Model save
Browse files- README.md +80 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +815 -0
README.md
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
model-index:
|
| 9 |
+
- name: zephyr-7b-dpo-full-gpt-reward-scale-05
|
| 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-reward-scale-05
|
| 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: 0.5099
|
| 21 |
+
- Rewards/chosen: -2.9128
|
| 22 |
+
- Rewards/rejected: -4.2868
|
| 23 |
+
- Rewards/accuracies: 0.7328
|
| 24 |
+
- Rewards/margins: 1.3740
|
| 25 |
+
- Logps/rejected: -674.3233
|
| 26 |
+
- Logps/chosen: -575.2830
|
| 27 |
+
- Logits/rejected: 2.9146
|
| 28 |
+
- Logits/chosen: 0.8634
|
| 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 |
+
| 0.6688 | 0.1147 | 50 | 0.6561 | -0.0264 | -0.1297 | 0.6767 | 0.1033 | -258.6208 | -286.6466 | -2.5069 | -2.6066 |
|
| 66 |
+
| 0.5822 | 0.2294 | 100 | 0.5813 | -0.9069 | -1.4729 | 0.6724 | 0.5660 | -392.9425 | -374.6986 | 0.4795 | -0.0936 |
|
| 67 |
+
| 0.5512 | 0.3440 | 150 | 0.5534 | -0.9841 | -1.8057 | 0.7284 | 0.8216 | -426.2166 | -382.4128 | 2.9841 | 1.9098 |
|
| 68 |
+
| 0.5364 | 0.4587 | 200 | 0.5368 | -1.4185 | -2.3616 | 0.7328 | 0.9430 | -481.8062 | -425.8600 | 3.8766 | 2.4303 |
|
| 69 |
+
| 0.5308 | 0.5734 | 250 | 0.5235 | -2.3552 | -3.5698 | 0.7284 | 1.2145 | -602.6266 | -519.5305 | 4.2409 | 2.4548 |
|
| 70 |
+
| 0.532 | 0.6881 | 300 | 0.5117 | -2.5133 | -3.7138 | 0.7198 | 1.2004 | -617.0262 | -535.3408 | 3.0154 | 1.1080 |
|
| 71 |
+
| 0.5064 | 0.8028 | 350 | 0.5117 | -2.6361 | -3.9100 | 0.7241 | 1.2739 | -636.6490 | -547.6198 | 2.6854 | 0.7127 |
|
| 72 |
+
| 0.5105 | 0.9174 | 400 | 0.5099 | -2.9128 | -4.2868 | 0.7328 | 1.3740 | -674.3233 | -575.2830 | 2.9146 | 0.8634 |
|
| 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": 0.5604413999330013,
|
| 5 |
+
"train_runtime": 11415.5934,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.884,
|
| 8 |
+
"train_steps_per_second": 0.038
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"transformers_version": "4.44.0.dev0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5604413999330013,
|
| 5 |
+
"train_runtime": 11415.5934,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.884,
|
| 8 |
+
"train_steps_per_second": 0.038
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,815 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 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": 5.353972534143438,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.6582446098327637,
|
| 16 |
+
"logits/rejected": -2.612395763397217,
|
| 17 |
+
"logps/chosen": -310.3081359863281,
|
| 18 |
+
"logps/rejected": -241.6246337890625,
|
| 19 |
+
"loss": 0.6932,
|
| 20 |
+
"rewards/accuracies": 0.3499999940395355,
|
| 21 |
+
"rewards/chosen": -0.00043685571290552616,
|
| 22 |
+
"rewards/margins": -0.0005496515659615397,
|
| 23 |
+
"rewards/rejected": 0.0001127958094002679,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045871559633027525,
|
| 28 |
+
"grad_norm": 6.431385284276218,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.690976142883301,
|
| 31 |
+
"logits/rejected": -2.615501880645752,
|
| 32 |
+
"logps/chosen": -293.55859375,
|
| 33 |
+
"logps/rejected": -265.65789794921875,
|
| 34 |
+
"loss": 0.6924,
|
| 35 |
+
"rewards/accuracies": 0.612500011920929,
|
| 36 |
+
"rewards/chosen": 0.001354431384243071,
|
| 37 |
+
"rewards/margins": 0.0023786118254065514,
|
| 38 |
+
"rewards/rejected": -0.0010241802083328366,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06880733944954129,
|
| 43 |
+
"grad_norm": 5.140938328988767,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.6976418495178223,
|
| 46 |
+
"logits/rejected": -2.6304168701171875,
|
| 47 |
+
"logps/chosen": -277.8341064453125,
|
| 48 |
+
"logps/rejected": -297.1772155761719,
|
| 49 |
+
"loss": 0.6892,
|
| 50 |
+
"rewards/accuracies": 0.637499988079071,
|
| 51 |
+
"rewards/chosen": 0.004078004974871874,
|
| 52 |
+
"rewards/margins": 0.009664928540587425,
|
| 53 |
+
"rewards/rejected": -0.005586923565715551,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09174311926605505,
|
| 58 |
+
"grad_norm": 5.971632655809275,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.616170883178711,
|
| 61 |
+
