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-low-curriculum
|
| 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-low-curriculum
|
| 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: 0.5229
|
| 21 |
+
- Rewards/chosen: -0.8152
|
| 22 |
+
- Rewards/rejected: -1.5392
|
| 23 |
+
- Rewards/accuracies: 0.7069
|
| 24 |
+
- Rewards/margins: 0.7241
|
| 25 |
+
- Logps/rejected: -399.5724
|
| 26 |
+
- Logps/chosen: -365.5233
|
| 27 |
+
- Logits/rejected: 1.4072
|
| 28 |
+
- Logits/chosen: 0.3892
|
| 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 |
+
| 0.6558 | 0.1147 | 50 | 0.6455 | 0.0044 | -0.0965 | 0.6810 | 0.1009 | -255.3015 | -283.5690 | -2.4863 | -2.5882 |
|
| 66 |
+
| 0.5907 | 0.2294 | 100 | 0.5894 | -0.2321 | -0.5376 | 0.7069 | 0.3055 | -299.4117 | -307.2200 | -2.4655 | -2.5910 |
|
| 67 |
+
| 0.5657 | 0.3440 | 150 | 0.5474 | -0.5168 | -1.0293 | 0.7198 | 0.5125 | -348.5750 | -335.6879 | -0.6546 | -1.0350 |
|
| 68 |
+
| 0.5303 | 0.4587 | 200 | 0.5414 | -1.0659 | -1.7181 | 0.75 | 0.6522 | -417.4532 | -390.5937 | 0.7246 | 0.0707 |
|
| 69 |
+
| 0.5472 | 0.5734 | 250 | 0.5268 | -0.8095 | -1.4718 | 0.7155 | 0.6623 | -392.8294 | -364.9606 | 1.2657 | 0.4213 |
|
| 70 |
+
| 0.5517 | 0.6881 | 300 | 0.5284 | -0.8914 | -1.6145 | 0.7112 | 0.7231 | -407.0940 | -373.1438 | 1.3137 | 0.2994 |
|
| 71 |
+
| 0.4943 | 0.8028 | 350 | 0.5237 | -0.8328 | -1.5668 | 0.7112 | 0.7339 | -402.3227 | -367.2895 | 1.4252 | 0.4044 |
|
| 72 |
+
| 0.5335 | 0.9174 | 400 | 0.5229 | -0.8152 | -1.5392 | 0.7069 | 0.7241 | -399.5724 | -365.5233 | 1.4072 | 0.3892 |
|
| 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": 0.5600318115785581,
|
| 5 |
+
"train_runtime": 11378.3731,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.9,
|
| 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": 0.5600318115785581,
|
| 5 |
+
"train_runtime": 11378.3731,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.9,
|
| 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": 8.800843666556295,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.6664202213287354,
|
| 16 |
+
"logits/rejected": -2.5855507850646973,
|
| 17 |
+
"logps/chosen": -290.6496276855469,
|
| 18 |
+
"logps/rejected": -275.44122314453125,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.39375001192092896,
|
| 21 |
+
"rewards/chosen": 0.00033013062784448266,
|
| 22 |
+
"rewards/margins": 0.00021790717437397689,
|
| 23 |
+
"rewards/rejected": 0.00011222347529837862,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045871559633027525,
|
| 28 |
+
"grad_norm": 8.175869900678,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.6562132835388184,
|
| 31 |
+
"logits/rejected": -2.5835225582122803,
|
| 32 |
+
"logps/chosen": -250.1300048828125,
|
| 33 |
+
"logps/rejected": -232.3563995361328,
|
| 34 |
+
"loss": 0.6921,
|
| 35 |
+
"rewards/accuracies": 0.59375,
|
| 36 |
+
"rewards/chosen": 0.0010141785023733974,
|
| 37 |
+
"rewards/margins": 0.002109699649736285,
|
| 38 |
+
"rewards/rejected": -0.001095521030947566,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06880733944954129,
|
| 43 |
+
"grad_norm": 7.970196521590508,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.6696975231170654,
|
| 46 |
+
"logits/rejected": -2.6478660106658936,
|
| 47 |
+
"logps/chosen": -275.9632263183594,
|
| 48 |
+
"logps/rejected": -266.75457763671875,
|
| 49 |
+
"loss": 0.6876,
|
| 50 |
+
"rewards/accuracies": 0.59375,
|
| 51 |
+
"rewards/chosen": 0.00607306556776166,
|
| 52 |
+
"rewards/margins": 0.009838912636041641,
|
| 53 |
+
"rewards/rejected": -0.003765846835449338,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09174311926605505,
|
| 58 |
+
"grad_norm": 8.67000085322498,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.595798969268799,
|
| 61 |
+
"logits/rejected": -2.5203123092651367,
|
| 62 |
+
"logps/chosen": -286.19683837890625,
|
| 63 |
+
"logps/rejected": -249.7381591796875,
|
| 64 |
+
"loss": 0.6763,
|
| 65 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 66 |
+
"rewards/chosen": 0.028836142271757126,
|
| 67 |
+
"rewards/margins": 0.03639785572886467,
|
| 68 |
+
"rewards/rejected": -0.007561707403510809,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11467889908256881,
