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_consistent-reward-scale-1-rpo-gamma-2
|
| 10 |
+
results: []
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 14 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 15 |
+
|
| 16 |
+
# zephyr-7b-dpo-full-gpt_consistent-reward-scale-1-rpo-gamma-2
|
| 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: 0.1135
|
| 21 |
+
- Rewards/chosen: -0.4357
|
| 22 |
+
- Rewards/rejected: -0.9844
|
| 23 |
+
- Rewards/accuracies: 0.75
|
| 24 |
+
- Rewards/margins: 0.5488
|
| 25 |
+
- Logps/rejected: -344.9655
|
| 26 |
+
- Logps/chosen: -328.6565
|
| 27 |
+
- Logits/rejected: -1.1351
|
| 28 |
+
- Logits/chosen: -1.6080
|
| 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 |
+
| 0.1732 | 0.1147 | 50 | 0.1640 | 0.0148 | -0.1111 | 0.7069 | 0.1258 | -257.6288 | -283.6147 | -2.4924 | -2.5722 |
|
| 66 |
+
| 0.1403 | 0.2294 | 100 | 0.1362 | -0.1370 | -0.4667 | 0.6940 | 0.3297 | -293.1888 | -298.7873 | -1.8344 | -2.0560 |
|
| 67 |
+
| 0.1324 | 0.3440 | 150 | 0.1286 | -0.4769 | -0.9509 | 0.7371 | 0.4740 | -341.6123 | -332.7828 | -1.2887 | -1.6554 |
|
| 68 |
+
| 0.1249 | 0.4587 | 200 | 0.1217 | -0.2893 | -0.7611 | 0.7241 | 0.4719 | -322.6352 | -314.0176 | -1.4798 | -1.8578 |
|
| 69 |
+
| 0.1189 | 0.5734 | 250 | 0.1175 | -0.4263 | -0.9754 | 0.7629 | 0.5491 | -344.0638 | -327.7221 | -1.2227 | -1.6727 |
|
| 70 |
+
| 0.1252 | 0.6881 | 300 | 0.1154 | -0.4298 | -0.9852 | 0.7543 | 0.5554 | -345.0454 | -328.0691 | -1.1891 | -1.6634 |
|
| 71 |
+
| 0.1226 | 0.8028 | 350 | 0.1137 | -0.4793 | -1.0328 | 0.7543 | 0.5535 | -349.7979 | -333.0171 | -1.0590 | -1.5759 |
|
| 72 |
+
| 0.1206 | 0.9174 | 400 | 0.1135 | -0.4357 | -0.9844 | 0.75 | 0.5488 | -344.9655 | -328.6565 | -1.1351 | -1.6080 |
|
| 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": 0.13275344205012016,
|
| 5 |
+
"train_runtime": 11631.0591,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.794,
|
| 8 |
+
"train_steps_per_second": 0.037
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"transformers_version": "4.44.0.dev0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.13275344205012016,
|
| 5 |
+
"train_runtime": 11631.0591,
|
| 6 |
+
"train_samples": 55758,
|
| 7 |
+
"train_samples_per_second": 4.794,
|
| 8 |
+
"train_steps_per_second": 0.037
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,815 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 50,
|
| 6 |
+
"global_step": 436,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.022935779816513763,
|
| 13 |
+
"grad_norm": 5.594593346811413,
|
| 14 |
+
"learning_rate": 1.1363636363636363e-07,
|
| 15 |
+
"logits/chosen": -2.619579315185547,
|
| 16 |
+
"logits/rejected": -2.552640676498413,
|
| 17 |
+
"logps/chosen": -265.4322814941406,
|
| 18 |
+
"logps/rejected": -236.1514434814453,
|
| 19 |
+
"loss": 0.1942,
|
| 20 |
+
"rewards/accuracies": 0.42500001192092896,
|
| 21 |
+
"rewards/chosen": 0.0001355366694042459,
|
| 22 |
+
"rewards/margins": 0.0003500286547932774,
|
| 23 |
+
"rewards/rejected": -0.00021449197083711624,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.045871559633027525,
|
| 28 |
+
"grad_norm": 5.257491443462003,
|
| 29 |
+
"learning_rate": 2.2727272727272726e-07,
|
| 30 |
+
"logits/chosen": -2.6575067043304443,
|
| 31 |
+
"logits/rejected": -2.575622081756592,
|
| 32 |
+
"logps/chosen": -298.8291320800781,
|
| 33 |
+
"logps/rejected": -274.30499267578125,
|
| 34 |
+
"loss": 0.1958,
|
| 35 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 36 |
+
"rewards/chosen": -0.0004384421627037227,
|
| 37 |
+
"rewards/margins": 0.0015119289746508002,
|
| 38 |
+
"rewards/rejected": -0.001950371079146862,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06880733944954129,
|
| 43 |
+
"grad_norm": 4.9882040114909,
|
| 44 |
+
"learning_rate": 3.4090909090909085e-07,
|
| 45 |
+
"logits/chosen": -2.6754746437072754,
|
| 46 |
+
"logits/rejected": -2.60184383392334,
|
| 47 |
+
"logps/chosen": -290.3965759277344,
|
| 48 |
+
"logps/rejected": -234.4066619873047,
|
| 49 |
+
"loss": 0.1923,
|
| 50 |
+
"rewards/accuracies": 0.706250011920929,
|
| 51 |
+
"rewards/chosen": 0.009362037293612957,
|
| 52 |
+
"rewards/margins": 0.013609759509563446,
|
| 53 |
+
"rewards/rejected": -0.004247722681611776,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09174311926605505,
|
| 58 |
+
"grad_norm": 5.21855757124415,
|
| 59 |
+
"learning_rate": 4.545454545454545e-07,
|
| 60 |
+
"logits/chosen": -2.6603915691375732,
|
| 61 |
+
"logits/rejected": -2.610896348953247,
|
| 62 |