"logits/rejected": -2.5451369285583496,
|
| 62 |
+
"logps/chosen": -283.9632568359375,
|
| 63 |
+
"logps/rejected": -259.82861328125,
|
| 64 |
+
"loss": 0.6798,
|
| 65 |
+
"rewards/accuracies": 0.6875,
|
| 66 |
+
"rewards/chosen": 0.03654901683330536,
|
| 67 |
+
"rewards/margins": 0.045721281319856644,
|
| 68 |
+
"rewards/rejected": -0.00917226541787386,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11467889908256881,
|
| 73 |
+
"grad_norm": 5.916656852320022,
|
| 74 |
+
"learning_rate": 4.997110275491701e-07,
|
| 75 |
+
"logits/chosen": -2.5970985889434814,
|
| 76 |
+
"logits/rejected": -2.5133914947509766,
|
| 77 |
+
"logps/chosen": -285.24835205078125,
|
| 78 |
+
"logps/rejected": -247.303466796875,
|
| 79 |
+
"loss": 0.6688,
|
| 80 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 81 |
+
"rewards/chosen": -0.006145569030195475,
|
| 82 |
+
"rewards/margins": 0.0578111931681633,
|
| 83 |
+
"rewards/rejected": -0.06395676732063293,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11467889908256881,
|
| 88 |
+
"eval_logits/chosen": -2.6066324710845947,
|
| 89 |
+
"eval_logits/rejected": -2.506901979446411,
|
| 90 |
+
"eval_logps/chosen": -286.6465759277344,
|
| 91 |
+
"eval_logps/rejected": -258.62078857421875,
|
| 92 |
+
"eval_loss": 0.6561177968978882,
|
| 93 |
+
"eval_rewards/accuracies": 0.6767241358757019,
|
| 94 |
+
"eval_rewards/chosen": -0.02640603668987751,
|
| 95 |
+
"eval_rewards/margins": 0.10332722216844559,
|
| 96 |
+
"eval_rewards/rejected": -0.12973324954509735,
|
| 97 |
+
"eval_runtime": 91.0244,
|
| 98 |
+
"eval_samples_per_second": 19.973,
|
| 99 |
+
"eval_steps_per_second": 0.319,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13761467889908258,
|
| 104 |
+
"grad_norm": 7.499634288772489,
|
| 105 |
+
"learning_rate": 4.979475034558115e-07,
|
| 106 |
+
"logits/chosen": -2.582371234893799,
|
| 107 |
+
"logits/rejected": -2.5081627368927,
|
| 108 |
+
"logps/chosen": -292.10491943359375,
|
| 109 |
+
"logps/rejected": -282.31195068359375,
|
| 110 |
+
"loss": 0.6423,
|
| 111 |
+
"rewards/accuracies": 0.675000011920929,
|
| 112 |
+
"rewards/chosen": -0.11516664922237396,
|
| 113 |
+
"rewards/margins": 0.19041100144386292,
|
| 114 |
+
"rewards/rejected": -0.30557766556739807,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16055045871559634,
|
| 119 |
+
"grad_norm": 17.30037068758165,
|
| 120 |
+
"learning_rate": 4.945923025551788e-07,
|
| 121 |
+
"logits/chosen": -2.4502875804901123,
|
| 122 |
+
"logits/rejected": -2.3790054321289062,
|
| 123 |
+
"logps/chosen": -298.32244873046875,
|
| 124 |
+
"logps/rejected": -273.11859130859375,
|
| 125 |
+
"loss": 0.6397,
|
| 126 |
+
"rewards/accuracies": 0.675000011920929,
|
| 127 |
+
"rewards/chosen": -0.2997247576713562,
|
| 128 |
+
"rewards/margins": 0.23786215484142303,
|
| 129 |
+
"rewards/rejected": -0.5375869870185852,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.1834862385321101,
|
| 134 |
+
"grad_norm": 11.228813057299567,
|
| 135 |
+
"learning_rate": 4.896669632591651e-07,
|
| 136 |
+
"logits/chosen": -2.5100908279418945,
|
| 137 |
+
"logits/rejected": -2.4027259349823,
|
| 138 |
+
"logps/chosen": -306.67510986328125,
|
| 139 |
+
"logps/rejected": -322.7925720214844,
|
| 140 |
+
"loss": 0.6257,
|
| 141 |
+
"rewards/accuracies": 0.6875,
|
| 142 |
+
"rewards/chosen": -0.36647987365722656,
|
| 143 |
+
"rewards/margins": 0.28450754284858704,
|
| 144 |
+
"rewards/rejected": -0.6509873867034912,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.20642201834862386,
|
| 149 |
+
"grad_norm": 15.58920411413326,
|
| 150 |
+
"learning_rate": 4.832031033425662e-07,
|
| 151 |
+
"logits/chosen": -1.5505931377410889,
|
| 152 |
+
"logits/rejected": -1.3694034814834595,
|
| 153 |
+
"logps/chosen": -357.6716613769531,
|
| 154 |
+
"logps/rejected": -372.05133056640625,
|
| 155 |
+
"loss": 0.5967,
|
| 156 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 157 |
+
"rewards/chosen": -0.5348917245864868,
|
| 158 |
+
"rewards/margins": 0.4341323971748352,
|
| 159 |
+
"rewards/rejected": -0.9690243005752563,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22935779816513763,
|
| 164 |
+
"grad_norm": 13.006583087547677,
|
| 165 |
+
"learning_rate": 4.752422169756047e-07,
|
| 166 |
