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-01
|
| 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-01
|
| 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: 0.5383
|
| 21 |
+
- Rewards/chosen: -1.3950
|
| 22 |
+
- Rewards/rejected: -2.4464
|
| 23 |
+
- Rewards/accuracies: 0.7241
|
| 24 |
+
- Rewards/margins: 1.0514
|
| 25 |
+
- Logps/rejected: -490.2826
|
| 26 |
+
- Logps/chosen: -423.5039
|
| 27 |
+
- Logits/rejected: 1.2749
|
| 28 |
+
- Logits/chosen: -0.2527
|
| 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 |
+
| 0.674 | 0.1147 | 50 | 0.6644 | -0.0456 | -0.1442 | 0.6724 | 0.0986 | -260.0650 | -288.5642 | -2.5039 | -2.6023 |
|
| 66 |
+
| 0.5874 | 0.2294 | 100 | 0.5920 | -0.9820 | -1.5650 | 0.6810 | 0.5830 | -402.1482 | -382.2076 | 0.3008 | -0.2226 |
|
| 67 |
+
| 0.5612 | 0.3440 | 150 | 0.5695 | -1.4677 | -2.3665 | 0.6897 | 0.8989 | -482.2998 | -430.7732 | 2.3140 | 1.4310 |
|
| 68 |
+
| 0.5427 | 0.4587 | 200 | 0.5523 | -1.3469 | -2.2624 | 0.7241 | 0.9156 | -471.8922 | -418.6947 | 0.9223 | -0.3630 |
|
| 69 |
+
| 0.5474 | 0.5734 | 250 | 0.5430 | -1.0958 | -2.0370 | 0.6897 | 0.9412 | -449.3501 | -393.5861 | 0.9071 | -0.4403 |
|
| 70 |
+
| 0.5556 | 0.6881 | 300 | 0.5404 | -1.3959 | -2.3862 | 0.7198 | 0.9903 | -484.2666 | -423.5919 | 1.1950 | -0.1993 |
|
| 71 |
+
| 0.5373 | 0.8028 | 350 | 0.5416 | -1.5583 | -2.5998 | 0.7284 | 1.0414 | -505.6230 | -439.8387 | 1.7159 | 0.2396 |
|
| 72 |
+
| 0.5405 | 0.9174 | 400 | 0.5383 | -1.3950 | -2.4464 | 0.7241 | 1.0514 | -490.2826 | -423.5039 | 1.2749 | -0.2527 |
|
| 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": 0.5759701597581216,
|
| 5 |
+
"train_runtime": 11541.3114,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.831,
|
| 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": 0.5759701597581216,
|
| 5 |
+
"train_runtime": 11541.3114,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.831,
|
| 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": 4.5521214161319365,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.658930778503418,
|
| 16 |
+
"logits/rejected": -2.6127209663391113,
|
| 17 |
+
"logps/chosen": -310.2911682128906,
|
| 18 |
+
"logps/rejected": -241.6273651123047,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.3812499940395355,
|
| 21 |
+
"rewards/chosen": -0.0002677589072845876,
|
| 22 |
+
"rewards/margins": -0.0003533684357535094,
|
| 23 |
+
"rewards/rejected": 8.560952846892178e-05,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045871559633027525,
|
| 28 |
+
"grad_norm": 5.31294051195239,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.691143751144409,
|
| 31 |
+
"logits/rejected": -2.6153030395507812,
|
| 32 |
+
"logps/chosen": -293.57037353515625,
|
| 33 |
+
"logps/rejected": -265.63885498046875,
|
| 34 |
+
"loss": 0.6926,
|
| 35 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 36 |
+
"rewards/chosen": 0.00123660231474787,
|
| 37 |
+
"rewards/margins": 0.00207112031057477,
|
| 38 |
+
"rewards/rejected": -0.0008345181122422218,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06880733944954129,
|
| 43 |
+
"grad_norm": 4.285586645859575,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.6962637901306152,
|
| 46 |
+
"logits/rejected": -2.628978729248047,
|
| 47 |
+
"logps/chosen": -277.83209228515625,
|
| 48 |
+
"logps/rejected": -297.1653137207031,
|
| 49 |
+
"loss": 0.6901,
|
| 50 |
+
"rewards/accuracies": 0.643750011920929,
|
| 51 |
+
"rewards/chosen": 0.004098366480320692,
|
| 52 |
+
"rewards/margins": 0.00956634059548378,
|
| 53 |
+
"rewards/rejected": -0.0054679736495018005,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09174311926605505,
|
| 58 |
+
"grad_norm": 5.111796420680735,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.6144208908081055,
|
| 61 |
+
"logits/rejected": -2.543553352355957,
|
| 62 |
+
"logps/chosen": -284.1578674316406,
|
| 63 |
+
"logps/rejected": -259.81475830078125,
|
| 64 |
+
"loss": 0.6826,
|
| 65 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 66 |
+
"rewards/chosen": 0.03460276871919632,
|
| 67 |
+
"rewards/margins": 0.0436362698674202,
|
| 68 |
+
"rewards/rejected": -0.009033503010869026,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11467889908256881,