+
"logps/chosen": -281.05230712890625,
|
| 63 |
+
"logps/rejected": -267.76873779296875,
|
| 64 |
+
"loss": 0.1836,
|
| 65 |
+
"rewards/accuracies": 0.675000011920929,
|
| 66 |
+
"rewards/chosen": 0.04116874188184738,
|
| 67 |
+
"rewards/margins": 0.04092608764767647,
|
| 68 |
+
"rewards/rejected": 0.00024265461252070963,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.11467889908256881,
|
| 73 |
+
"grad_norm": 5.5623018036583085,
|
| 74 |
+
"learning_rate": 4.997110275491701e-07,
|
| 75 |
+
"logits/chosen": -2.619241714477539,
|
| 76 |
+
"logits/rejected": -2.6116530895233154,
|
| 77 |
+
"logps/chosen": -293.481689453125,
|
| 78 |
+
"logps/rejected": -303.96014404296875,
|
| 79 |
+
"loss": 0.1732,
|
| 80 |
+
"rewards/accuracies": 0.6875,
|
| 81 |
+
"rewards/chosen": 0.02769497036933899,
|
| 82 |
+
"rewards/margins": 0.07545463740825653,
|
| 83 |
+
"rewards/rejected": -0.047759659588336945,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.11467889908256881,
|
| 88 |
+
"eval_logits/chosen": -2.5722098350524902,
|
| 89 |
+
"eval_logits/rejected": -2.492361068725586,
|
| 90 |
+
"eval_logps/chosen": -283.61468505859375,
|
| 91 |
+
"eval_logps/rejected": -257.6287841796875,
|
| 92 |
+
"eval_loss": 0.16397468745708466,
|
| 93 |
+
"eval_rewards/accuracies": 0.7068965435028076,
|
| 94 |
+
"eval_rewards/chosen": 0.014754011295735836,
|
| 95 |
+
"eval_rewards/margins": 0.12581734359264374,
|
| 96 |
+
"eval_rewards/rejected": -0.11106333881616592,
|
| 97 |
+
"eval_runtime": 94.5224,
|
| 98 |
+
"eval_samples_per_second": 19.234,
|
| 99 |
+
"eval_steps_per_second": 0.307,
|
| 100 |
+
"step": 50
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.13761467889908258,
|
| 104 |
+
"grad_norm": 7.227949877380974,
|
| 105 |
+
"learning_rate": 4.979475034558115e-07,
|
| 106 |
+
"logits/chosen": -2.5612680912017822,
|
| 107 |
+
"logits/rejected": -2.5027718544006348,
|
| 108 |
+
"logps/chosen": -293.4398193359375,
|
| 109 |
+
"logps/rejected": -273.11016845703125,
|
| 110 |
+
"loss": 0.1593,
|
| 111 |
+
"rewards/accuracies": 0.699999988079071,
|
| 112 |
+
"rewards/chosen": -0.03955172747373581,
|
| 113 |
+
"rewards/margins": 0.11598478257656097,
|
| 114 |
+
"rewards/rejected": -0.155536487698555,
|
| 115 |
+
"step": 60
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.16055045871559634,
|
| 119 |
+
"grad_norm": 7.344770125290469,
|
| 120 |
+
"learning_rate": 4.945923025551788e-07,
|
| 121 |
+
"logits/chosen": -2.451911449432373,
|
| 122 |
+
"logits/rejected": -2.4021246433258057,
|
| 123 |
+
"logps/chosen": -327.6131591796875,
|
| 124 |
+
"logps/rejected": -284.70257568359375,
|
| 125 |
+
"loss": 0.1564,
|
| 126 |
+
"rewards/accuracies": 0.699999988079071,
|
| 127 |
+
"rewards/chosen": -0.00042213575216010213,
|
| 128 |
+
"rewards/margins": 0.2325272113084793,
|
| 129 |
+
"rewards/rejected": -0.23294934630393982,
|
| 130 |
+
"step": 70
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.1834862385321101,
|
| 134 |
+
"grad_norm": 8.03952466935575,
|
| 135 |
+
"learning_rate": 4.896669632591651e-07,
|
| 136 |
+
"logits/chosen": -2.312901020050049,
|
| 137 |
+
"logits/rejected": -2.1888813972473145,
|
| 138 |
+
"logps/chosen": -298.2128601074219,
|
| 139 |
+
"logps/rejected": -291.5619201660156,
|
| 140 |
+
"loss": 0.1475,
|
| 141 |
+
"rewards/accuracies": 0.731249988079071,
|
| 142 |
+
"rewards/chosen": -0.1236300840973854,
|
| 143 |
+
"rewards/margins": 0.2654283344745636,
|
| 144 |
+
"rewards/rejected": -0.3890584111213684,
|
| 145 |
+
"step": 80
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.20642201834862386,
|
| 149 |
+
"grad_norm": 10.413197257520066,
|
| 150 |
+
"learning_rate": 4.832031033425662e-07,
|
| 151 |
+
"logits/chosen": -1.9742157459259033,
|
| 152 |
+
"logits/rejected": -1.9053027629852295,
|
| 153 |
+
"logps/chosen": -299.2644958496094,
|
| 154 |
+
"logps/rejected": -292.5104064941406,
|
| 155 |
+
"loss": 0.1436,
|
| 156 |
+
"rewards/accuracies": 0.706250011920929,
|
| 157 |
+
"rewards/chosen": -0.11671074479818344,
|
| 158 |
+
"rewards/margins": 0.34038084745407104,
|
| 159 |
+
"rewards/rejected": -0.4570915102958679,
|
| 160 |
+
"step": 90
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.22935779816513763,
|
| 164 |
+
"grad_norm": 8.350562582651383,
|
| 165 |
+
"learning_rate": 4.752422169756047e-07,
|
| 166 |
+
"logits/chosen": -1.6017181873321533,
|
| 167 |
+
"logits/rejected": -1.5160282850265503,
|
| 168 |
+
"logps/chosen": -289.7087097167969,
|
| 169 |
+
"logps/rejected": -317.5405578613281,
|