+
"logits/chosen": -0.7837198972702026,
|
| 167 |
+
"logits/rejected": -0.35428792238235474,
|
| 168 |
+
"logps/chosen": -326.9918518066406,
|
| 169 |
+
"logps/rejected": -346.36737060546875,
|
| 170 |
+
"loss": 0.5822,
|
| 171 |
+
"rewards/accuracies": 0.65625,
|
| 172 |
+
"rewards/chosen": -0.4772585332393646,
|
| 173 |
+
"rewards/margins": 0.4704399108886719,
|
| 174 |
+
"rewards/rejected": -0.9476984143257141,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22935779816513763,
|
| 179 |
+
"eval_logits/chosen": -0.09357786923646927,
|
| 180 |
+
"eval_logits/rejected": 0.4795497953891754,
|
| 181 |
+
"eval_logps/chosen": -374.6986083984375,
|
| 182 |
+
"eval_logps/rejected": -392.9424743652344,
|
| 183 |
+
"eval_loss": 0.5813368558883667,
|
| 184 |
+
"eval_rewards/accuracies": 0.6724137663841248,
|
| 185 |
+
"eval_rewards/chosen": -0.9069267511367798,
|
| 186 |
+
"eval_rewards/margins": 0.5660232305526733,
|
| 187 |
+
"eval_rewards/rejected": -1.4729499816894531,
|
| 188 |
+
"eval_runtime": 91.4662,
|
| 189 |
+
"eval_samples_per_second": 19.876,
|
| 190 |
+
"eval_steps_per_second": 0.317,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25229357798165136,
|
| 195 |
+
"grad_norm": 14.7655268267239,
|
| 196 |
+
"learning_rate": 4.658354083558188e-07,
|
| 197 |
+
"logits/chosen": -0.23613190650939941,
|
| 198 |
+
"logits/rejected": 0.2948758006095886,
|
| 199 |
+
"logps/chosen": -371.15667724609375,
|
| 200 |
+
"logps/rejected": -427.76885986328125,
|
| 201 |
+
"loss": 0.5606,
|
| 202 |
+
"rewards/accuracies": 0.675000011920929,
|
| 203 |
+
"rewards/chosen": -0.7391853332519531,
|
| 204 |
+
"rewards/margins": 0.7208150625228882,
|
| 205 |
+
"rewards/rejected": -1.4600005149841309,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.27522935779816515,
|
| 210 |
+
"grad_norm": 28.232913631245626,
|
| 211 |
+
"learning_rate": 4.550430636492389e-07,
|
| 212 |
+
"logits/chosen": 0.3172193467617035,
|
| 213 |
+
"logits/rejected": 1.228100299835205,
|
| 214 |
+
"logps/chosen": -412.1929626464844,
|
| 215 |
+
"logps/rejected": -428.08056640625,
|
| 216 |
+
"loss": 0.5789,
|
| 217 |
+
"rewards/accuracies": 0.675000011920929,
|
| 218 |
+
"rewards/chosen": -1.0178120136260986,
|
| 219 |
+
"rewards/margins": 0.6625908613204956,
|
| 220 |
+
"rewards/rejected": -1.6804027557373047,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.2981651376146789,
|
| 225 |
+
"grad_norm": 23.106046920597972,
|
| 226 |
+
"learning_rate": 4.429344633468004e-07,
|
| 227 |
+
"logits/chosen": 1.296276330947876,
|
| 228 |
+
"logits/rejected": 2.0952706336975098,
|
| 229 |
+
"logps/chosen": -377.18572998046875,
|
| 230 |
+
"logps/rejected": -435.5022888183594,
|
| 231 |
+
"loss": 0.5712,
|
| 232 |
+
"rewards/accuracies": 0.737500011920929,
|
| 233 |
+
"rewards/chosen": -0.8953048586845398,
|
| 234 |
+
"rewards/margins": 0.8917394876480103,
|
| 235 |
+
"rewards/rejected": -1.7870445251464844,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.3211009174311927,
|
| 240 |
+
"grad_norm": 21.3617509080007,
|
| 241 |
+
"learning_rate": 4.2958733752443187e-07,
|
| 242 |
+
"logits/chosen": 1.0354318618774414,
|
| 243 |
+
"logits/rejected": 2.103768825531006,
|
| 244 |
+
"logps/chosen": -374.42938232421875,
|
| 245 |
+
"logps/rejected": -408.32342529296875,
|
| 246 |
+
"loss": 0.5477,
|
| 247 |
+
"rewards/accuracies": 0.71875,
|
| 248 |
+
"rewards/chosen": -0.9035611152648926,
|
| 249 |
+
"rewards/margins": 0.7404158115386963,
|
| 250 |
+
"rewards/rejected": -1.643977165222168,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.3440366972477064,
|
| 255 |
+
"grad_norm": 21.53708307235743,
|
| 256 |
+
"learning_rate": 4.150873668617898e-07,
|
| 257 |
+
"logits/chosen": 0.8976553678512573,
|
| 258 |
+
"logits/rejected": 2.0599629878997803,
|
| 259 |
+
"logps/chosen": -370.0615234375,
|
| 260 |
+
"logps/rejected": -411.820068359375,
|
| 261 |
+
"loss": 0.5512,
|
| 262 |
+
"rewards/accuracies": 0.71875,
|
| 263 |
+
"rewards/chosen": -0.7934576869010925,
|
| 264 |
+
"rewards/margins": 0.716931939125061,
|
| 265 |
+
"rewards/rejected": -1.5103896856307983,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.3440366972477064,
|
| 270 |
+
"eval_logits/chosen": 1.9097994565963745,
|
| 271 |