|
| 73 |
+
"grad_norm": 4.717770708541697,
|
| 74 |
+
"learning_rate": 4.997110275491701e-07,
|
| 75 |
+
"logits/chosen": -2.591963291168213,
|
| 76 |
+
"logits/rejected": -2.5082640647888184,
|
| 77 |
+
"logps/chosen": -285.80047607421875,
|
| 78 |
+
"logps/rejected": -247.5881805419922,
|
| 79 |
+
"loss": 0.674,
|
| 80 |
+
"rewards/accuracies": 0.606249988079071,
|
| 81 |
+
"rewards/chosen": -0.011667141690850258,
|
| 82 |
+
"rewards/margins": 0.055136788636446,
|
| 83 |
+
"rewards/rejected": -0.06680391728878021,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11467889908256881,
|
| 88 |
+
"eval_logits/chosen": -2.6023333072662354,
|
| 89 |
+
"eval_logits/rejected": -2.5038623809814453,
|
| 90 |
+
"eval_logps/chosen": -288.564208984375,
|
| 91 |
+
"eval_logps/rejected": -260.0650329589844,
|
| 92 |
+
"eval_loss": 0.6643866300582886,
|
| 93 |
+
"eval_rewards/accuracies": 0.6724137663841248,
|
| 94 |
+
"eval_rewards/chosen": -0.04558267444372177,
|
| 95 |
+
"eval_rewards/margins": 0.09859247505664825,
|
| 96 |
+
"eval_rewards/rejected": -0.14417517185211182,
|
| 97 |
+
"eval_runtime": 91.2705,
|
| 98 |
+
"eval_samples_per_second": 19.919,
|
| 99 |
+
"eval_steps_per_second": 0.318,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13761467889908258,
|
| 104 |
+
"grad_norm": 6.057042762669634,
|
| 105 |
+
"learning_rate": 4.979475034558115e-07,
|
| 106 |
+
"logits/chosen": -2.584840774536133,
|
| 107 |
+
"logits/rejected": -2.5118155479431152,
|
| 108 |
+
"logps/chosen": -291.8733215332031,
|
| 109 |
+
"logps/rejected": -281.57476806640625,
|
| 110 |
+
"loss": 0.6519,
|
| 111 |
+
"rewards/accuracies": 0.699999988079071,
|
| 112 |
+
"rewards/chosen": -0.11285074800252914,
|
| 113 |
+
"rewards/margins": 0.18535511195659637,
|
| 114 |
+
"rewards/rejected": -0.2982058525085449,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16055045871559634,
|
| 119 |
+
"grad_norm": 12.297420969455343,
|
| 120 |
+
"learning_rate": 4.945923025551788e-07,
|
| 121 |
+
"logits/chosen": -2.4776272773742676,
|
| 122 |
+
"logits/rejected": -2.4097015857696533,
|
| 123 |
+
"logps/chosen": -299.2547912597656,
|
| 124 |
+
"logps/rejected": -271.7998046875,
|
| 125 |
+
"loss": 0.6509,
|
| 126 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 127 |
+
"rewards/chosen": -0.30904772877693176,
|
| 128 |
+
"rewards/margins": 0.21535193920135498,
|
| 129 |
+
"rewards/rejected": -0.5243996381759644,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.1834862385321101,
|
| 134 |
+
"grad_norm": 8.87671354702832,
|
| 135 |
+
"learning_rate": 4.896669632591651e-07,
|
| 136 |
+
"logits/chosen": -2.565410852432251,
|
| 137 |
+
"logits/rejected": -2.4615464210510254,
|
| 138 |
+
"logps/chosen": -310.5340270996094,
|
| 139 |
+
"logps/rejected": -324.5120544433594,
|
| 140 |
+
"loss": 0.637,
|
| 141 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 142 |
+
"rewards/chosen": -0.40506935119628906,
|
| 143 |
+
"rewards/margins": 0.2631128430366516,
|
| 144 |
+
"rewards/rejected": -0.6681821942329407,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.20642201834862386,
|
| 149 |
+
"grad_norm": 13.125498153678837,
|
| 150 |
+
"learning_rate": 4.832031033425662e-07,
|
| 151 |
+
"logits/chosen": -1.73556649684906,
|
| 152 |
+
"logits/rejected": -1.575148105621338,
|
| 153 |
+
"logps/chosen": -354.5548400878906,
|
| 154 |
+
"logps/rejected": -374.177978515625,
|
| 155 |
+
"loss": 0.6117,
|
| 156 |
+
"rewards/accuracies": 0.706250011920929,
|
| 157 |
+
"rewards/chosen": -0.5037232041358948,
|
| 158 |
+
"rewards/margins": 0.48656755685806274,
|
| 159 |
+
"rewards/rejected": -0.9902908205986023,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22935779816513763,
|
| 164 |
+
"grad_norm": 13.953968120873695,
|
| 165 |
+
"learning_rate": 4.752422169756047e-07,
|
| 166 |
+
"logits/chosen": -0.5573834180831909,
|
| 167 |
+
"logits/rejected": -0.19994211196899414,
|
| 168 |
+
"logps/chosen": -354.9537048339844,
|
| 169 |
+
"logps/rejected": -379.1808166503906,
|
| 170 |
+
"loss": 0.5874,
|
| 171 |
+
"rewards/accuracies": 0.6875,
|
| 172 |
+
"rewards/chosen": -0.7568774819374084,
|
| 173 |
+