| 170 |
+
"loss": 0.1403,
|
| 171 |
+
"rewards/accuracies": 0.668749988079071,
|
| 172 |
+
"rewards/chosen": -0.17284031212329865,
|
| 173 |
+
"rewards/margins": 0.28328073024749756,
|
| 174 |
+
"rewards/rejected": -0.4561210572719574,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.22935779816513763,
|
| 179 |
+
"eval_logits/chosen": -2.056002616882324,
|
| 180 |
+
"eval_logits/rejected": -1.8343770503997803,
|
| 181 |
+
"eval_logps/chosen": -298.7873229980469,
|
| 182 |
+
"eval_logps/rejected": -293.1888427734375,
|
| 183 |
+
"eval_loss": 0.1361575871706009,
|
| 184 |
+
"eval_rewards/accuracies": 0.693965494632721,
|
| 185 |
+
"eval_rewards/chosen": -0.13697203993797302,
|
| 186 |
+
"eval_rewards/margins": 0.3296918570995331,
|
| 187 |
+
"eval_rewards/rejected": -0.4666639268398285,
|
| 188 |
+
"eval_runtime": 95.6937,
|
| 189 |
+
"eval_samples_per_second": 18.998,
|
| 190 |
+
"eval_steps_per_second": 0.303,
|
| 191 |
+
"step": 100
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 0.25229357798165136,
|
| 195 |
+
"grad_norm": 8.176356995673753,
|
| 196 |
+
"learning_rate": 4.658354083558188e-07,
|
| 197 |
+
"logits/chosen": -2.1571879386901855,
|
| 198 |
+
"logits/rejected": -1.9664814472198486,
|
| 199 |
+
"logps/chosen": -285.4144287109375,
|
| 200 |
+
"logps/rejected": -280.06732177734375,
|
| 201 |
+
"loss": 0.1397,
|
| 202 |
+
"rewards/accuracies": 0.706250011920929,
|
| 203 |
+
"rewards/chosen": -0.2022629678249359,
|
| 204 |
+
"rewards/margins": 0.2610785961151123,
|
| 205 |
+
"rewards/rejected": -0.46334153413772583,
|
| 206 |
+
"step": 110
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.27522935779816515,
|
| 210 |
+
"grad_norm": 11.47061183691114,
|
| 211 |
+
"learning_rate": 4.550430636492389e-07,
|
| 212 |
+
"logits/chosen": -1.4951406717300415,
|
| 213 |
+
"logits/rejected": -1.340619444847107,
|
| 214 |
+
"logps/chosen": -316.77838134765625,
|
| 215 |
+
"logps/rejected": -329.01495361328125,
|
| 216 |
+
"loss": 0.1386,
|
| 217 |
+
"rewards/accuracies": 0.675000011920929,
|
| 218 |
+
"rewards/chosen": -0.4095904231071472,
|
| 219 |
+
"rewards/margins": 0.3659195899963379,
|
| 220 |
+
"rewards/rejected": -0.7755100131034851,
|
| 221 |
+
"step": 120
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.2981651376146789,
|
| 225 |
+
"grad_norm": 12.47670192079271,
|
| 226 |
+
"learning_rate": 4.429344633468004e-07,
|
| 227 |
+
"logits/chosen": -1.376880168914795,
|
| 228 |
+
"logits/rejected": -1.1415055990219116,
|
| 229 |
+
"logps/chosen": -303.68109130859375,
|
| 230 |
+
"logps/rejected": -326.821533203125,
|
| 231 |
+
"loss": 0.1383,
|
| 232 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 233 |
+
"rewards/chosen": -0.47500982880592346,
|
| 234 |
+
"rewards/margins": 0.3670799136161804,
|
| 235 |
+
"rewards/rejected": -0.8420897722244263,
|
| 236 |
+
"step": 130
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 0.3211009174311927,
|
| 240 |
+
"grad_norm": 10.567717397857333,
|
| 241 |
+
"learning_rate": 4.2958733752443187e-07,
|
| 242 |
+
"logits/chosen": -1.6405372619628906,
|
| 243 |
+
"logits/rejected": -1.487768530845642,
|
| 244 |
+
"logps/chosen": -310.8406677246094,
|
| 245 |
+
"logps/rejected": -290.4248962402344,
|
| 246 |
+
"loss": 0.1328,
|
| 247 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 248 |
+
"rewards/chosen": -0.3465554118156433,
|
| 249 |
+
"rewards/margins": 0.31472834944725037,
|
| 250 |
+
"rewards/rejected": -0.6612837910652161,
|
| 251 |
+
"step": 140
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 0.3440366972477064,
|
| 255 |
+
"grad_norm": 10.103632857127238,
|
| 256 |
+
"learning_rate": 4.150873668617898e-07,
|
| 257 |
+
"logits/chosen": -1.5787580013275146,
|
| 258 |
+
"logits/rejected": -1.270272970199585,
|
| 259 |
+
"logps/chosen": -316.16461181640625,
|
| 260 |
+
"logps/rejected": -323.0017395019531,
|
| 261 |
+
"loss": 0.1324,
|
| 262 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 263 |
+
"rewards/chosen": -0.49621015787124634,
|
| 264 |
+
"rewards/margins": 0.4019313454627991,
|
| 265 |
+
"rewards/rejected": -0.8981413841247559,
|
| 266 |
+
"step": 150
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 0.3440366972477064,
|
| 270 |
+
"eval_logits/chosen": -1.6554096937179565,
|
| 271 |
+
"eval_logits/rejected": -1.2887206077575684,
|
| 272 |
+
"eval_logps/chosen": -332.78277587890625,
|
| 273 |
+
"eval_logps/rejected": -341.6122741699219,
|
| 274 |
+
"eval_loss": 0.12861265242099762,
|
| 275 |
+
"eval_rewards/accuracies": 0.7370689511299133,