+
"eval_logits/rejected": 2.9840593338012695,
|
| 272 |
+
"eval_logps/chosen": -382.4128112792969,
|
| 273 |
+
"eval_logps/rejected": -426.216552734375,
|
| 274 |
+
"eval_loss": 0.5533820390701294,
|
| 275 |
+
"eval_rewards/accuracies": 0.7284482717514038,
|
| 276 |
+
"eval_rewards/chosen": -0.9840683937072754,
|
| 277 |
+
"eval_rewards/margins": 0.8216219544410706,
|
| 278 |
+
"eval_rewards/rejected": -1.8056902885437012,
|
| 279 |
+
"eval_runtime": 91.5586,
|
| 280 |
+
"eval_samples_per_second": 19.856,
|
| 281 |
+
"eval_steps_per_second": 0.317,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3669724770642202,
|
| 286 |
+
"grad_norm": 17.634177990437703,
|
| 287 |
+
"learning_rate": 3.9952763262280397e-07,
|
| 288 |
+
"logits/chosen": 1.8957267999649048,
|
| 289 |
+
"logits/rejected": 2.8357367515563965,
|
| 290 |
+
"logps/chosen": -408.04705810546875,
|
| 291 |
+
"logps/rejected": -450.3290100097656,
|
| 292 |
+
"loss": 0.5609,
|
| 293 |
+
"rewards/accuracies": 0.731249988079071,
|
| 294 |
+
"rewards/chosen": -1.1164201498031616,
|
| 295 |
+
"rewards/margins": 0.8357731103897095,
|
| 296 |
+
"rewards/rejected": -1.952193021774292,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38990825688073394,
|
| 301 |
+
"grad_norm": 24.6798606158854,
|
| 302 |
+
"learning_rate": 3.8300801912883414e-07,
|
| 303 |
+
"logits/chosen": 1.3192155361175537,
|
| 304 |
+
"logits/rejected": 2.3843648433685303,
|
| 305 |
+
"logps/chosen": -356.1672058105469,
|
| 306 |
+
"logps/rejected": -387.1358947753906,
|
| 307 |
+
"loss": 0.5414,
|
| 308 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 309 |
+
"rewards/chosen": -0.9024198651313782,
|
| 310 |
+
"rewards/margins": 0.7546060681343079,
|
| 311 |
+
"rewards/rejected": -1.6570260524749756,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.41284403669724773,
|
| 316 |
+
"grad_norm": 22.218831723434445,
|
| 317 |
+
"learning_rate": 3.6563457256020884e-07,
|
| 318 |
+
"logits/chosen": 1.3455697298049927,
|
| 319 |
+
"logits/rejected": 2.5438590049743652,
|
| 320 |
+
"logps/chosen": -351.62774658203125,
|
| 321 |
+
"logps/rejected": -430.138427734375,
|
| 322 |
+
"loss": 0.5396,
|
| 323 |
+
"rewards/accuracies": 0.731249988079071,
|
| 324 |
+
"rewards/chosen": -0.9148648381233215,
|
| 325 |
+
"rewards/margins": 0.9510505795478821,
|
| 326 |
+
"rewards/rejected": -1.8659156560897827,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.43577981651376146,
|
| 331 |
+
"grad_norm": 25.297690497973328,
|
| 332 |
+
"learning_rate": 3.475188202022617e-07,
|
| 333 |
+
"logits/chosen": 1.7387921810150146,
|
| 334 |
+
"logits/rejected": 2.998396396636963,
|
| 335 |
+
"logps/chosen": -333.116455078125,
|
| 336 |
+
"logps/rejected": -437.8639221191406,
|
| 337 |
+
"loss": 0.5399,
|
| 338 |
+
"rewards/accuracies": 0.737500011920929,
|
| 339 |
+
"rewards/chosen": -0.7444877624511719,
|
| 340 |
+
"rewards/margins": 1.007294774055481,
|
| 341 |
+
"rewards/rejected": -1.7517824172973633,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45871559633027525,
|
| 346 |
+
"grad_norm": 25.947440207471573,
|
| 347 |
+
"learning_rate": 3.287770545059052e-07,
|
| 348 |
+
"logits/chosen": 1.9480648040771484,
|
| 349 |
+
"logits/rejected": 2.9033870697021484,
|
| 350 |
+
"logps/chosen": -380.21185302734375,
|
| 351 |
+
"logps/rejected": -424.3128967285156,
|
| 352 |
+
"loss": 0.5364,
|
| 353 |
+
"rewards/accuracies": 0.699999988079071,
|
| 354 |
+
"rewards/chosen": -1.1435054540634155,
|
| 355 |
+
"rewards/margins": 0.736918032169342,
|
| 356 |
+
"rewards/rejected": -1.8804235458374023,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45871559633027525,
|
| 361 |
+
"eval_logits/chosen": 2.4303267002105713,
|
| 362 |
+
"eval_logits/rejected": 3.876626491546631,
|
| 363 |
+
"eval_logps/chosen": -425.8599853515625,
|
| 364 |
+
"eval_logps/rejected": -481.8061828613281,
|
| 365 |
+
"eval_loss": 0.5367683172225952,
|
| 366 |
+
"eval_rewards/accuracies": 0.732758641242981,
|
| 367 |
+
"eval_rewards/chosen": -1.4185398817062378,
|
| 368 |
+
"eval_rewards/margins": 0.9430465698242188,
|
| 369 |
+
"eval_rewards/rejected": -2.361586570739746,
|
| 370 |
+
"eval_runtime": 91.0832,
|
| 371 |
+
"eval_samples_per_second": 19.96,
|
| 372 |
+