|
| 73 |
+
"grad_norm": 13.337487209496269,
|
| 74 |
+
"learning_rate": 4.997110275491701e-07,
|
| 75 |
+
"logits/chosen": -2.536128520965576,
|
| 76 |
+
"logits/rejected": -2.4935240745544434,
|
| 77 |
+
"logps/chosen": -288.22052001953125,
|
| 78 |
+
"logps/rejected": -272.15167236328125,
|
| 79 |
+
"loss": 0.6558,
|
| 80 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 81 |
+
"rewards/chosen": 0.006760952528566122,
|
| 82 |
+
"rewards/margins": 0.0824705958366394,
|
| 83 |
+
"rewards/rejected": -0.07570964097976685,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11467889908256881,
|
| 88 |
+
"eval_logits/chosen": -2.5882346630096436,
|
| 89 |
+
"eval_logits/rejected": -2.4863200187683105,
|
| 90 |
+
"eval_logps/chosen": -283.5689697265625,
|
| 91 |
+
"eval_logps/rejected": -255.30152893066406,
|
| 92 |
+
"eval_loss": 0.6454581022262573,
|
| 93 |
+
"eval_rewards/accuracies": 0.681034505367279,
|
| 94 |
+
"eval_rewards/chosen": 0.00436991173774004,
|
| 95 |
+
"eval_rewards/margins": 0.10091028362512589,
|
| 96 |
+
"eval_rewards/rejected": -0.09654037654399872,
|
| 97 |
+
"eval_runtime": 90.5984,
|
| 98 |
+
"eval_samples_per_second": 20.067,
|
| 99 |
+
"eval_steps_per_second": 0.32,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13761467889908258,
|
| 104 |
+
"grad_norm": 11.707830850446049,
|
| 105 |
+
"learning_rate": 4.979475034558115e-07,
|
| 106 |
+
"logits/chosen": -2.5649328231811523,
|
| 107 |
+
"logits/rejected": -2.482309341430664,
|
| 108 |
+
"logps/chosen": -296.80865478515625,
|
| 109 |
+
"logps/rejected": -277.1481628417969,
|
| 110 |
+
"loss": 0.635,
|
| 111 |
+
"rewards/accuracies": 0.637499988079071,
|
| 112 |
+
"rewards/chosen": -0.026310335844755173,
|
| 113 |
+
"rewards/margins": 0.16292977333068848,
|
| 114 |
+
"rewards/rejected": -0.18924009799957275,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16055045871559634,
|
| 119 |
+
"grad_norm": 15.087943087489696,
|
| 120 |
+
"learning_rate": 4.945923025551788e-07,
|
| 121 |
+
"logits/chosen": -2.5654821395874023,
|
| 122 |
+
"logits/rejected": -2.473752498626709,
|
| 123 |
+
"logps/chosen": -302.06201171875,
|
| 124 |
+
"logps/rejected": -300.6982421875,
|
| 125 |
+
"loss": 0.6205,
|
| 126 |
+
"rewards/accuracies": 0.699999988079071,
|
| 127 |
+
"rewards/chosen": -0.1509528011083603,
|
| 128 |
+
"rewards/margins": 0.2617207169532776,
|
| 129 |
+
"rewards/rejected": -0.4126734733581543,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.1834862385321101,
|
| 134 |
+
"grad_norm": 12.495701165097614,
|
| 135 |
+
"learning_rate": 4.896669632591651e-07,
|
| 136 |
+
"logits/chosen": -2.513960361480713,
|
| 137 |
+
"logits/rejected": -2.4418704509735107,
|
| 138 |
+
"logps/chosen": -310.827392578125,
|
| 139 |
+
"logps/rejected": -304.6408996582031,
|
| 140 |
+
"loss": 0.6186,
|
| 141 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 142 |
+
"rewards/chosen": -0.3509863018989563,
|
| 143 |
+
"rewards/margins": 0.16184711456298828,
|
| 144 |
+
"rewards/rejected": -0.5128334760665894,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.20642201834862386,
|
| 149 |
+
"grad_norm": 18.19180173568921,
|
| 150 |
+
"learning_rate": 4.832031033425662e-07,
|
| 151 |
+
"logits/chosen": -2.5777266025543213,
|
| 152 |
+
"logits/rejected": -2.5101935863494873,
|
| 153 |
+
"logps/chosen": -290.23126220703125,
|
| 154 |
+
"logps/rejected": -271.0259704589844,
|
| 155 |
+
"loss": 0.5969,
|
| 156 |
+
"rewards/accuracies": 0.706250011920929,
|
| 157 |
+
"rewards/chosen": -0.19850726425647736,
|
| 158 |
+
"rewards/margins": 0.29140809178352356,
|
| 159 |
+
"rewards/rejected": -0.4899153709411621,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22935779816513763,
|
| 164 |
+
"grad_norm": 20.909839179480045,
|
| 165 |
+
"learning_rate": 4.752422169756047e-07,
|
| 166 |
+
"logits/chosen": -2.548067808151245,
|
| 167 |
+
"logits/rejected": -2.48946475982666,
|
| 168 |
+
"logps/chosen": -313.51727294921875,
|
| 169 |
+
"logps/rejected": -314.1280212402344,
|
| 170 |
+
"loss": 0.5907,
|
| 171 |
+
"rewards/accuracies": 0.668749988079071,
|
| 172 |
+
"rewards/chosen": -0.2900985777378082,
|
| 173 |
+