|
| 276 |
+
"eval_rewards/chosen": -0.47692668437957764,
|
| 277 |
+
"eval_rewards/margins": 0.47397124767303467,
|
| 278 |
+
"eval_rewards/rejected": -0.9508979916572571,
|
| 279 |
+
"eval_runtime": 94.5745,
|
| 280 |
+
"eval_samples_per_second": 19.223,
|
| 281 |
+
"eval_steps_per_second": 0.307,
|
| 282 |
+
"step": 150
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 0.3669724770642202,
|
| 286 |
+
"grad_norm": 12.610771394953652,
|
| 287 |
+
"learning_rate": 3.9952763262280397e-07,
|
| 288 |
+
"logits/chosen": -1.6438308954238892,
|
| 289 |
+
"logits/rejected": -1.378089427947998,
|
| 290 |
+
"logps/chosen": -342.8047790527344,
|
| 291 |
+
"logps/rejected": -378.4410705566406,
|
| 292 |
+
"loss": 0.125,
|
| 293 |
+
"rewards/accuracies": 0.731249988079071,
|
| 294 |
+
"rewards/chosen": -0.46253666281700134,
|
| 295 |
+
"rewards/margins": 0.4621877670288086,
|
| 296 |
+
"rewards/rejected": -0.9247244000434875,
|
| 297 |
+
"step": 160
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38990825688073394,
|
| 301 |
+
"grad_norm": 7.3876281210799375,
|
| 302 |
+
"learning_rate": 3.8300801912883414e-07,
|
| 303 |
+
"logits/chosen": -1.8106731176376343,
|
| 304 |
+
"logits/rejected": -1.5751054286956787,
|
| 305 |
+
"logps/chosen": -300.5574035644531,
|
| 306 |
+
"logps/rejected": -343.215576171875,
|
| 307 |
+
"loss": 0.12,
|
| 308 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 309 |
+
"rewards/chosen": -0.28956499695777893,
|
| 310 |
+
"rewards/margins": 0.4419015049934387,
|
| 311 |
+
"rewards/rejected": -0.73146653175354,
|
| 312 |
+
"step": 170
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 0.41284403669724773,
|
| 316 |
+
"grad_norm": 9.731458591388959,
|
| 317 |
+
"learning_rate": 3.6563457256020884e-07,
|
| 318 |
+
"logits/chosen": -1.6891987323760986,
|
| 319 |
+
"logits/rejected": -1.344957709312439,
|
| 320 |
+
"logps/chosen": -351.0888671875,
|
| 321 |
+
"logps/rejected": -328.3114013671875,
|
| 322 |
+
"loss": 0.129,
|
| 323 |
+
"rewards/accuracies": 0.731249988079071,
|
| 324 |
+
"rewards/chosen": -0.47730937600135803,
|
| 325 |
+
"rewards/margins": 0.42085570096969604,
|
| 326 |
+
"rewards/rejected": -0.8981650471687317,
|
| 327 |
+
"step": 180
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.43577981651376146,
|
| 331 |
+
"grad_norm": 10.618153958266253,
|
| 332 |
+
"learning_rate": 3.475188202022617e-07,
|
| 333 |
+
"logits/chosen": -1.5754601955413818,
|
| 334 |
+
"logits/rejected": -1.322837471961975,
|
| 335 |
+
"logps/chosen": -299.17999267578125,
|
| 336 |
+
"logps/rejected": -353.1037292480469,
|
| 337 |
+
"loss": 0.1248,
|
| 338 |
+
"rewards/accuracies": 0.768750011920929,
|
| 339 |
+
"rewards/chosen": -0.41065484285354614,
|
| 340 |
+
"rewards/margins": 0.4769380986690521,
|
| 341 |
+
"rewards/rejected": -0.8875927925109863,
|
| 342 |
+
"step": 190
|
| 343 |
+
},
|
| 344 |
+
{
|
| 345 |
+
"epoch": 0.45871559633027525,
|
| 346 |
+
"grad_norm": 8.299502199250934,
|
| 347 |
+
"learning_rate": 3.287770545059052e-07,
|
| 348 |
+
"logits/chosen": -1.7796523571014404,
|
| 349 |
+
"logits/rejected": -1.4286386966705322,
|
| 350 |
+
"logps/chosen": -312.58306884765625,
|
| 351 |
+
"logps/rejected": -324.02142333984375,
|
| 352 |
+
"loss": 0.1249,
|
| 353 |
+
"rewards/accuracies": 0.762499988079071,
|
| 354 |
+
"rewards/chosen": -0.3430142104625702,
|
| 355 |
+
"rewards/margins": 0.4586402475833893,
|
| 356 |
+
"rewards/rejected": -0.8016544580459595,
|
| 357 |
+
"step": 200
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.45871559633027525,
|
| 361 |
+
"eval_logits/chosen": -1.8577966690063477,
|
| 362 |
+
"eval_logits/rejected": -1.4797794818878174,
|
| 363 |
+
"eval_logps/chosen": -314.017578125,
|
| 364 |
+
"eval_logps/rejected": -322.6352233886719,
|
| 365 |
+
"eval_loss": 0.12169274687767029,
|
| 366 |
+
"eval_rewards/accuracies": 0.7241379022598267,
|
| 367 |
+
"eval_rewards/chosen": -0.2892747223377228,
|
| 368 |
+
"eval_rewards/margins": 0.4718529284000397,
|
| 369 |
+
"eval_rewards/rejected": -0.7611277103424072,
|
| 370 |
+
"eval_runtime": 95.5757,
|
| 371 |
+
"eval_samples_per_second": 19.022,
|
| 372 |
+
"eval_steps_per_second": 0.303,
|
| 373 |
+
"step": 200
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"epoch": 0.481651376146789,
|
| 377 |
+
"grad_norm": 8.358424851083543,
|
| 378 |
+
"learning_rate": 3.0952958655864954e-07,
|
| 379 |
+
"logits/chosen": -1.7499780654907227,
|
| 380 |
+
"logits/rejected": -1.4417035579681396,
|
| 381 |
+