"eval_steps_per_second": 0.318,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.481651376146789,
|
| 377 |
+
"grad_norm": 19.563478098052215,
|
| 378 |
+
"learning_rate": 3.0952958655864954e-07,
|
| 379 |
+
"logits/chosen": 3.3911328315734863,
|
| 380 |
+
"logits/rejected": 4.194566249847412,
|
| 381 |
+
"logps/chosen": -451.25238037109375,
|
| 382 |
+
"logps/rejected": -539.4793701171875,
|
| 383 |
+
"loss": 0.5321,
|
| 384 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 385 |
+
"rewards/chosen": -1.744564414024353,
|
| 386 |
+
"rewards/margins": 0.8036998510360718,
|
| 387 |
+
"rewards/rejected": -2.548264503479004,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5045871559633027,
|
| 392 |
+
"grad_norm": 24.94888206530693,
|
| 393 |
+
"learning_rate": 2.898999737583448e-07,
|
| 394 |
+
"logits/chosen": 2.1577422618865967,
|
| 395 |
+
"logits/rejected": 3.670943021774292,
|
| 396 |
+
"logps/chosen": -426.3487854003906,
|
| 397 |
+
"logps/rejected": -508.80780029296875,
|
| 398 |
+
"loss": 0.5365,
|
| 399 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 400 |
+
"rewards/chosen": -1.5799643993377686,
|
| 401 |
+
"rewards/margins": 0.9677878618240356,
|
| 402 |
+
"rewards/rejected": -2.5477521419525146,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5275229357798165,
|
| 407 |
+
"grad_norm": 23.826139773858404,
|
| 408 |
+
"learning_rate": 2.7001422664752333e-07,
|
| 409 |
+
"logits/chosen": 0.8777297735214233,
|
| 410 |
+
"logits/rejected": 2.3443570137023926,
|
| 411 |
+
"logps/chosen": -384.8174743652344,
|
| 412 |
+
"logps/rejected": -467.87298583984375,
|
| 413 |
+
"loss": 0.5416,
|
| 414 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 415 |
+
"rewards/chosen": -1.1064906120300293,
|
| 416 |
+
"rewards/margins": 1.0363706350326538,
|
| 417 |
+
"rewards/rejected": -2.1428613662719727,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5504587155963303,
|
| 422 |
+
"grad_norm": 18.150746967508407,
|
| 423 |
+
"learning_rate": 2.5e-07,
|
| 424 |
+
"logits/chosen": 2.9278922080993652,
|
| 425 |
+
"logits/rejected": 3.511791706085205,
|
| 426 |
+
"logps/chosen": -450.8653259277344,
|
| 427 |
+
"logps/rejected": -547.5108642578125,
|
| 428 |
+
"loss": 0.5638,
|
| 429 |
+
"rewards/accuracies": 0.668749988079071,
|
| 430 |
+
"rewards/chosen": -1.810927391052246,
|
| 431 |
+
"rewards/margins": 1.0066088438034058,
|
| 432 |
+
"rewards/rejected": -2.8175363540649414,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.573394495412844,
|
| 437 |
+
"grad_norm": 21.73286552315769,
|
| 438 |
+
"learning_rate": 2.2998577335247667e-07,
|
| 439 |
+
"logits/chosen": 2.6097209453582764,
|
| 440 |
+
"logits/rejected": 3.895547389984131,
|
| 441 |
+
"logps/chosen": -484.319091796875,
|
| 442 |
+
"logps/rejected": -561.3115234375,
|
| 443 |
+
"loss": 0.5308,
|
| 444 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 445 |
+
"rewards/chosen": -2.1941843032836914,
|
| 446 |
+
"rewards/margins": 0.9737985730171204,
|
| 447 |
+
"rewards/rejected": -3.167982578277588,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.573394495412844,
|
| 452 |
+
"eval_logits/chosen": 2.454780101776123,
|
| 453 |
+
"eval_logits/rejected": 4.2408928871154785,
|
| 454 |
+
"eval_logps/chosen": -519.5304565429688,
|
| 455 |
+
"eval_logps/rejected": -602.6265869140625,
|
| 456 |
+
"eval_loss": 0.5234553217887878,
|
| 457 |
+
"eval_rewards/accuracies": 0.7284482717514038,
|
| 458 |
+
"eval_rewards/chosen": -2.3552448749542236,
|
| 459 |
+
"eval_rewards/margins": 1.2145458459854126,
|
| 460 |
+
"eval_rewards/rejected": -3.5697906017303467,
|
| 461 |
+
"eval_runtime": 91.7528,
|
| 462 |
+
"eval_samples_per_second": 19.814,
|
| 463 |
+
"eval_steps_per_second": 0.316,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5963302752293578,
|
| 468 |
+
"grad_norm": 29.190042715796082,
|
| 469 |
+
"learning_rate": 2.1010002624165524e-07,
|
| 470 |
+
"logits/chosen": 2.5975215435028076,
|
| 471 |
+
"logits/rejected": 4.360453128814697,
|
| 472 |
+
"logps/chosen": -554.3533935546875,
|
| 473 |
+
"logps/rejected": -655.7716064453125,
|
| 474 |
+
"loss": 0.5307,
|
| 475 |
+
"rewards/accuracies": 0.800000011920929,
|
| 476 |
+
"rewards/chosen": -2.5185165405273438,
|
| 477 |
