|
| 791 |
+
"train_steps_per_second": 0.037
|
| 792 |
+
}
|
| 793 |
+
],
|
| 794 |
+
"logging_steps": 10,
|
| 795 |
+
"max_steps": 436,
|
| 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 |
+
}
|