"rewards/margins": 0.5189547538757324,
|
| 174 |
+
"rewards/rejected": -1.2758322954177856,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22935779816513763,
|
| 179 |
+
"eval_logits/chosen": -0.22263701260089874,
|
| 180 |
+
"eval_logits/rejected": 0.3007841110229492,
|
| 181 |
+
"eval_logps/chosen": -382.2076416015625,
|
| 182 |
+
"eval_logps/rejected": -402.1482238769531,
|
| 183 |
+
"eval_loss": 0.5920498371124268,
|
| 184 |
+
"eval_rewards/accuracies": 0.681034505367279,
|
| 185 |
+
"eval_rewards/chosen": -0.9820166230201721,
|
| 186 |
+
"eval_rewards/margins": 0.5829907059669495,
|
| 187 |
+
"eval_rewards/rejected": -1.5650073289871216,
|
| 188 |
+
"eval_runtime": 91.5375,
|
| 189 |
+
"eval_samples_per_second": 19.861,
|
| 190 |
+
"eval_steps_per_second": 0.317,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25229357798165136,
|
| 195 |
+
"grad_norm": 23.915856678856265,
|
| 196 |
+
"learning_rate": 4.658354083558188e-07,
|
| 197 |
+
"logits/chosen": -0.40947216749191284,
|
| 198 |
+
"logits/rejected": 0.038003671914339066,
|
| 199 |
+
"logps/chosen": -379.87451171875,
|
| 200 |
+
"logps/rejected": -438.7657165527344,
|
| 201 |
+
"loss": 0.5763,
|
| 202 |
+
"rewards/accuracies": 0.675000011920929,
|
| 203 |
+
"rewards/chosen": -0.826363742351532,
|
| 204 |
+
"rewards/margins": 0.7436057329177856,
|
| 205 |
+
"rewards/rejected": -1.5699694156646729,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.27522935779816515,
|
| 210 |
+
"grad_norm": 14.304332284980681,
|
| 211 |
+
"learning_rate": 4.550430636492389e-07,
|
| 212 |
+
"logits/chosen": -0.18220779299736023,
|
| 213 |
+
"logits/rejected": 0.7561357021331787,
|
| 214 |
+
"logps/chosen": -424.419677734375,
|
| 215 |
+
"logps/rejected": -439.09320068359375,
|
| 216 |
+
"loss": 0.5937,
|
| 217 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 218 |
+
"rewards/chosen": -1.1400790214538574,
|
| 219 |
+
"rewards/margins": 0.6504500508308411,
|
| 220 |
+
"rewards/rejected": -1.7905290126800537,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.2981651376146789,
|
| 225 |
+
"grad_norm": 15.490141656078896,
|
| 226 |
+
"learning_rate": 4.429344633468004e-07,
|
| 227 |
+
"logits/chosen": 0.37701982259750366,
|
| 228 |
+
"logits/rejected": 1.372661828994751,
|
| 229 |
+
"logps/chosen": -391.7457580566406,
|
| 230 |
+
"logps/rejected": -451.080810546875,
|
| 231 |
+
"loss": 0.5829,
|
| 232 |
+
"rewards/accuracies": 0.71875,
|
| 233 |
+
"rewards/chosen": -1.0409055948257446,
|
| 234 |
+
"rewards/margins": 0.9019233584403992,
|
| 235 |
+
"rewards/rejected": -1.942828893661499,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.3211009174311927,
|
| 240 |
+
"grad_norm": 15.271366347742205,
|
| 241 |
+
"learning_rate": 4.2958733752443187e-07,
|
| 242 |
+
"logits/chosen": -0.13376641273498535,
|
| 243 |
+
"logits/rejected": 1.0010212659835815,
|
| 244 |
+
"logps/chosen": -383.65301513671875,
|
| 245 |
+
"logps/rejected": -419.6103515625,
|
| 246 |
+
"loss": 0.5623,
|
| 247 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 248 |
+
"rewards/chosen": -0.9957970380783081,
|
| 249 |
+
"rewards/margins": 0.7610489130020142,
|
| 250 |
+
"rewards/rejected": -1.7568460702896118,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.3440366972477064,
|
| 255 |
+
"grad_norm": 22.33400136982847,
|
| 256 |
+
"learning_rate": 4.150873668617898e-07,
|
| 257 |
+
"logits/chosen": 0.09550192952156067,
|
| 258 |
+
"logits/rejected": 1.2170594930648804,
|
| 259 |
+
"logps/chosen": -387.51849365234375,
|
| 260 |
+
"logps/rejected": -434.84136962890625,
|
| 261 |
+
"loss": 0.5612,
|
| 262 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 263 |
+
"rewards/chosen": -0.9680271148681641,
|
| 264 |
+
"rewards/margins": 0.772575318813324,
|
| 265 |
+
"rewards/rejected": -1.7406024932861328,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.3440366972477064,
|
| 270 |
+
"eval_logits/chosen": 1.4310089349746704,
|
| 271 |
+
"eval_logits/rejected": 2.314030885696411,
|
| 272 |
+
"eval_logps/chosen": -430.7732238769531,
|
| 273 |
+
"eval_logps/rejected": -482.2997741699219,
|
| 274 |
+
"eval_loss": 0.5694876909255981,
|
| 275 |
+
"eval_rewards/accuracies": 0.6896551847457886,