"rewards/margins": 0.3498932421207428,
|
| 174 |
+
"rewards/rejected": -0.6399917602539062,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22935779816513763,
|
| 179 |
+
"eval_logits/chosen": -2.590954065322876,
|
| 180 |
+
"eval_logits/rejected": -2.465471029281616,
|
| 181 |
+
"eval_logps/chosen": -307.22003173828125,
|
| 182 |
+
"eval_logps/rejected": -299.41168212890625,
|
| 183 |
+
"eval_loss": 0.5893865823745728,
|
| 184 |
+
"eval_rewards/accuracies": 0.7068965435028076,
|
| 185 |
+
"eval_rewards/chosen": -0.2321406602859497,
|
| 186 |
+
"eval_rewards/margins": 0.30550122261047363,
|
| 187 |
+
"eval_rewards/rejected": -0.5376418828964233,
|
| 188 |
+
"eval_runtime": 91.8942,
|
| 189 |
+
"eval_samples_per_second": 19.784,
|
| 190 |
+
"eval_steps_per_second": 0.316,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25229357798165136,
|
| 195 |
+
"grad_norm": 21.128492078002974,
|
| 196 |
+
"learning_rate": 4.658354083558188e-07,
|
| 197 |
+
"logits/chosen": -2.2998366355895996,
|
| 198 |
+
"logits/rejected": -2.1539642810821533,
|
| 199 |
+
"logps/chosen": -310.5582580566406,
|
| 200 |
+
"logps/rejected": -343.615234375,
|
| 201 |
+
"loss": 0.58,
|
| 202 |
+
"rewards/accuracies": 0.699999988079071,
|
| 203 |
+
"rewards/chosen": -0.4958949089050293,
|
| 204 |
+
"rewards/margins": 0.41762369871139526,
|
| 205 |
+
"rewards/rejected": -0.9135186076164246,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.27522935779816515,
|
| 210 |
+
"grad_norm": 22.76420803543292,
|
| 211 |
+
"learning_rate": 4.550430636492389e-07,
|
| 212 |
+
"logits/chosen": -1.5430848598480225,
|
| 213 |
+
"logits/rejected": -1.318800926208496,
|
| 214 |
+
"logps/chosen": -358.74884033203125,
|
| 215 |
+
"logps/rejected": -403.6099548339844,
|
| 216 |
+
"loss": 0.576,
|
| 217 |
+
"rewards/accuracies": 0.706250011920929,
|
| 218 |
+
"rewards/chosen": -0.6421557664871216,
|
| 219 |
+
"rewards/margins": 0.5780781507492065,
|
| 220 |
+
"rewards/rejected": -1.2202337980270386,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.2981651376146789,
|
| 225 |
+
"grad_norm": 18.25021654810524,
|
| 226 |
+
"learning_rate": 4.429344633468004e-07,
|
| 227 |
+
"logits/chosen": -1.3487799167633057,
|
| 228 |
+
"logits/rejected": -1.0795713663101196,
|
| 229 |
+
"logps/chosen": -370.9684143066406,
|
| 230 |
+
"logps/rejected": -413.46026611328125,
|
| 231 |
+
"loss": 0.5503,
|
| 232 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 233 |
+
"rewards/chosen": -0.7266704440116882,
|
| 234 |
+
"rewards/margins": 0.6452904939651489,
|
| 235 |
+
"rewards/rejected": -1.371960997581482,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.3211009174311927,
|
| 240 |
+
"grad_norm": 26.911406799310377,
|
| 241 |
+
"learning_rate": 4.2958733752443187e-07,
|
| 242 |
+
"logits/chosen": -1.0087560415267944,
|
| 243 |
+
"logits/rejected": -0.8604623079299927,
|
| 244 |
+
"logps/chosen": -331.37322998046875,
|
| 245 |
+
"logps/rejected": -340.6692810058594,
|
| 246 |
+
"loss": 0.5608,
|
| 247 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 248 |
+
"rewards/chosen": -0.7133186459541321,
|
| 249 |
+
"rewards/margins": 0.42135268449783325,
|
| 250 |
+
"rewards/rejected": -1.1346712112426758,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.3440366972477064,
|
| 255 |
+
"grad_norm": 21.7192249273613,
|
| 256 |
+
"learning_rate": 4.150873668617898e-07,
|
| 257 |
+
"logits/chosen": -1.0606236457824707,
|
| 258 |
+
"logits/rejected": -0.8972805738449097,
|
| 259 |
+
"logps/chosen": -345.69580078125,
|
| 260 |
+
"logps/rejected": -376.9483947753906,
|
| 261 |
+
"loss": 0.5657,
|
| 262 |
+
"rewards/accuracies": 0.706250011920929,
|
| 263 |
+
"rewards/chosen": -0.6985127329826355,
|
| 264 |
+
"rewards/margins": 0.4048144221305847,
|
| 265 |
+
"rewards/rejected": -1.1033271551132202,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.3440366972477064,
|
| 270 |
+
"eval_logits/chosen": -1.0350327491760254,
|
| 271 |
+
"eval_logits/rejected": -0.6546276807785034,
|
| 272 |
+
"eval_logps/chosen": -335.6878662109375,
|
| 273 |
+
"eval_logps/rejected": -348.57501220703125,
|
| 274 |
+
"eval_loss": 0.5474189519882202,
|
| 275 |
+
"eval_rewards/accuracies": 0.7198275923728943,