"logps/chosen": -303.82586669921875,
|
| 382 |
+
"logps/rejected": -331.75396728515625,
|
| 383 |
+
"loss": 0.1201,
|
| 384 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 385 |
+
"rewards/chosen": -0.30521491169929504,
|
| 386 |
+
"rewards/margins": 0.46890267729759216,
|
| 387 |
+
"rewards/rejected": -0.7741175293922424,
|
| 388 |
+
"step": 210
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.5045871559633027,
|
| 392 |
+
"grad_norm": 9.67017691007605,
|
| 393 |
+
"learning_rate": 2.898999737583448e-07,
|
| 394 |
+
"logits/chosen": -1.6984879970550537,
|
| 395 |
+
"logits/rejected": -1.103266954421997,
|
| 396 |
+
"logps/chosen": -371.213623046875,
|
| 397 |
+
"logps/rejected": -387.9047546386719,
|
| 398 |
+
"loss": 0.1164,
|
| 399 |
+
"rewards/accuracies": 0.78125,
|
| 400 |
+
"rewards/chosen": -0.4853332042694092,
|
| 401 |
+
"rewards/margins": 0.5440121293067932,
|
| 402 |
+
"rewards/rejected": -1.0293452739715576,
|
| 403 |
+
"step": 220
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 0.5275229357798165,
|
| 407 |
+
"grad_norm": 9.051099459635976,
|
| 408 |
+
"learning_rate": 2.7001422664752333e-07,
|
| 409 |
+
"logits/chosen": -1.5834382772445679,
|
| 410 |
+
"logits/rejected": -1.3102777004241943,
|
| 411 |
+
"logps/chosen": -317.35888671875,
|
| 412 |
+
"logps/rejected": -356.3561096191406,
|
| 413 |
+
"loss": 0.1216,
|
| 414 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 415 |
+
"rewards/chosen": -0.4941536784172058,
|
| 416 |
+
"rewards/margins": 0.4132082462310791,
|
| 417 |
+
"rewards/rejected": -0.9073619842529297,
|
| 418 |
+
"step": 230
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"epoch": 0.5504587155963303,
|
| 422 |
+
"grad_norm": 8.670604935851552,
|
| 423 |
+
"learning_rate": 2.5e-07,
|
| 424 |
+
"logits/chosen": -1.79386305809021,
|
| 425 |
+
"logits/rejected": -1.4528472423553467,
|
| 426 |
+
"logps/chosen": -328.12200927734375,
|
| 427 |
+
"logps/rejected": -345.81500244140625,
|
| 428 |
+
"loss": 0.1217,
|
| 429 |
+
"rewards/accuracies": 0.71875,
|
| 430 |
+
"rewards/chosen": -0.40590184926986694,
|
| 431 |
+
"rewards/margins": 0.43874797224998474,
|
| 432 |
+
"rewards/rejected": -0.8446499109268188,
|
| 433 |
+
"step": 240
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.573394495412844,
|
| 437 |
+
"grad_norm": 8.325091240643497,
|
| 438 |
+
"learning_rate": 2.2998577335247667e-07,
|
| 439 |
+
"logits/chosen": -1.7739427089691162,
|
| 440 |
+
"logits/rejected": -1.3005108833312988,
|
| 441 |
+
"logps/chosen": -344.14129638671875,
|
| 442 |
+
"logps/rejected": -352.2928771972656,
|
| 443 |
+
"loss": 0.1189,
|
| 444 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 445 |
+
"rewards/chosen": -0.4034748673439026,
|
| 446 |
+
"rewards/margins": 0.515015721321106,
|
| 447 |
+
"rewards/rejected": -0.9184905886650085,
|
| 448 |
+
"step": 250
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"epoch": 0.573394495412844,
|
| 452 |
+
"eval_logits/chosen": -1.6726821660995483,
|
| 453 |
+
"eval_logits/rejected": -1.2226585149765015,
|
| 454 |
+
"eval_logps/chosen": -327.7221374511719,
|
| 455 |
+
"eval_logps/rejected": -344.0638427734375,
|
| 456 |
+
"eval_loss": 0.11745046824216843,
|
| 457 |
+
"eval_rewards/accuracies": 0.7629310488700867,
|
| 458 |
+
"eval_rewards/chosen": -0.4263203740119934,
|
| 459 |
+
"eval_rewards/margins": 0.5490937232971191,
|
| 460 |
+
"eval_rewards/rejected": -0.9754140973091125,
|
| 461 |
+
"eval_runtime": 95.1095,
|
| 462 |
+
"eval_samples_per_second": 19.115,
|
| 463 |
+
"eval_steps_per_second": 0.305,
|
| 464 |
+
"step": 250
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 0.5963302752293578,
|
| 468 |
+
"grad_norm": 8.867617693309038,
|
| 469 |
+
"learning_rate": 2.1010002624165524e-07,
|
| 470 |
+
"logits/chosen": -1.6618998050689697,
|
| 471 |
+
"logits/rejected": -1.3614200353622437,
|
| 472 |
+
"logps/chosen": -315.7143249511719,
|
| 473 |
+
"logps/rejected": -377.2685546875,
|
| 474 |
+
"loss": 0.117,
|
| 475 |
+
"rewards/accuracies": 0.800000011920929,
|
| 476 |
+
"rewards/chosen": -0.44999808073043823,
|
| 477 |
+
"rewards/margins": 0.5787504315376282,
|
| 478 |
+
"rewards/rejected": -1.0287485122680664,
|
| 479 |
+
"step": 260
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.6192660550458715,
|
| 483 |
+
"grad_norm": 8.935506104756715,
|
| 484 |
+
"learning_rate": 1.9047041344135043e-07,
|
| 485 |
+
"logits/chosen": -1.6130893230438232,
|
| 486 |
+
"logits/rejected": -1.4423035383224487,
|
| 487 |
+
"logps/chosen": -308.60711669921875,