+
"rewards/margins": 1.447409987449646,
|
| 478 |
+
"rewards/rejected": -3.9659264087677,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6192660550458715,
|
| 483 |
+
"grad_norm": 20.081922974931803,
|
| 484 |
+
"learning_rate": 1.9047041344135043e-07,
|
| 485 |
+
"logits/chosen": 2.110996723175049,
|
| 486 |
+
"logits/rejected": 3.4121272563934326,
|
| 487 |
+
"logps/chosen": -542.6881103515625,
|
| 488 |
+
"logps/rejected": -613.2174682617188,
|
| 489 |
+
"loss": 0.5514,
|
| 490 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 491 |
+
"rewards/chosen": -2.664341449737549,
|
| 492 |
+
"rewards/margins": 0.9934176206588745,
|
| 493 |
+
"rewards/rejected": -3.657759189605713,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.6422018348623854,
|
| 498 |
+
"grad_norm": 20.797156741141926,
|
| 499 |
+
"learning_rate": 1.7122294549409482e-07,
|
| 500 |
+
"logits/chosen": 1.95541250705719,
|
| 501 |
+
"logits/rejected": 3.574702024459839,
|
| 502 |
+
"logps/chosen": -541.0426025390625,
|
| 503 |
+
"logps/rejected": -644.5146484375,
|
| 504 |
+
"loss": 0.522,
|
| 505 |
+
"rewards/accuracies": 0.699999988079071,
|
| 506 |
+
"rewards/chosen": -2.637174129486084,
|
| 507 |
+
"rewards/margins": 1.117336630821228,
|
| 508 |
+
"rewards/rejected": -3.7545104026794434,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6651376146788991,
|
| 513 |
+
"grad_norm": 22.250484161252675,
|
| 514 |
+
"learning_rate": 1.524811797977383e-07,
|
| 515 |
+
"logits/chosen": 1.9210926294326782,
|
| 516 |
+
"logits/rejected": 3.2735812664031982,
|
| 517 |
+
"logps/chosen": -576.992431640625,
|
| 518 |
+
"logps/rejected": -663.1363525390625,
|
| 519 |
+
"loss": 0.5191,
|
| 520 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 521 |
+
"rewards/chosen": -2.8690013885498047,
|
| 522 |
+
"rewards/margins": 1.0208569765090942,
|
| 523 |
+
"rewards/rejected": -3.8898582458496094,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6880733944954128,
|
| 528 |
+
"grad_norm": 24.39105290704078,
|
| 529 |
+
"learning_rate": 1.3436542743979125e-07,
|
| 530 |
+
"logits/chosen": 1.646095871925354,
|
| 531 |
+
"logits/rejected": 3.443913221359253,
|
| 532 |
+
"logps/chosen": -558.287353515625,
|
| 533 |
+
"logps/rejected": -657.3651123046875,
|
| 534 |
+
"loss": 0.532,
|
| 535 |
+
"rewards/accuracies": 0.731249988079071,
|
| 536 |
+
"rewards/chosen": -2.7306606769561768,
|
| 537 |
+
"rewards/margins": 1.3434104919433594,
|
| 538 |
+
"rewards/rejected": -4.074070930480957,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6880733944954128,
|
| 543 |
+
"eval_logits/chosen": 1.1080348491668701,
|
| 544 |
+
"eval_logits/rejected": 3.015399694442749,
|
| 545 |
+
"eval_logps/chosen": -535.3407592773438,
|
| 546 |
+
"eval_logps/rejected": -617.0262451171875,
|
| 547 |
+
"eval_loss": 0.5116756558418274,
|
| 548 |
+
"eval_rewards/accuracies": 0.7198275923728943,
|
| 549 |
+
"eval_rewards/chosen": -2.513347864151001,
|
| 550 |
+
"eval_rewards/margins": 1.200439691543579,
|
| 551 |
+
"eval_rewards/rejected": -3.713787794113159,
|
| 552 |
+
"eval_runtime": 91.1655,
|
| 553 |
+
"eval_samples_per_second": 19.942,
|
| 554 |
+
"eval_steps_per_second": 0.318,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7110091743119266,
|
| 559 |
+
"grad_norm": 27.989896858900266,
|
| 560 |
+
"learning_rate": 1.1699198087116588e-07,
|
| 561 |
+
"logits/chosen": 2.1850762367248535,
|
| 562 |
+
"logits/rejected": 3.5708484649658203,
|
| 563 |
+
"logps/chosen": -533.7103271484375,
|
| 564 |
+
"logps/rejected": -638.1661376953125,
|
| 565 |
+
"loss": 0.54,
|
| 566 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 567 |
+
"rewards/chosen": -2.7955188751220703,
|
| 568 |
+
"rewards/margins": 1.163648247718811,
|
| 569 |
+
"rewards/rejected": -3.959167003631592,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7339449541284404,
|
| 574 |
+
"grad_norm": 26.423781855718417,
|
| 575 |
+
"learning_rate": 1.00472367377196e-07,
|
| 576 |
+
"logits/chosen": 1.8076130151748657,
|
| 577 |
+
"logits/rejected": 4.071971893310547,
|
| 578 |
+
"logps/chosen": -614.3458251953125,
|
| 579 |
+
"logps/rejected": -702.5189819335938,
|
| 580 |
+
"loss": 0.5138,
|
| 581 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 582 |