|
| 276 |
+
"eval_rewards/chosen": -0.5168190598487854,
|
| 277 |
+
"eval_rewards/margins": 0.5124561190605164,
|
| 278 |
+
"eval_rewards/rejected": -1.0292751789093018,
|
| 279 |
+
"eval_runtime": 90.811,
|
| 280 |
+
"eval_samples_per_second": 20.02,
|
| 281 |
+
"eval_steps_per_second": 0.319,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3669724770642202,
|
| 286 |
+
"grad_norm": 20.42265728106707,
|
| 287 |
+
"learning_rate": 3.9952763262280397e-07,
|
| 288 |
+
"logits/chosen": -0.5147255063056946,
|
| 289 |
+
"logits/rejected": -0.3840788006782532,
|
| 290 |
+
"logps/chosen": -278.23956298828125,
|
| 291 |
+
"logps/rejected": -349.45599365234375,
|
| 292 |
+
"loss": 0.557,
|
| 293 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 294 |
+
"rewards/chosen": -0.5052815675735474,
|
| 295 |
+
"rewards/margins": 0.5687552690505981,
|
| 296 |
+
"rewards/rejected": -1.0740368366241455,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38990825688073394,
|
| 301 |
+
"grad_norm": 27.546124980926876,
|
| 302 |
+
"learning_rate": 3.8300801912883414e-07,
|
| 303 |
+
"logits/chosen": -0.2255803644657135,
|
| 304 |
+
"logits/rejected": 0.25452661514282227,
|
| 305 |
+
"logps/chosen": -356.0308837890625,
|
| 306 |
+
"logps/rejected": -378.4934997558594,
|
| 307 |
+
"loss": 0.5436,
|
| 308 |
+
"rewards/accuracies": 0.6875,
|
| 309 |
+
"rewards/chosen": -0.6629082560539246,
|
| 310 |
+
"rewards/margins": 0.6058169007301331,
|
| 311 |
+
"rewards/rejected": -1.268725037574768,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.41284403669724773,
|
| 316 |
+
"grad_norm": 26.686484538643338,
|
| 317 |
+
"learning_rate": 3.6563457256020884e-07,
|
| 318 |
+
"logits/chosen": -0.08900181949138641,
|
| 319 |
+
"logits/rejected": 0.5574623942375183,
|
| 320 |
+
"logps/chosen": -388.6288146972656,
|
| 321 |
+
"logps/rejected": -392.36724853515625,
|
| 322 |
+
"loss": 0.5357,
|
| 323 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 324 |
+
"rewards/chosen": -0.8669939041137695,
|
| 325 |
+
"rewards/margins": 0.6011493802070618,
|
| 326 |
+
"rewards/rejected": -1.4681432247161865,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.43577981651376146,
|
| 331 |
+
"grad_norm": 26.88225056627165,
|
| 332 |
+
"learning_rate": 3.475188202022617e-07,
|
| 333 |
+
"logits/chosen": -0.923051655292511,
|
| 334 |
+
"logits/rejected": -0.5179102420806885,
|
| 335 |
+
"logps/chosen": -328.10980224609375,
|
| 336 |
+
"logps/rejected": -390.62261962890625,
|
| 337 |
+
"loss": 0.5355,
|
| 338 |
+
"rewards/accuracies": 0.762499988079071,
|
| 339 |
+
"rewards/chosen": -0.5608564615249634,
|
| 340 |
+
"rewards/margins": 0.6695758104324341,
|
| 341 |
+
"rewards/rejected": -1.2304322719573975,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45871559633027525,
|
| 346 |
+
"grad_norm": 28.83910063760549,
|
| 347 |
+
"learning_rate": 3.287770545059052e-07,
|
| 348 |
+
"logits/chosen": 0.1071801632642746,
|
| 349 |
+
"logits/rejected": 0.8112713098526001,
|
| 350 |
+
"logps/chosen": -398.0206604003906,
|
| 351 |
+
"logps/rejected": -412.6197814941406,
|
| 352 |
+
"loss": 0.5303,
|
| 353 |
+
"rewards/accuracies": 0.762499988079071,
|
| 354 |
+
"rewards/chosen": -1.0966007709503174,
|
| 355 |
+
"rewards/margins": 0.6390595436096191,
|
| 356 |
+
"rewards/rejected": -1.735660195350647,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45871559633027525,
|
| 361 |
+
"eval_logits/chosen": 0.07066183537244797,
|
| 362 |
+
"eval_logits/rejected": 0.7245904207229614,
|
| 363 |
+
"eval_logps/chosen": -390.5937194824219,
|
| 364 |
+
"eval_logps/rejected": -417.4531555175781,
|
| 365 |
+
"eval_loss": 0.5413529872894287,
|
| 366 |
+
"eval_rewards/accuracies": 0.75,
|
| 367 |
+
"eval_rewards/chosen": -1.0658774375915527,
|
| 368 |
+
"eval_rewards/margins": 0.6521791815757751,
|
| 369 |
+
"eval_rewards/rejected": -1.7180566787719727,
|
| 370 |
+
"eval_runtime": 91.7218,
|
| 371 |
+
"eval_samples_per_second": 19.821,
|
| 372 |
+
"eval_steps_per_second": 0.316,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.481651376146789,
|
| 377 |
+
"grad_norm": 22.60643364808249,
|
| 378 |
+
"learning_rate": 3.0952958655864954e-07,