|
| 276 |
+
"eval_rewards/chosen": -1.4676721096038818,
|
| 277 |
+
"eval_rewards/margins": 0.8988507390022278,
|
| 278 |
+
"eval_rewards/rejected": -2.366523027420044,
|
| 279 |
+
"eval_runtime": 91.6244,
|
| 280 |
+
"eval_samples_per_second": 19.842,
|
| 281 |
+
"eval_steps_per_second": 0.317,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3669724770642202,
|
| 286 |
+
"grad_norm": 23.657529687976723,
|
| 287 |
+
"learning_rate": 3.9952763262280397e-07,
|
| 288 |
+
"logits/chosen": 1.280305027961731,
|
| 289 |
+
"logits/rejected": 2.0924124717712402,
|
| 290 |
+
"logps/chosen": -457.1259765625,
|
| 291 |
+
"logps/rejected": -502.7013244628906,
|
| 292 |
+
"loss": 0.5762,
|
| 293 |
+
"rewards/accuracies": 0.699999988079071,
|
| 294 |
+
"rewards/chosen": -1.6072089672088623,
|
| 295 |
+
"rewards/margins": 0.868707001209259,
|
| 296 |
+
"rewards/rejected": -2.4759163856506348,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38990825688073394,
|
| 301 |
+
"grad_norm": 20.186960313509775,
|
| 302 |
+
"learning_rate": 3.8300801912883414e-07,
|
| 303 |
+
"logits/chosen": -0.41927653551101685,
|
| 304 |
+
"logits/rejected": 0.5089612007141113,
|
| 305 |
+
"logps/chosen": -360.0221252441406,
|
| 306 |
+
"logps/rejected": -393.9572448730469,
|
| 307 |
+
"loss": 0.5498,
|
| 308 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 309 |
+
"rewards/chosen": -0.9409689903259277,
|
| 310 |
+
"rewards/margins": 0.7842710614204407,
|
| 311 |
+
"rewards/rejected": -1.7252399921417236,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.41284403669724773,
|
| 316 |
+
"grad_norm": 19.134949732610515,
|
| 317 |
+
"learning_rate": 3.6563457256020884e-07,
|
| 318 |
+
"logits/chosen": -0.13538230955600739,
|
| 319 |
+
"logits/rejected": 0.8745505213737488,
|
| 320 |
+
"logps/chosen": -365.5416259765625,
|
| 321 |
+
"logps/rejected": -449.64080810546875,
|
| 322 |
+
"loss": 0.5571,
|
| 323 |
+
"rewards/accuracies": 0.706250011920929,
|
| 324 |
+
"rewards/chosen": -1.0540037155151367,
|
| 325 |
+
"rewards/margins": 1.0069358348846436,
|
| 326 |
+
"rewards/rejected": -2.060939311981201,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.43577981651376146,
|
| 331 |
+
"grad_norm": 18.989463695568457,
|
| 332 |
+
"learning_rate": 3.475188202022617e-07,
|
| 333 |
+
"logits/chosen": 0.555556058883667,
|
| 334 |
+
"logits/rejected": 1.7799046039581299,
|
| 335 |
+
"logps/chosen": -379.05133056640625,
|
| 336 |
+
"logps/rejected": -495.5801696777344,
|
| 337 |
+
"loss": 0.5566,
|
| 338 |
+
"rewards/accuracies": 0.762499988079071,
|
| 339 |
+
"rewards/chosen": -1.2038367986679077,
|
| 340 |
+
"rewards/margins": 1.1251088380813599,
|
| 341 |
+
"rewards/rejected": -2.3289456367492676,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45871559633027525,
|
| 346 |
+
"grad_norm": 20.721540306309656,
|
| 347 |
+
"learning_rate": 3.287770545059052e-07,
|
| 348 |
+
"logits/chosen": 0.1556408703327179,
|
| 349 |
+
"logits/rejected": 1.0401029586791992,
|
| 350 |
+
"logps/chosen": -403.26068115234375,
|
| 351 |
+
"logps/rejected": -452.82684326171875,
|
| 352 |
+
"loss": 0.5427,
|
| 353 |
+
"rewards/accuracies": 0.699999988079071,
|
| 354 |
+
"rewards/chosen": -1.3739937543869019,
|
| 355 |
+
"rewards/margins": 0.7915691137313843,
|
| 356 |
+
"rewards/rejected": -2.1655631065368652,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45871559633027525,
|
| 361 |
+
"eval_logits/chosen": -0.3630250096321106,
|
| 362 |
+
"eval_logits/rejected": 0.9222813248634338,
|
| 363 |
+
"eval_logps/chosen": -418.6947021484375,
|
| 364 |
+
"eval_logps/rejected": -471.8922119140625,
|
| 365 |
+
"eval_loss": 0.5523015260696411,
|
| 366 |
+
"eval_rewards/accuracies": 0.7241379022598267,
|
| 367 |
+
"eval_rewards/chosen": -1.346887469291687,
|
| 368 |
+
"eval_rewards/margins": 0.9155599474906921,
|
| 369 |
+
"eval_rewards/rejected": -2.2624475955963135,
|
| 370 |
+
"eval_runtime": 92.3698,
|
| 371 |
+
"eval_samples_per_second": 19.682,
|
| 372 |
+
"eval_steps_per_second": 0.314,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.481651376146789,
|
| 377 |
+
"grad_norm": 17.428734986299386,
|
| 378 |
+
"learning_rate": 3.0952958655864954e-07,