+
"rewards/chosen": -3.0845131874084473,
|
| 583 |
+
"rewards/margins": 1.4407538175582886,
|
| 584 |
+
"rewards/rejected": -4.525267601013184,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7568807339449541,
|
| 589 |
+
"grad_norm": 24.484829935546955,
|
| 590 |
+
"learning_rate": 8.49126331382102e-08,
|
| 591 |
+
"logits/chosen": 2.0375962257385254,
|
| 592 |
+
"logits/rejected": 3.5112037658691406,
|
| 593 |
+
"logps/chosen": -607.512939453125,
|
| 594 |
+
"logps/rejected": -720.7943115234375,
|
| 595 |
+
"loss": 0.5187,
|
| 596 |
+
"rewards/accuracies": 0.6875,
|
| 597 |
+
"rewards/chosen": -3.2878499031066895,
|
| 598 |
+
"rewards/margins": 1.2010066509246826,
|
| 599 |
+
"rewards/rejected": -4.488856315612793,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7798165137614679,
|
| 604 |
+
"grad_norm": 23.41145949202815,
|
| 605 |
+
"learning_rate": 7.041266247556812e-08,
|
| 606 |
+
"logits/chosen": 1.657248854637146,
|
| 607 |
+
"logits/rejected": 3.65797758102417,
|
| 608 |
+
"logps/chosen": -553.1179809570312,
|
| 609 |
+
"logps/rejected": -692.9093017578125,
|
| 610 |
+
"loss": 0.5258,
|
| 611 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 612 |
+
"rewards/chosen": -2.909062623977661,
|
| 613 |
+
"rewards/margins": 1.4310705661773682,
|
| 614 |
+
"rewards/rejected": -4.340132713317871,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8027522935779816,
|
| 619 |
+
"grad_norm": 28.908000369301963,
|
| 620 |
+
"learning_rate": 5.706553665319955e-08,
|
| 621 |
+
"logits/chosen": 0.9193560481071472,
|
| 622 |
+
"logits/rejected": 3.4261555671691895,
|
| 623 |
+
"logps/chosen": -548.4246215820312,
|
| 624 |
+
"logps/rejected": -669.2120361328125,
|
| 625 |
+
"loss": 0.5064,
|
| 626 |
+
"rewards/accuracies": 0.8125,
|
| 627 |
+
"rewards/chosen": -2.572448968887329,
|
| 628 |
+
"rewards/margins": 1.67291259765625,
|
| 629 |
+
"rewards/rejected": -4.245361804962158,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8027522935779816,
|
| 634 |
+
"eval_logits/chosen": 0.7126501202583313,
|
| 635 |
+
"eval_logits/rejected": 2.685429096221924,
|
| 636 |
+
"eval_logps/chosen": -547.6198120117188,
|
| 637 |
+
"eval_logps/rejected": -636.6490478515625,
|
| 638 |
+
"eval_loss": 0.5116574168205261,
|
| 639 |
+
"eval_rewards/accuracies": 0.7241379022598267,
|
| 640 |
+
"eval_rewards/chosen": -2.636138677597046,
|
| 641 |
+
"eval_rewards/margins": 1.2738765478134155,
|
| 642 |
+
"eval_rewards/rejected": -3.9100148677825928,
|
| 643 |
+
"eval_runtime": 90.9536,
|
| 644 |
+
"eval_samples_per_second": 19.988,
|
| 645 |
+
"eval_steps_per_second": 0.319,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8256880733944955,
|
| 650 |
+
"grad_norm": 24.395714355317615,
|
| 651 |
+
"learning_rate": 4.4956936350761005e-08,
|
| 652 |
+
"logits/chosen": 0.9300888180732727,
|
| 653 |
+
"logits/rejected": 2.3748581409454346,
|
| 654 |
+
"logps/chosen": -543.5307006835938,
|
| 655 |
+
"logps/rejected": -659.3692016601562,
|
| 656 |
+
"loss": 0.5084,
|
| 657 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 658 |
+
"rewards/chosen": -2.6378746032714844,
|
| 659 |
+
"rewards/margins": 1.2973625659942627,
|
| 660 |
+
"rewards/rejected": -3.935237169265747,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8486238532110092,
|
| 665 |
+
"grad_norm": 22.09369129566989,
|
| 666 |
+
"learning_rate": 3.416459164418123e-08,
|
| 667 |
+
"logits/chosen": 0.2887948155403137,
|
| 668 |
+
"logits/rejected": 2.4889461994171143,
|
| 669 |
+
"logps/chosen": -591.7805786132812,
|
| 670 |
+
"logps/rejected": -667.9407348632812,
|
| 671 |
+
"loss": 0.5109,
|
| 672 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 673 |
+
"rewards/chosen": -2.5994677543640137,
|
| 674 |
+
"rewards/margins": 1.3468105792999268,
|
| 675 |
+
"rewards/rejected": -3.9462783336639404,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8715596330275229,
|
| 680 |
+
"grad_norm": 26.209345843998328,
|
| 681 |
+
"learning_rate": 2.475778302439524e-08,
|
| 682 |
+
"logits/chosen": 0.8550162315368652,
|
| 683 |
+
"logits/rejected": 3.1779205799102783,
|
| 684 |
+
"logps/chosen": -580.2351684570312,
|
| 685 |
+
"logps/rejected": -637.8566284179688,
|
| 686 |
+
"loss": 0.5075,
|
| 687 |