|
| 379 |
+
"logits/chosen": -0.1810363531112671,
|
| 380 |
+
"logits/rejected": 0.5873329043388367,
|
| 381 |
+
"logps/chosen": -370.56695556640625,
|
| 382 |
+
"logps/rejected": -393.52349853515625,
|
| 383 |
+
"loss": 0.52,
|
| 384 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 385 |
+
"rewards/chosen": -0.848964512348175,
|
| 386 |
+
"rewards/margins": 0.7796515226364136,
|
| 387 |
+
"rewards/rejected": -1.6286159753799438,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5045871559633027,
|
| 392 |
+
"grad_norm": 25.261348780053293,
|
| 393 |
+
"learning_rate": 2.898999737583448e-07,
|
| 394 |
+
"logits/chosen": 0.12911781668663025,
|
| 395 |
+
"logits/rejected": 0.8453273773193359,
|
| 396 |
+
"logps/chosen": -356.42694091796875,
|
| 397 |
+
"logps/rejected": -401.22625732421875,
|
| 398 |
+
"loss": 0.5152,
|
| 399 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 400 |
+
"rewards/chosen": -0.7875608801841736,
|
| 401 |
+
"rewards/margins": 0.8164966702461243,
|
| 402 |
+
"rewards/rejected": -1.6040576696395874,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5275229357798165,
|
| 407 |
+
"grad_norm": 27.60361008184946,
|
| 408 |
+
"learning_rate": 2.7001422664752333e-07,
|
| 409 |
+
"logits/chosen": 0.1694536954164505,
|
| 410 |
+
"logits/rejected": 0.745651125907898,
|
| 411 |
+
"logps/chosen": -374.00701904296875,
|
| 412 |
+
"logps/rejected": -404.31109619140625,
|
| 413 |
+
"loss": 0.5324,
|
| 414 |
+
"rewards/accuracies": 0.731249988079071,
|
| 415 |
+
"rewards/chosen": -0.8075408935546875,
|
| 416 |
+
"rewards/margins": 0.6526550054550171,
|
| 417 |
+
"rewards/rejected": -1.4601958990097046,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5504587155963303,
|
| 422 |
+
"grad_norm": 26.900360490730417,
|
| 423 |
+
"learning_rate": 2.5e-07,
|
| 424 |
+
"logits/chosen": 0.5097543001174927,
|
| 425 |
+
"logits/rejected": 1.0124820470809937,
|
| 426 |
+
"logps/chosen": -356.8577575683594,
|
| 427 |
+
"logps/rejected": -414.31085205078125,
|
| 428 |
+
"loss": 0.5326,
|
| 429 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 430 |
+
"rewards/chosen": -0.8185293078422546,
|
| 431 |
+
"rewards/margins": 0.7143917083740234,
|
| 432 |
+
"rewards/rejected": -1.5329210758209229,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.573394495412844,
|
| 437 |
+
"grad_norm": 21.852049902948178,
|
| 438 |
+
"learning_rate": 2.2998577335247667e-07,
|
| 439 |
+
"logits/chosen": 0.41708582639694214,
|
| 440 |
+
"logits/rejected": 1.3239974975585938,
|
| 441 |
+
"logps/chosen": -396.1962890625,
|
| 442 |
+
"logps/rejected": -424.10107421875,
|
| 443 |
+
"loss": 0.5472,
|
| 444 |
+
"rewards/accuracies": 0.699999988079071,
|
| 445 |
+
"rewards/chosen": -0.8911197781562805,
|
| 446 |
+
"rewards/margins": 0.6929961442947388,
|
| 447 |
+
"rewards/rejected": -1.5841158628463745,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.573394495412844,
|
| 452 |
+
"eval_logits/chosen": 0.42131370306015015,
|
| 453 |
+
"eval_logits/rejected": 1.265686273574829,
|
| 454 |
+
"eval_logps/chosen": -364.9606018066406,
|
| 455 |
+
"eval_logps/rejected": -392.8293762207031,
|
| 456 |
+
"eval_loss": 0.5268200635910034,
|
| 457 |
+
"eval_rewards/accuracies": 0.7155172228813171,
|
| 458 |
+
"eval_rewards/chosen": -0.8095463514328003,
|
| 459 |
+
"eval_rewards/margins": 0.6622725129127502,
|
| 460 |
+
"eval_rewards/rejected": -1.4718190431594849,
|
| 461 |
+
"eval_runtime": 91.4885,
|
| 462 |
+
"eval_samples_per_second": 19.871,
|
| 463 |
+
"eval_steps_per_second": 0.317,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5963302752293578,
|
| 468 |
+
"grad_norm": 23.156708624000842,
|
| 469 |
+
"learning_rate": 2.1010002624165524e-07,
|
| 470 |
+
"logits/chosen": 0.08591889590024948,
|
| 471 |
+
"logits/rejected": 0.6472758054733276,
|
| 472 |
+
"logps/chosen": -366.85699462890625,
|
| 473 |
+
"logps/rejected": -412.9854431152344,
|
| 474 |
+
"loss": 0.5217,
|
| 475 |
+
"rewards/accuracies": 0.737500011920929,
|
| 476 |
+
"rewards/chosen": -0.7147234678268433,
|
| 477 |
+
"rewards/margins": 0.6098002195358276,
|
| 478 |
+
"rewards/rejected": -1.324523687362671,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6192660550458715,