|
| 379 |
+
"logits/chosen": 0.16718950867652893,
|
| 380 |
+
"logits/rejected": 1.1019595861434937,
|
| 381 |
+
"logps/chosen": -430.6678771972656,
|
| 382 |
+
"logps/rejected": -517.311767578125,
|
| 383 |
+
"loss": 0.5458,
|
| 384 |
+
"rewards/accuracies": 0.71875,
|
| 385 |
+
"rewards/chosen": -1.5387192964553833,
|
| 386 |
+
"rewards/margins": 0.787868857383728,
|
| 387 |
+
"rewards/rejected": -2.3265881538391113,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5045871559633027,
|
| 392 |
+
"grad_norm": 18.835466832309084,
|
| 393 |
+
"learning_rate": 2.898999737583448e-07,
|
| 394 |
+
"logits/chosen": -0.19043000042438507,
|
| 395 |
+
"logits/rejected": 1.0126752853393555,
|
| 396 |
+
"logps/chosen": -419.5379943847656,
|
| 397 |
+
"logps/rejected": -484.6988830566406,
|
| 398 |
+
"loss": 0.5528,
|
| 399 |
+
"rewards/accuracies": 0.6875,
|
| 400 |
+
"rewards/chosen": -1.511856198310852,
|
| 401 |
+
"rewards/margins": 0.7948068976402283,
|
| 402 |
+
"rewards/rejected": -2.3066630363464355,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5275229357798165,
|
| 407 |
+
"grad_norm": 23.197020188064432,
|
| 408 |
+
"learning_rate": 2.7001422664752333e-07,
|
| 409 |
+
"logits/chosen": -0.30360710620880127,
|
| 410 |
+
"logits/rejected": 1.0719916820526123,
|
| 411 |
+
"logps/chosen": -386.96087646484375,
|
| 412 |
+
"logps/rejected": -477.74029541015625,
|
| 413 |
+
"loss": 0.5482,
|
| 414 |
+
"rewards/accuracies": 0.762499988079071,
|
| 415 |
+
"rewards/chosen": -1.1279244422912598,
|
| 416 |
+
"rewards/margins": 1.1136093139648438,
|
| 417 |
+
"rewards/rejected": -2.2415337562561035,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5504587155963303,
|
| 422 |
+
"grad_norm": 23.485163773762974,
|
| 423 |
+
"learning_rate": 2.5e-07,
|
| 424 |
+
"logits/chosen": 0.9384336471557617,
|
| 425 |
+
"logits/rejected": 1.5441758632659912,
|
| 426 |
+
"logps/chosen": -413.4400329589844,
|
| 427 |
+
"logps/rejected": -505.8965759277344,
|
| 428 |
+
"loss": 0.5778,
|
| 429 |
+
"rewards/accuracies": 0.6875,
|
| 430 |
+
"rewards/chosen": -1.4366745948791504,
|
| 431 |
+
"rewards/margins": 0.9647194147109985,
|
| 432 |
+
"rewards/rejected": -2.4013938903808594,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.573394495412844,
|
| 437 |
+
"grad_norm": 18.02833446289089,
|
| 438 |
+
"learning_rate": 2.2998577335247667e-07,
|
| 439 |
+
"logits/chosen": 0.6526413559913635,
|
| 440 |
+
"logits/rejected": 1.6761877536773682,
|
| 441 |
+
"logps/chosen": -408.7395324707031,
|
| 442 |
+
"logps/rejected": -484.34417724609375,
|
| 443 |
+
"loss": 0.5474,
|
| 444 |
+
"rewards/accuracies": 0.6875,
|
| 445 |
+
"rewards/chosen": -1.438388705253601,
|
| 446 |
+
"rewards/margins": 0.9599205851554871,
|
| 447 |
+
"rewards/rejected": -2.3983092308044434,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.573394495412844,
|
| 452 |
+
"eval_logits/chosen": -0.44025513529777527,
|
| 453 |
+
"eval_logits/rejected": 0.9070881605148315,
|
| 454 |
+
"eval_logps/chosen": -393.58612060546875,
|
| 455 |
+
"eval_logps/rejected": -449.3501281738281,
|
| 456 |
+
"eval_loss": 0.5429869890213013,
|
| 457 |
+
"eval_rewards/accuracies": 0.6896551847457886,
|
| 458 |
+
"eval_rewards/chosen": -1.0958014726638794,
|
| 459 |
+
"eval_rewards/margins": 0.9412251114845276,
|
| 460 |
+
"eval_rewards/rejected": -2.0370266437530518,
|
| 461 |
+
"eval_runtime": 91.0397,
|
| 462 |
+
"eval_samples_per_second": 19.969,
|
| 463 |
+
"eval_steps_per_second": 0.319,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5963302752293578,
|
| 468 |
+
"grad_norm": 21.769039546192015,
|
| 469 |
+
"learning_rate": 2.1010002624165524e-07,
|
| 470 |
+
"logits/chosen": 0.1161709800362587,
|
| 471 |
+
"logits/rejected": 1.337697982788086,
|
| 472 |
+
"logps/chosen": -419.63995361328125,
|
| 473 |
+
"logps/rejected": -492.03363037109375,
|
| 474 |
+
"loss": 0.5494,
|
| 475 |
+
"rewards/accuracies": 0.8125,
|
| 476 |
+
"rewards/chosen": -1.1713824272155762,
|
| 477 |
+
"rewards/margins": 1.1571643352508545,
|
| 478 |
+
"rewards/rejected": -2.3285470008850098,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6192660550458715,
|
| 483 |
+