+
"rewards/accuracies": 0.668749988079071,
|
| 688 |
+
"rewards/chosen": -2.8695898056030273,
|
| 689 |
+
"rewards/margins": 1.257644534111023,
|
| 690 |
+
"rewards/rejected": -4.127234935760498,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8944954128440367,
|
| 695 |
+
"grad_norm": 21.12858235158005,
|
| 696 |
+
"learning_rate": 1.6796896657433805e-08,
|
| 697 |
+
"logits/chosen": 0.47364893555641174,
|
| 698 |
+
"logits/rejected": 2.78879451751709,
|
| 699 |
+
"logps/chosen": -587.971435546875,
|
| 700 |
+
"logps/rejected": -724.6598510742188,
|
| 701 |
+
"loss": 0.5014,
|
| 702 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 703 |
+
"rewards/chosen": -2.7999932765960693,
|
| 704 |
+
"rewards/margins": 1.9162429571151733,
|
| 705 |
+
"rewards/rejected": -4.716236591339111,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9174311926605505,
|
| 710 |
+
"grad_norm": 23.688779288096637,
|
| 711 |
+
"learning_rate": 1.0333036740834855e-08,
|
| 712 |
+
"logits/chosen": 1.2185232639312744,
|
| 713 |
+
"logits/rejected": 2.7246413230895996,
|
| 714 |
+
"logps/chosen": -606.762939453125,
|
| 715 |
+
"logps/rejected": -717.7750244140625,
|
| 716 |
+
"loss": 0.5105,
|
| 717 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 718 |
+
"rewards/chosen": -3.1490612030029297,
|
| 719 |
+
"rewards/margins": 1.1964399814605713,
|
| 720 |
+
"rewards/rejected": -4.345500946044922,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9174311926605505,
|
| 725 |
+
"eval_logits/chosen": 0.863433837890625,
|
| 726 |
+
"eval_logits/rejected": 2.9146454334259033,
|
| 727 |
+
"eval_logps/chosen": -575.282958984375,
|
| 728 |
+
"eval_logps/rejected": -674.3233032226562,
|
| 729 |
+
"eval_loss": 0.5098804235458374,
|
| 730 |
+
"eval_rewards/accuracies": 0.732758641242981,
|
| 731 |
+
"eval_rewards/chosen": -2.9127700328826904,
|
| 732 |
+
"eval_rewards/margins": 1.3739889860153198,
|
| 733 |
+
"eval_rewards/rejected": -4.286758899688721,
|
| 734 |
+
"eval_runtime": 91.1821,
|
| 735 |
+
"eval_samples_per_second": 19.938,
|
| 736 |
+
"eval_steps_per_second": 0.318,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9403669724770642,
|
| 741 |
+
"grad_norm": 23.71509307695075,
|
| 742 |
+
"learning_rate": 5.4076974448211685e-09,
|
| 743 |
+
"logits/chosen": 1.313011884689331,
|
| 744 |
+
"logits/rejected": 2.788435459136963,
|
| 745 |
+
"logps/chosen": -609.2062377929688,
|
| 746 |
+
"logps/rejected": -684.1138305664062,
|
| 747 |
+
"loss": 0.5379,
|
| 748 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 749 |
+
"rewards/chosen": -3.2180449962615967,
|
| 750 |
+
"rewards/margins": 1.0637754201889038,
|
| 751 |
+
"rewards/rejected": -4.281820297241211,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.963302752293578,
|
| 756 |
+
"grad_norm": 27.57431143958426,
|
| 757 |
+
"learning_rate": 2.052496544188487e-09,
|
| 758 |
+
"logits/chosen": 1.1412100791931152,
|
| 759 |
+
"logits/rejected": 3.4668610095977783,
|
| 760 |
+
"logps/chosen": -616.3365478515625,
|
| 761 |
+
"logps/rejected": -685.5581665039062,
|
| 762 |
+
"loss": 0.5129,
|
| 763 |
+
"rewards/accuracies": 0.75,
|
| 764 |
+
"rewards/chosen": -3.115288496017456,
|
| 765 |
+
"rewards/margins": 1.404497504234314,
|
| 766 |
+
"rewards/rejected": -4.5197858810424805,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9862385321100917,
|
| 771 |
+
"grad_norm": 19.004038331375536,
|
| 772 |
+
"learning_rate": 2.889724508297886e-10,
|
| 773 |
+
"logits/chosen": 1.3194568157196045,
|
| 774 |
+
"logits/rejected": 2.8764185905456543,
|
| 775 |
+
"logps/chosen": -561.5572509765625,
|
| 776 |
+
"logps/rejected": -680.3653564453125,
|
| 777 |
+
"loss": 0.5165,
|
| 778 |
+
"rewards/accuracies": 0.768750011920929,
|
| 779 |
+
"rewards/chosen": -2.9964187145233154,
|
| 780 |
+
"rewards/margins": 1.2785086631774902,
|
| 781 |
+
"rewards/rejected": -4.274927616119385,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 1.0,
|
| 786 |
+
"step": 436,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.5604413999330013,
|
| 789 |
+
"train_runtime": 11415.5934,
|
| 790 |
+
"train_samples_per_second": 4.884,
|
| 791 |
+
"train_steps_per_second": 0.038
|
| 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 |
+
}
|