|
| 483 |
+
"grad_norm": 26.15521858032112,
|
| 484 |
+
"learning_rate": 1.9047041344135043e-07,
|
| 485 |
+
"logits/chosen": 0.2586960792541504,
|
| 486 |
+
"logits/rejected": 1.1518855094909668,
|
| 487 |
+
"logps/chosen": -362.0467224121094,
|
| 488 |
+
"logps/rejected": -416.77716064453125,
|
| 489 |
+
"loss": 0.5074,
|
| 490 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 491 |
+
"rewards/chosen": -0.6773337125778198,
|
| 492 |
+
"rewards/margins": 0.9160529375076294,
|
| 493 |
+
"rewards/rejected": -1.5933866500854492,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.6422018348623854,
|
| 498 |
+
"grad_norm": 22.25288149573626,
|
| 499 |
+
"learning_rate": 1.7122294549409482e-07,
|
| 500 |
+
"logits/chosen": 0.54390949010849,
|
| 501 |
+
"logits/rejected": 1.2391692399978638,
|
| 502 |
+
"logps/chosen": -429.69769287109375,
|
| 503 |
+
"logps/rejected": -436.3062438964844,
|
| 504 |
+
"loss": 0.5485,
|
| 505 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 506 |
+
"rewards/chosen": -1.0754292011260986,
|
| 507 |
+
"rewards/margins": 0.619503915309906,
|
| 508 |
+
"rewards/rejected": -1.6949331760406494,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6651376146788991,
|
| 513 |
+
"grad_norm": 23.40120757809907,
|
| 514 |
+
"learning_rate": 1.524811797977383e-07,
|
| 515 |
+
"logits/chosen": 0.3535264730453491,
|
| 516 |
+
"logits/rejected": 1.0907752513885498,
|
| 517 |
+
"logps/chosen": -380.4679260253906,
|
| 518 |
+
"logps/rejected": -438.02178955078125,
|
| 519 |
+
"loss": 0.5205,
|
| 520 |
+
"rewards/accuracies": 0.731249988079071,
|
| 521 |
+
"rewards/chosen": -0.9396567344665527,
|
| 522 |
+
"rewards/margins": 0.6416970491409302,
|
| 523 |
+
"rewards/rejected": -1.581353783607483,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6880733944954128,
|
| 528 |
+
"grad_norm": 28.751131516456226,
|
| 529 |
+
"learning_rate": 1.3436542743979125e-07,
|
| 530 |
+
"logits/chosen": 0.7159294486045837,
|
| 531 |
+
"logits/rejected": 1.3497685194015503,
|
| 532 |
+
"logps/chosen": -370.821533203125,
|
| 533 |
+
"logps/rejected": -406.04010009765625,
|
| 534 |
+
"loss": 0.5517,
|
| 535 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 536 |
+
"rewards/chosen": -1.1309765577316284,
|
| 537 |
+
"rewards/margins": 0.6069232821464539,
|
| 538 |
+
"rewards/rejected": -1.7378997802734375,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6880733944954128,
|
| 543 |
+
"eval_logits/chosen": 0.29943859577178955,
|
| 544 |
+
"eval_logits/rejected": 1.3137139081954956,
|
| 545 |
+
"eval_logps/chosen": -373.143798828125,
|
| 546 |
+
"eval_logps/rejected": -407.0939636230469,
|
| 547 |
+
"eval_loss": 0.5283679962158203,
|
| 548 |
+
"eval_rewards/accuracies": 0.7112069129943848,
|
| 549 |
+
"eval_rewards/chosen": -0.8913781642913818,
|
| 550 |
+
"eval_rewards/margins": 0.723086416721344,
|
| 551 |
+
"eval_rewards/rejected": -1.6144647598266602,
|
| 552 |
+
"eval_runtime": 91.561,
|
| 553 |
+
"eval_samples_per_second": 19.856,
|
| 554 |
+
"eval_steps_per_second": 0.317,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7110091743119266,
|
| 559 |
+
"grad_norm": 31.58697998681577,
|
| 560 |
+
"learning_rate": 1.1699198087116588e-07,
|
| 561 |
+
"logits/chosen": 0.399633526802063,
|
| 562 |
+
"logits/rejected": 1.2564690113067627,
|
| 563 |
+
"logps/chosen": -335.18218994140625,
|
| 564 |
+
"logps/rejected": -429.2320861816406,
|
| 565 |
+
"loss": 0.5096,
|
| 566 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 567 |
+
"rewards/chosen": -0.8466620445251465,
|
| 568 |
+
"rewards/margins": 0.8323869705200195,
|
| 569 |
+
"rewards/rejected": -1.6790491342544556,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7339449541284404,
|
| 574 |
+
"grad_norm": 25.23384166425554,
|
| 575 |
+
"learning_rate": 1.00472367377196e-07,
|
| 576 |
+
"logits/chosen": 0.13118520379066467,
|
| 577 |
+
"logits/rejected": 0.912071704864502,
|
| 578 |
+
"logps/chosen": -374.3765869140625,
|
| 579 |
+
"logps/rejected": -386.69384765625,
|
| 580 |
+
"loss": 0.5351,
|
| 581 |
+
"rewards/accuracies": 0.71875,
|
| 582 |
+
"rewards/chosen": -0.8689004182815552,
|
| 583 |
+