"grad_norm": 19.115539205351443,
|
| 484 |
+
"learning_rate": 1.9047041344135043e-07,
|
| 485 |
+
"logits/chosen": 1.2381702661514282,
|
| 486 |
+
"logits/rejected": 2.2026009559631348,
|
| 487 |
+
"logps/chosen": -462.31072998046875,
|
| 488 |
+
"logps/rejected": -520.6209716796875,
|
| 489 |
+
"loss": 0.5714,
|
| 490 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 491 |
+
"rewards/chosen": -1.8605678081512451,
|
| 492 |
+
"rewards/margins": 0.8712261319160461,
|
| 493 |
+
"rewards/rejected": -2.7317938804626465,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.6422018348623854,
|
| 498 |
+
"grad_norm": 20.348758953970556,
|
| 499 |
+
"learning_rate": 1.7122294549409482e-07,
|
| 500 |
+
"logits/chosen": 1.1374260187149048,
|
| 501 |
+
"logits/rejected": 2.329713821411133,
|
| 502 |
+
"logps/chosen": -465.6629333496094,
|
| 503 |
+
"logps/rejected": -563.9349365234375,
|
| 504 |
+
"loss": 0.543,
|
| 505 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 506 |
+
"rewards/chosen": -1.8833774328231812,
|
| 507 |
+
"rewards/margins": 1.0653350353240967,
|
| 508 |
+
"rewards/rejected": -2.948712110519409,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6651376146788991,
|
| 513 |
+
"grad_norm": 18.788678367954812,
|
| 514 |
+
"learning_rate": 1.524811797977383e-07,
|
| 515 |
+
"logits/chosen": -0.17822352051734924,
|
| 516 |
+
"logits/rejected": 0.844377338886261,
|
| 517 |
+
"logps/chosen": -437.96435546875,
|
| 518 |
+
"logps/rejected": -507.1295471191406,
|
| 519 |
+
"loss": 0.542,
|
| 520 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 521 |
+
"rewards/chosen": -1.4787204265594482,
|
| 522 |
+
"rewards/margins": 0.8510695695877075,
|
| 523 |
+
"rewards/rejected": -2.329789876937866,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6880733944954128,
|
| 528 |
+
"grad_norm": 22.907322461677648,
|
| 529 |
+
"learning_rate": 1.3436542743979125e-07,
|
| 530 |
+
"logits/chosen": 0.011055344715714455,
|
| 531 |
+
"logits/rejected": 1.3963500261306763,
|
| 532 |
+
"logps/chosen": -432.23321533203125,
|
| 533 |
+
"logps/rejected": -509.43939208984375,
|
| 534 |
+
"loss": 0.5556,
|
| 535 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 536 |
+
"rewards/chosen": -1.4701194763183594,
|
| 537 |
+
"rewards/margins": 1.124694585800171,
|
| 538 |
+
"rewards/rejected": -2.5948143005371094,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6880733944954128,
|
| 543 |
+
"eval_logits/chosen": -0.1993160843849182,
|
| 544 |
+
"eval_logits/rejected": 1.1949646472930908,
|
| 545 |
+
"eval_logps/chosen": -423.5918884277344,
|
| 546 |
+
"eval_logps/rejected": -484.2666320800781,
|
| 547 |
+
"eval_loss": 0.5403606295585632,
|
| 548 |
+
"eval_rewards/accuracies": 0.7198275923728943,
|
| 549 |
+
"eval_rewards/chosen": -1.3958592414855957,
|
| 550 |
+
"eval_rewards/margins": 0.9903314709663391,
|
| 551 |
+
"eval_rewards/rejected": -2.38619065284729,
|
| 552 |
+
"eval_runtime": 90.6243,
|
| 553 |
+
"eval_samples_per_second": 20.061,
|
| 554 |
+
"eval_steps_per_second": 0.32,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7110091743119266,
|
| 559 |
+
"grad_norm": 27.111269521337316,
|
| 560 |
+
"learning_rate": 1.1699198087116588e-07,
|
| 561 |
+
"logits/chosen": 0.3549458086490631,
|
| 562 |
+
"logits/rejected": 1.506676435470581,
|
| 563 |
+
"logps/chosen": -407.77020263671875,
|
| 564 |
+
"logps/rejected": -494.737548828125,
|
| 565 |
+
"loss": 0.5603,
|
| 566 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 567 |
+
"rewards/chosen": -1.536117434501648,
|
| 568 |
+
"rewards/margins": 0.9887636303901672,
|
| 569 |
+
"rewards/rejected": -2.524880886077881,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7339449541284404,
|
| 574 |
+
"grad_norm": 20.93281634880908,
|
| 575 |
+
"learning_rate": 1.00472367377196e-07,
|
| 576 |
+
"logits/chosen": -0.10006646066904068,
|
| 577 |
+
"logits/rejected": 1.6550134420394897,
|
| 578 |
+
"logps/chosen": -442.34283447265625,
|
| 579 |
+
"logps/rejected": -508.07061767578125,
|
| 580 |
+
"loss": 0.5395,
|
| 581 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 582 |
+
"rewards/chosen": -1.3644827604293823,
|
| 583 |
+
"rewards/margins": 1.2163000106811523,