|
| 488 |
+
"logps/rejected": -346.1998291015625,
|
| 489 |
+
"loss": 0.1177,
|
| 490 |
+
"rewards/accuracies": 0.793749988079071,
|
| 491 |
+
"rewards/chosen": -0.4254485070705414,
|
| 492 |
+
"rewards/margins": 0.5180791020393372,
|
| 493 |
+
"rewards/rejected": -0.9435275793075562,
|
| 494 |
+
"step": 270
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 0.6422018348623854,
|
| 498 |
+
"grad_norm": 9.329726294195847,
|
| 499 |
+
"learning_rate": 1.7122294549409482e-07,
|
| 500 |
+
"logits/chosen": -1.6924329996109009,
|
| 501 |
+
"logits/rejected": -1.2535055875778198,
|
| 502 |
+
"logps/chosen": -311.0262451171875,
|
| 503 |
+
"logps/rejected": -363.9076232910156,
|
| 504 |
+
"loss": 0.1216,
|
| 505 |
+
"rewards/accuracies": 0.793749988079071,
|
| 506 |
+
"rewards/chosen": -0.3802736699581146,
|
| 507 |
+
"rewards/margins": 0.6208445429801941,
|
| 508 |
+
"rewards/rejected": -1.0011183023452759,
|
| 509 |
+
"step": 280
|
| 510 |
+
},
|
| 511 |
+
{
|
| 512 |
+
"epoch": 0.6651376146788991,
|
| 513 |
+
"grad_norm": 11.266324554606376,
|
| 514 |
+
"learning_rate": 1.524811797977383e-07,
|
| 515 |
+
"logits/chosen": -1.6415001153945923,
|
| 516 |
+
"logits/rejected": -1.2690132856369019,
|
| 517 |
+
"logps/chosen": -327.433837890625,
|
| 518 |
+
"logps/rejected": -346.80560302734375,
|
| 519 |
+
"loss": 0.1149,
|
| 520 |
+
"rewards/accuracies": 0.768750011920929,
|
| 521 |
+
"rewards/chosen": -0.3970295786857605,
|
| 522 |
+
"rewards/margins": 0.5318586826324463,
|
| 523 |
+
"rewards/rejected": -0.928888201713562,
|
| 524 |
+
"step": 290
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 0.6880733944954128,
|
| 528 |
+
"grad_norm": 9.171141301041647,
|
| 529 |
+
"learning_rate": 1.3436542743979125e-07,
|
| 530 |
+
"logits/chosen": -1.6442855596542358,
|
| 531 |
+
"logits/rejected": -1.4366796016693115,
|
| 532 |
+
"logps/chosen": -349.1554870605469,
|
| 533 |
+
"logps/rejected": -343.85443115234375,
|
| 534 |
+
"loss": 0.1252,
|
| 535 |
+
"rewards/accuracies": 0.737500011920929,
|
| 536 |
+
"rewards/chosen": -0.4320616126060486,
|
| 537 |
+
"rewards/margins": 0.4178841710090637,
|
| 538 |
+
"rewards/rejected": -0.8499458432197571,
|
| 539 |
+
"step": 300
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 0.6880733944954128,
|
| 543 |
+
"eval_logits/chosen": -1.663388729095459,
|
| 544 |
+
"eval_logits/rejected": -1.1890878677368164,
|
| 545 |
+
"eval_logps/chosen": -328.0691223144531,
|
| 546 |
+
"eval_logps/rejected": -345.0453796386719,
|
| 547 |
+
"eval_loss": 0.11537355184555054,
|
| 548 |
+
"eval_rewards/accuracies": 0.7543103694915771,
|
| 549 |
+
"eval_rewards/chosen": -0.4297903776168823,
|
| 550 |
+
"eval_rewards/margins": 0.5554393529891968,
|
| 551 |
+
"eval_rewards/rejected": -0.9852296710014343,
|
| 552 |
+
"eval_runtime": 94.7636,
|
| 553 |
+
"eval_samples_per_second": 19.185,
|
| 554 |
+
"eval_steps_per_second": 0.306,
|
| 555 |
+
"step": 300
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 0.7110091743119266,
|
| 559 |
+
"grad_norm": 8.786286019785855,
|
| 560 |
+
"learning_rate": 1.1699198087116588e-07,
|
| 561 |
+
"logits/chosen": -1.6734371185302734,
|
| 562 |
+
"logits/rejected": -1.3057066202163696,
|
| 563 |
+
"logps/chosen": -325.54986572265625,
|
| 564 |
+
"logps/rejected": -359.43572998046875,
|
| 565 |
+
"loss": 0.123,
|
| 566 |
+
"rewards/accuracies": 0.731249988079071,
|
| 567 |
+
"rewards/chosen": -0.47584667801856995,
|
| 568 |
+
"rewards/margins": 0.4540133476257324,
|
| 569 |
+
"rewards/rejected": -0.92985999584198,
|
| 570 |
+
"step": 310
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 0.7339449541284404,
|
| 574 |
+
"grad_norm": 9.586519281194345,
|
| 575 |
+
"learning_rate": 1.00472367377196e-07,
|
| 576 |
+
"logits/chosen": -1.605583906173706,
|
| 577 |
+
"logits/rejected": -1.1544862985610962,
|
| 578 |
+
"logps/chosen": -318.45654296875,
|
| 579 |
+
"logps/rejected": -339.86163330078125,
|
| 580 |
+
"loss": 0.1189,
|
| 581 |
+
"rewards/accuracies": 0.768750011920929,
|
| 582 |
+
"rewards/chosen": -0.4422459006309509,
|
| 583 |
+
"rewards/margins": 0.611468493938446,
|
| 584 |
+
"rewards/rejected": -1.053714394569397,
|
| 585 |
+
"step": 320
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.7568807339449541,
|
| 589 |
+
"grad_norm": 9.694294177484474,
|
| 590 |
+
"learning_rate": 8.49126331382102e-08,
|
| 591 |
+
"logits/chosen": -1.499674677848816,
|
| 592 |
+
"logits/rejected": -1.1590025424957275,
|
| 593 |
+
"logps/chosen": -327.3614807128906,
|
| 594 |
+