"rewards/margins": 0.6572145223617554,
|
| 584 |
+
"rewards/rejected": -1.5261149406433105,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7568807339449541,
|
| 589 |
+
"grad_norm": 23.043674301052096,
|
| 590 |
+
"learning_rate": 8.49126331382102e-08,
|
| 591 |
+
"logits/chosen": 0.2135738581418991,
|
| 592 |
+
"logits/rejected": 1.1428295373916626,
|
| 593 |
+
"logps/chosen": -332.0506286621094,
|
| 594 |
+
"logps/rejected": -378.5418701171875,
|
| 595 |
+
"loss": 0.5351,
|
| 596 |
+
"rewards/accuracies": 0.75,
|
| 597 |
+
"rewards/chosen": -0.7963577508926392,
|
| 598 |
+
"rewards/margins": 0.6999517679214478,
|
| 599 |
+
"rewards/rejected": -1.4963096380233765,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7798165137614679,
|
| 604 |
+
"grad_norm": 22.897014093425614,
|
| 605 |
+
"learning_rate": 7.041266247556812e-08,
|
| 606 |
+
"logits/chosen": 0.228415846824646,
|
| 607 |
+
"logits/rejected": 1.0023858547210693,
|
| 608 |
+
"logps/chosen": -356.47955322265625,
|
| 609 |
+
"logps/rejected": -407.20928955078125,
|
| 610 |
+
"loss": 0.5292,
|
| 611 |
+
"rewards/accuracies": 0.71875,
|
| 612 |
+
"rewards/chosen": -0.906414806842804,
|
| 613 |
+
"rewards/margins": 0.5302497744560242,
|
| 614 |
+
"rewards/rejected": -1.4366645812988281,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8027522935779816,
|
| 619 |
+
"grad_norm": 24.081238898664065,
|
| 620 |
+
"learning_rate": 5.706553665319955e-08,
|
| 621 |
+
"logits/chosen": 0.30141282081604004,
|
| 622 |
+
"logits/rejected": 1.139885663986206,
|
| 623 |
+
"logps/chosen": -382.5127868652344,
|
| 624 |
+
"logps/rejected": -426.60723876953125,
|
| 625 |
+
"loss": 0.4943,
|
| 626 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 627 |
+
"rewards/chosen": -0.8278465270996094,
|
| 628 |
+
"rewards/margins": 0.7722162008285522,
|
| 629 |
+
"rewards/rejected": -1.6000627279281616,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8027522935779816,
|
| 634 |
+
"eval_logits/chosen": 0.40436258912086487,
|
| 635 |
+
"eval_logits/rejected": 1.4252060651779175,
|
| 636 |
+
"eval_logps/chosen": -367.28948974609375,
|
| 637 |
+
"eval_logps/rejected": -402.3227233886719,
|
| 638 |
+
"eval_loss": 0.5236544609069824,
|
| 639 |
+
"eval_rewards/accuracies": 0.7112069129943848,
|
| 640 |
+
"eval_rewards/chosen": -0.8328355550765991,
|
| 641 |
+
"eval_rewards/margins": 0.7339165806770325,
|
| 642 |
+
"eval_rewards/rejected": -1.5667520761489868,
|
| 643 |
+
"eval_runtime": 91.5336,
|
| 644 |
+
"eval_samples_per_second": 19.862,
|
| 645 |
+
"eval_steps_per_second": 0.317,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8256880733944955,
|
| 650 |
+
"grad_norm": 25.9302171157121,
|
| 651 |
+
"learning_rate": 4.4956936350761005e-08,
|
| 652 |
+
"logits/chosen": 0.17887906730175018,
|
| 653 |
+
"logits/rejected": 1.2261936664581299,
|
| 654 |
+
"logps/chosen": -403.9292297363281,
|
| 655 |
+
"logps/rejected": -456.3939514160156,
|
| 656 |
+
"loss": 0.5292,
|
| 657 |
+
"rewards/accuracies": 0.699999988079071,
|
| 658 |
+
"rewards/chosen": -0.9248501062393188,
|
| 659 |
+
"rewards/margins": 0.827883243560791,
|
| 660 |
+
"rewards/rejected": -1.7527334690093994,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8486238532110092,
|
| 665 |
+
"grad_norm": 25.758246129198,
|
| 666 |
+
"learning_rate": 3.416459164418123e-08,
|
| 667 |
+
"logits/chosen": 0.7486616373062134,
|
| 668 |
+
"logits/rejected": 1.5815865993499756,
|
| 669 |
+
"logps/chosen": -384.26458740234375,
|
| 670 |
+
"logps/rejected": -433.817626953125,
|
| 671 |
+
"loss": 0.5355,
|
| 672 |
+
"rewards/accuracies": 0.737500011920929,
|
| 673 |
+
"rewards/chosen": -0.9148642420768738,
|
| 674 |
+
"rewards/margins": 0.8338562846183777,
|
| 675 |
+
"rewards/rejected": -1.7487205266952515,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8715596330275229,
|
| 680 |
+
"grad_norm": 27.707849331024978,
|
| 681 |
+
"learning_rate": 2.475778302439524e-08,
|
| 682 |
+
"logits/chosen": 0.8750450015068054,
|
| 683 |
+
"logits/rejected": 1.3391921520233154,
|
| 684 |
+
"logps/chosen": -381.8013610839844,
|
| 685 |
+
"logps/rejected": -423.6830139160156,
|
| 686 |
+
"loss": 0.5201,
|
| 687 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 688 |