|
| 584 |
+
"rewards/rejected": -2.580782651901245,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7568807339449541,
|
| 589 |
+
"grad_norm": 22.09617953181825,
|
| 590 |
+
"learning_rate": 8.49126331382102e-08,
|
| 591 |
+
"logits/chosen": 0.12130077183246613,
|
| 592 |
+
"logits/rejected": 1.2904590368270874,
|
| 593 |
+
"logps/chosen": -437.9993591308594,
|
| 594 |
+
"logps/rejected": -525.1572875976562,
|
| 595 |
+
"loss": 0.5394,
|
| 596 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 597 |
+
"rewards/chosen": -1.592714548110962,
|
| 598 |
+
"rewards/margins": 0.9397724270820618,
|
| 599 |
+
"rewards/rejected": -2.5324866771698,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7798165137614679,
|
| 604 |
+
"grad_norm": 21.35703312189058,
|
| 605 |
+
"learning_rate": 7.041266247556812e-08,
|
| 606 |
+
"logits/chosen": 0.1835191398859024,
|
| 607 |
+
"logits/rejected": 1.8147304058074951,
|
| 608 |
+
"logps/chosen": -403.4627380371094,
|
| 609 |
+
"logps/rejected": -524.2403564453125,
|
| 610 |
+
"loss": 0.5415,
|
| 611 |
+
"rewards/accuracies": 0.793749988079071,
|
| 612 |
+
"rewards/chosen": -1.4125099182128906,
|
| 613 |
+
"rewards/margins": 1.2409336566925049,
|
| 614 |
+
"rewards/rejected": -2.6534433364868164,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8027522935779816,
|
| 619 |
+
"grad_norm": 21.04805960149964,
|
| 620 |
+
"learning_rate": 5.706553665319955e-08,
|
| 621 |
+
"logits/chosen": 0.23725076019763947,
|
| 622 |
+
"logits/rejected": 2.2347147464752197,
|
| 623 |
+
"logps/chosen": -436.37939453125,
|
| 624 |
+
"logps/rejected": -537.32373046875,
|
| 625 |
+
"loss": 0.5373,
|
| 626 |
+
"rewards/accuracies": 0.800000011920929,
|
| 627 |
+
"rewards/chosen": -1.4519966840744019,
|
| 628 |
+
"rewards/margins": 1.4744818210601807,
|
| 629 |
+
"rewards/rejected": -2.926478624343872,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8027522935779816,
|
| 634 |
+
"eval_logits/chosen": 0.23955857753753662,
|
| 635 |
+
"eval_logits/rejected": 1.7159334421157837,
|
| 636 |
+
"eval_logps/chosen": -439.8386535644531,
|
| 637 |
+
"eval_logps/rejected": -505.62298583984375,
|
| 638 |
+
"eval_loss": 0.5416247844696045,
|
| 639 |
+
"eval_rewards/accuracies": 0.7284482717514038,
|
| 640 |
+
"eval_rewards/chosen": -1.5583269596099854,
|
| 641 |
+
"eval_rewards/margins": 1.041427493095398,
|
| 642 |
+
"eval_rewards/rejected": -2.599754810333252,
|
| 643 |
+
"eval_runtime": 92.2289,
|
| 644 |
+
"eval_samples_per_second": 19.712,
|
| 645 |
+
"eval_steps_per_second": 0.314,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8256880733944955,
|
| 650 |
+
"grad_norm": 28.15381228175136,
|
| 651 |
+
"learning_rate": 4.4956936350761005e-08,
|
| 652 |
+
"logits/chosen": 0.37452951073646545,
|
| 653 |
+
"logits/rejected": 1.507530689239502,
|
| 654 |
+
"logps/chosen": -437.071044921875,
|
| 655 |
+
"logps/rejected": -533.8405151367188,
|
| 656 |
+
"loss": 0.535,
|
| 657 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 658 |
+
"rewards/chosen": -1.5732779502868652,
|
| 659 |
+
"rewards/margins": 1.106673002243042,
|
| 660 |
+
"rewards/rejected": -2.679950714111328,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8486238532110092,
|
| 665 |
+
"grad_norm": 17.904935938367142,
|
| 666 |
+
"learning_rate": 3.416459164418123e-08,
|
| 667 |
+
"logits/chosen": -0.39112910628318787,
|
| 668 |
+
"logits/rejected": 1.349570631980896,
|
| 669 |
+
"logps/chosen": -471.1695861816406,
|
| 670 |
+
"logps/rejected": -532.53271484375,
|
| 671 |
+
"loss": 0.5285,
|
| 672 |
+
"rewards/accuracies": 0.793749988079071,
|
| 673 |
+
"rewards/chosen": -1.3933590650558472,
|
| 674 |
+
"rewards/margins": 1.198838472366333,
|
| 675 |
+
"rewards/rejected": -2.5921976566314697,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8715596330275229,
|
| 680 |
+
"grad_norm": 21.395515100278608,
|
| 681 |
+
"learning_rate": 2.475778302439524e-08,
|
| 682 |
+
"logits/chosen": -0.16580775380134583,
|
| 683 |
+
"logits/rejected": 1.5711965560913086,
|
| 684 |
+
"logps/chosen": -442.42633056640625,
|
| 685 |
+
"logps/rejected": -471.59808349609375,
|
| 686 |
+
"loss": 0.5345,
|