"logps/rejected": -342.82550048828125,
|
| 595 |
+
"loss": 0.1187,
|
| 596 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 597 |
+
"rewards/chosen": -0.5283932685852051,
|
| 598 |
+
"rewards/margins": 0.4288608431816101,
|
| 599 |
+
"rewards/rejected": -0.9572542309761047,
|
| 600 |
+
"step": 330
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 0.7798165137614679,
|
| 604 |
+
"grad_norm": 10.420796259880595,
|
| 605 |
+
"learning_rate": 7.041266247556812e-08,
|
| 606 |
+
"logits/chosen": -1.587501883506775,
|
| 607 |
+
"logits/rejected": -1.274778127670288,
|
| 608 |
+
"logps/chosen": -338.21112060546875,
|
| 609 |
+
"logps/rejected": -354.18890380859375,
|
| 610 |
+
"loss": 0.1144,
|
| 611 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 612 |
+
"rewards/chosen": -0.5021848678588867,
|
| 613 |
+
"rewards/margins": 0.439863920211792,
|
| 614 |
+
"rewards/rejected": -0.9420488476753235,
|
| 615 |
+
"step": 340
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 0.8027522935779816,
|
| 619 |
+
"grad_norm": 9.40520686958908,
|
| 620 |
+
"learning_rate": 5.706553665319955e-08,
|
| 621 |
+
"logits/chosen": -1.567800521850586,
|
| 622 |
+
"logits/rejected": -1.0267796516418457,
|
| 623 |
+
"logps/chosen": -327.0886535644531,
|
| 624 |
+
"logps/rejected": -339.1602478027344,
|
| 625 |
+
"loss": 0.1226,
|
| 626 |
+
"rewards/accuracies": 0.768750011920929,
|
| 627 |
+
"rewards/chosen": -0.5259757041931152,
|
| 628 |
+
"rewards/margins": 0.48901230096817017,
|
| 629 |
+
"rewards/rejected": -1.0149879455566406,
|
| 630 |
+
"step": 350
|
| 631 |
+
},
|
| 632 |
+
{
|
| 633 |
+
"epoch": 0.8027522935779816,
|
| 634 |
+
"eval_logits/chosen": -1.5758812427520752,
|
| 635 |
+
"eval_logits/rejected": -1.0590085983276367,
|
| 636 |
+
"eval_logps/chosen": -333.0171203613281,
|
| 637 |
+
"eval_logps/rejected": -349.79791259765625,
|
| 638 |
+
"eval_loss": 0.11367923766374588,
|
| 639 |
+
"eval_rewards/accuracies": 0.7543103694915771,
|
| 640 |
+
"eval_rewards/chosen": -0.47927024960517883,
|
| 641 |
+
"eval_rewards/margins": 0.5534845590591431,
|
| 642 |
+
"eval_rewards/rejected": -1.032754898071289,
|
| 643 |
+
"eval_runtime": 94.1796,
|
| 644 |
+
"eval_samples_per_second": 19.304,
|
| 645 |
+
"eval_steps_per_second": 0.308,
|
| 646 |
+
"step": 350
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"epoch": 0.8256880733944955,
|
| 650 |
+
"grad_norm": 7.926122004749844,
|
| 651 |
+
"learning_rate": 4.4956936350761005e-08,
|
| 652 |
+
"logits/chosen": -1.530390977859497,
|
| 653 |
+
"logits/rejected": -1.2470009326934814,
|
| 654 |
+
"logps/chosen": -297.60394287109375,
|
| 655 |
+
"logps/rejected": -349.6514587402344,
|
| 656 |
+
"loss": 0.1145,
|
| 657 |
+
"rewards/accuracies": 0.706250011920929,
|
| 658 |
+
"rewards/chosen": -0.45283880829811096,
|
| 659 |
+
"rewards/margins": 0.45830464363098145,
|
| 660 |
+
"rewards/rejected": -0.91114342212677,
|
| 661 |
+
"step": 360
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.8486238532110092,
|
| 665 |
+
"grad_norm": 8.501039777999448,
|
| 666 |
+
"learning_rate": 3.416459164418123e-08,
|
| 667 |
+
"logits/chosen": -1.637953519821167,
|
| 668 |
+
"logits/rejected": -1.3401291370391846,
|
| 669 |
+
"logps/chosen": -346.0074462890625,
|
| 670 |
+
"logps/rejected": -358.09075927734375,
|
| 671 |
+
"loss": 0.1125,
|
| 672 |
+
"rewards/accuracies": 0.75,
|
| 673 |
+
"rewards/chosen": -0.46422967314720154,
|
| 674 |
+
"rewards/margins": 0.4909321367740631,
|
| 675 |
+
"rewards/rejected": -0.9551618695259094,
|
| 676 |
+
"step": 370
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 0.8715596330275229,
|
| 680 |
+
"grad_norm": 10.294163163028395,
|
| 681 |
+
"learning_rate": 2.475778302439524e-08,
|
| 682 |
+
"logits/chosen": -1.6482412815093994,
|
| 683 |
+
"logits/rejected": -1.1318132877349854,
|
| 684 |
+
"logps/chosen": -337.7255859375,
|
| 685 |
+
"logps/rejected": -354.24090576171875,
|
| 686 |
+
"loss": 0.118,
|
| 687 |
+
"rewards/accuracies": 0.762499988079071,
|
| 688 |
+
"rewards/chosen": -0.4134751260280609,
|
| 689 |
+
"rewards/margins": 0.5523717403411865,
|
| 690 |
+
"rewards/rejected": -0.965846836566925,
|
| 691 |
+
"step": 380
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"epoch": 0.8944954128440367,
|
| 695 |
+
"grad_norm": 9.29566160345761,
|
| 696 |
+
"learning_rate": 1.6796896657433805e-08,
|
| 697 |
+
"logits/chosen": -1.5356476306915283,
|
| 698 |
+
"logits/rejected": -1.1930160522460938,
|
| 699 |
+
"logps/chosen": -296.85870361328125,
|
| 700 |
+