+
"rewards/chosen": -0.9749946594238281,
|
| 689 |
+
"rewards/margins": 0.6640049815177917,
|
| 690 |
+
"rewards/rejected": -1.638999581336975,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8944954128440367,
|
| 695 |
+
"grad_norm": 26.946369171438246,
|
| 696 |
+
"learning_rate": 1.6796896657433805e-08,
|
| 697 |
+
"logits/chosen": 0.8233477473258972,
|
| 698 |
+
"logits/rejected": 1.701456069946289,
|
| 699 |
+
"logps/chosen": -368.91595458984375,
|
| 700 |
+
"logps/rejected": -411.06951904296875,
|
| 701 |
+
"loss": 0.5176,
|
| 702 |
+
"rewards/accuracies": 0.737500011920929,
|
| 703 |
+
"rewards/chosen": -0.964392364025116,
|
| 704 |
+
"rewards/margins": 0.733443021774292,
|
| 705 |
+
"rewards/rejected": -1.6978353261947632,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9174311926605505,
|
| 710 |
+
"grad_norm": 28.888256646435288,
|
| 711 |
+
"learning_rate": 1.0333036740834855e-08,
|
| 712 |
+
"logits/chosen": 0.8198953866958618,
|
| 713 |
+
"logits/rejected": 1.1675937175750732,
|
| 714 |
+
"logps/chosen": -357.7641296386719,
|
| 715 |
+
"logps/rejected": -417.57720947265625,
|
| 716 |
+
"loss": 0.5335,
|
| 717 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 718 |
+
"rewards/chosen": -0.8717674016952515,
|
| 719 |
+
"rewards/margins": 0.6356672644615173,
|
| 720 |
+
"rewards/rejected": -1.5074348449707031,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9174311926605505,
|
| 725 |
+
"eval_logits/chosen": 0.3891567885875702,
|
| 726 |
+
"eval_logits/rejected": 1.407207727432251,
|
| 727 |
+
"eval_logps/chosen": -365.5233154296875,
|
| 728 |
+
"eval_logps/rejected": -399.57244873046875,
|
| 729 |
+
"eval_loss": 0.5228918790817261,
|
| 730 |
+
"eval_rewards/accuracies": 0.7068965435028076,
|
| 731 |
+
"eval_rewards/chosen": -0.8151733875274658,
|
| 732 |
+
"eval_rewards/margins": 0.7240758538246155,
|
| 733 |
+
"eval_rewards/rejected": -1.5392491817474365,
|
| 734 |
+
"eval_runtime": 90.7936,
|
| 735 |
+
"eval_samples_per_second": 20.023,
|
| 736 |
+
"eval_steps_per_second": 0.319,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9403669724770642,
|
| 741 |
+
"grad_norm": 26.17469084665507,
|
| 742 |
+
"learning_rate": 5.4076974448211685e-09,
|
| 743 |
+
"logits/chosen": 0.33993715047836304,
|
| 744 |
+
"logits/rejected": 1.2628642320632935,
|
| 745 |
+
"logps/chosen": -358.8694152832031,
|
| 746 |
+
"logps/rejected": -421.8302307128906,
|
| 747 |
+
"loss": 0.536,
|
| 748 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 749 |
+
"rewards/chosen": -0.8047143220901489,
|
| 750 |
+
"rewards/margins": 0.8802136182785034,
|
| 751 |
+
"rewards/rejected": -1.6849279403686523,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.963302752293578,
|
| 756 |
+
"grad_norm": 21.435784540966043,
|
| 757 |
+
"learning_rate": 2.052496544188487e-09,
|
| 758 |
+
"logits/chosen": 0.5830007791519165,
|
| 759 |
+
"logits/rejected": 1.3784797191619873,
|
| 760 |
+
"logps/chosen": -336.01025390625,
|
| 761 |
+
"logps/rejected": -399.4895324707031,
|
| 762 |
+
"loss": 0.5098,
|
| 763 |
+
"rewards/accuracies": 0.75,
|
| 764 |
+
"rewards/chosen": -0.856100857257843,
|
| 765 |
+
"rewards/margins": 0.8189223408699036,
|
| 766 |
+
"rewards/rejected": -1.675023078918457,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9862385321100917,
|
| 771 |
+
"grad_norm": 33.824936700213314,
|
| 772 |
+
"learning_rate": 2.889724508297886e-10,
|
| 773 |
+
"logits/chosen": 0.7273231744766235,
|
| 774 |
+
"logits/rejected": 1.6111886501312256,
|
| 775 |
+
"logps/chosen": -351.79278564453125,
|
| 776 |
+
"logps/rejected": -395.43890380859375,
|
| 777 |
+
"loss": 0.5241,
|
| 778 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 779 |
+
"rewards/chosen": -0.8372980356216431,
|
| 780 |
+
"rewards/margins": 0.6832476258277893,
|
| 781 |
+
"rewards/rejected": -1.5205457210540771,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 1.0,
|
| 786 |
+
"step": 436,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.5600318115785581,
|
| 789 |
+
"train_runtime": 11378.3731,
|
| 790 |
+
"train_samples_per_second": 4.9,
|
| 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 |
+
}
|