| 687 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 688 |
+
"rewards/chosen": -1.4915008544921875,
|
| 689 |
+
"rewards/margins": 0.9731476902961731,
|
| 690 |
+
"rewards/rejected": -2.464648723602295,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8944954128440367,
|
| 695 |
+
"grad_norm": 18.765474571368344,
|
| 696 |
+
"learning_rate": 1.6796896657433805e-08,
|
| 697 |
+
"logits/chosen": -0.5060194730758667,
|
| 698 |
+
"logits/rejected": 1.3254286050796509,
|
| 699 |
+
"logps/chosen": -439.9669494628906,
|
| 700 |
+
"logps/rejected": -542.6121826171875,
|
| 701 |
+
"loss": 0.5238,
|
| 702 |
+
"rewards/accuracies": 0.793749988079071,
|
| 703 |
+
"rewards/chosen": -1.3199481964111328,
|
| 704 |
+
"rewards/margins": 1.575811743736267,
|
| 705 |
+
"rewards/rejected": -2.8957600593566895,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9174311926605505,
|
| 710 |
+
"grad_norm": 22.250966254695054,
|
| 711 |
+
"learning_rate": 1.0333036740834855e-08,
|
| 712 |
+
"logits/chosen": 0.07749566435813904,
|
| 713 |
+
"logits/rejected": 1.2516423463821411,
|
| 714 |
+
"logps/chosen": -449.790283203125,
|
| 715 |
+
"logps/rejected": -540.7836303710938,
|
| 716 |
+
"loss": 0.5405,
|
| 717 |
+
"rewards/accuracies": 0.737500011920929,
|
| 718 |
+
"rewards/chosen": -1.5793349742889404,
|
| 719 |
+
"rewards/margins": 0.9962536692619324,
|
| 720 |
+
"rewards/rejected": -2.5755887031555176,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9174311926605505,
|
| 725 |
+
"eval_logits/chosen": -0.2526824176311493,
|
| 726 |
+
"eval_logits/rejected": 1.2749311923980713,
|
| 727 |
+
"eval_logps/chosen": -423.50390625,
|
| 728 |
+
"eval_logps/rejected": -490.2825927734375,
|
| 729 |
+
"eval_loss": 0.5382627844810486,
|
| 730 |
+
"eval_rewards/accuracies": 0.7241379022598267,
|
| 731 |
+
"eval_rewards/chosen": -1.3949792385101318,
|
| 732 |
+
"eval_rewards/margins": 1.0513713359832764,
|
| 733 |
+
"eval_rewards/rejected": -2.446350574493408,
|
| 734 |
+
"eval_runtime": 90.812,
|
| 735 |
+
"eval_samples_per_second": 20.019,
|
| 736 |
+
"eval_steps_per_second": 0.319,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9403669724770642,
|
| 741 |
+
"grad_norm": 20.214960371040412,
|
| 742 |
+
"learning_rate": 5.4076974448211685e-09,
|
| 743 |
+
"logits/chosen": 0.10587388277053833,
|
| 744 |
+
"logits/rejected": 1.2455976009368896,
|
| 745 |
+
"logps/chosen": -443.05621337890625,
|
| 746 |
+
"logps/rejected": -499.88531494140625,
|
| 747 |
+
"loss": 0.551,
|
| 748 |
+
"rewards/accuracies": 0.71875,
|
| 749 |
+
"rewards/chosen": -1.5565433502197266,
|
| 750 |
+
"rewards/margins": 0.8829916715621948,
|
| 751 |
+
"rewards/rejected": -2.4395346641540527,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.963302752293578,
|
| 756 |
+
"grad_norm": 24.19047240295495,
|
| 757 |
+
"learning_rate": 2.052496544188487e-09,
|
| 758 |
+
"logits/chosen": -0.0691867247223854,
|
| 759 |
+
"logits/rejected": 1.6679766178131104,
|
| 760 |
+
"logps/chosen": -455.8050231933594,
|
| 761 |
+
"logps/rejected": -492.05401611328125,
|
| 762 |
+
"loss": 0.5468,
|
| 763 |
+
"rewards/accuracies": 0.737500011920929,
|
| 764 |
+
"rewards/chosen": -1.5099737644195557,
|
| 765 |
+
"rewards/margins": 1.0747706890106201,
|
| 766 |
+
"rewards/rejected": -2.584744691848755,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9862385321100917,
|
| 771 |
+
"grad_norm": 16.600958628723166,
|
| 772 |
+
"learning_rate": 2.889724508297886e-10,
|
| 773 |
+
"logits/chosen": 0.0887439101934433,
|
| 774 |
+
"logits/rejected": 1.3473814725875854,
|
| 775 |
+
"logps/chosen": -406.8893737792969,
|
| 776 |
+
"logps/rejected": -502.24786376953125,
|
| 777 |
+
"loss": 0.5367,
|
| 778 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 779 |
+
"rewards/chosen": -1.4497406482696533,
|
| 780 |
+
"rewards/margins": 1.044012188911438,
|
| 781 |
+
"rewards/rejected": -2.4937527179718018,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 1.0,
|
| 786 |
+
"step": 436,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.5759701597581216,
|
| 789 |
+
"train_runtime": 11541.3114,
|
| 790 |
+
"train_samples_per_second": 4.831,
|
| 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 |
+
}
|