"logps/rejected": -320.10577392578125,
|
| 701 |
+
"loss": 0.1227,
|
| 702 |
+
"rewards/accuracies": 0.731249988079071,
|
| 703 |
+
"rewards/chosen": -0.4490123391151428,
|
| 704 |
+
"rewards/margins": 0.4589834213256836,
|
| 705 |
+
"rewards/rejected": -0.9079957008361816,
|
| 706 |
+
"step": 390
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 0.9174311926605505,
|
| 710 |
+
"grad_norm": 10.238826336953899,
|
| 711 |
+
"learning_rate": 1.0333036740834855e-08,
|
| 712 |
+
"logits/chosen": -1.455672264099121,
|
| 713 |
+
"logits/rejected": -1.2008806467056274,
|
| 714 |
+
"logps/chosen": -265.4068298339844,
|
| 715 |
+
"logps/rejected": -317.6711120605469,
|
| 716 |
+
"loss": 0.1206,
|
| 717 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 718 |
+
"rewards/chosen": -0.4084719121456146,
|
| 719 |
+
"rewards/margins": 0.4760831892490387,
|
| 720 |
+
"rewards/rejected": -0.8845551609992981,
|
| 721 |
+
"step": 400
|
| 722 |
+
},
|
| 723 |
+
{
|
| 724 |
+
"epoch": 0.9174311926605505,
|
| 725 |
+
"eval_logits/chosen": -1.608026385307312,
|
| 726 |
+
"eval_logits/rejected": -1.1351321935653687,
|
| 727 |
+
"eval_logps/chosen": -328.656494140625,
|
| 728 |
+
"eval_logps/rejected": -344.9654846191406,
|
| 729 |
+
"eval_loss": 0.11354311555624008,
|
| 730 |
+
"eval_rewards/accuracies": 0.75,
|
| 731 |
+
"eval_rewards/chosen": -0.43566375970840454,
|
| 732 |
+
"eval_rewards/margins": 0.5487664937973022,
|
| 733 |
+
"eval_rewards/rejected": -0.984430193901062,
|
| 734 |
+
"eval_runtime": 96.0499,
|
| 735 |
+
"eval_samples_per_second": 18.928,
|
| 736 |
+
"eval_steps_per_second": 0.302,
|
| 737 |
+
"step": 400
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 0.9403669724770642,
|
| 741 |
+
"grad_norm": 10.26611326963808,
|
| 742 |
+
"learning_rate": 5.4076974448211685e-09,
|
| 743 |
+
"logits/chosen": -1.4635480642318726,
|
| 744 |
+
"logits/rejected": -1.029957890510559,
|
| 745 |
+
"logps/chosen": -317.2907409667969,
|
| 746 |
+
"logps/rejected": -339.1365051269531,
|
| 747 |
+
"loss": 0.1172,
|
| 748 |
+
"rewards/accuracies": 0.75,
|
| 749 |
+
"rewards/chosen": -0.4633886218070984,
|
| 750 |
+
"rewards/margins": 0.5475090742111206,
|
| 751 |
+
"rewards/rejected": -1.0108975172042847,
|
| 752 |
+
"step": 410
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"epoch": 0.963302752293578,
|
| 756 |
+
"grad_norm": 11.907377094577166,
|
| 757 |
+
"learning_rate": 2.052496544188487e-09,
|
| 758 |
+
"logits/chosen": -1.477085828781128,
|
| 759 |
+
"logits/rejected": -0.9949935078620911,
|
| 760 |
+
"logps/chosen": -308.0985412597656,
|
| 761 |
+
"logps/rejected": -351.2358093261719,
|
| 762 |
+
"loss": 0.1158,
|
| 763 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 764 |
+
"rewards/chosen": -0.4865953028202057,
|
| 765 |
+
"rewards/margins": 0.5669893026351929,
|
| 766 |
+
"rewards/rejected": -1.0535845756530762,
|
| 767 |
+
"step": 420
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.9862385321100917,
|
| 771 |
+
"grad_norm": 9.339603996948918,
|
| 772 |
+
"learning_rate": 2.889724508297886e-10,
|
| 773 |
+
"logits/chosen": -1.5852940082550049,
|
| 774 |
+
"logits/rejected": -1.0293024778366089,
|
| 775 |
+
"logps/chosen": -347.1506042480469,
|
| 776 |
+
"logps/rejected": -338.89019775390625,
|
| 777 |
+
"loss": 0.1161,
|
| 778 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 779 |
+
"rewards/chosen": -0.439751535654068,
|
| 780 |
+
"rewards/margins": 0.5087853074073792,
|
| 781 |
+
"rewards/rejected": -0.94853675365448,
|
| 782 |
+
"step": 430
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 1.0,
|
| 786 |
+
"step": 436,
|
| 787 |
+
"total_flos": 0.0,
|
| 788 |
+
"train_loss": 0.13275344205012016,
|
| 789 |
+
"train_runtime": 11631.0591,
|
| 790 |
+
"train_samples_per_second": 4.794,
|
| 791 |
+
"train_steps_per_second": 0.037
|
| 792 |
+
}
|
| 793 |
+
],
|
| 794 |
+
"logging_steps": 10,
|
| 795 |
+
"max_steps": 436,
|
| 796 |
+
"num_input_tokens_seen": 0,
|
| 797 |
+
"num_train_epochs": 1,
|
| 798 |
+
"save_steps": 100,
|
| 799 |
+
"stateful_callbacks": {
|
| 800 |
+
"TrainerControl": {
|
| 801 |
+
"args": {
|
| 802 |
+
"should_epoch_stop": false,
|
| 803 |
+
"should_evaluate": false,
|
| 804 |
+
"should_log": false,
|
| 805 |
+
"should_save": true,
|
| 806 |
+
"should_training_stop": true
|
| 807 |
+
},
|
| 808 |
+
"attributes": {}
|
| 809 |
+
}
|
| 810 |
+
},
|
| 811 |
+
"total_flos": 0.0,
|
| 812 |
+
"train_batch_size": 8,
|
| 813 |
+
"trial_name": null,
|
| 814 |
+
"trial_params": null
|
| 815 |
+
}
|