Model save
Browse files- README.md +76 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +811 -0
README.md
ADDED
|
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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-ultrabin-3-avg-logprob-lr-same
|
| 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-ultrabin-3-avg-logprob-lr-same
|
| 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 an unknown dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.6219
|
| 21 |
+
- Rewards/chosen: -0.2287
|
| 22 |
+
- Rewards/rejected: -0.4214
|
| 23 |
+
- Rewards/accuracies: 0.7188
|
| 24 |
+
- Rewards/margins: 0.1928
|
| 25 |
+
- Logps/rejected: -43.2637
|
| 26 |
+
- Logps/chosen: -23.8237
|
| 27 |
+
- Logits/rejected: 2.7050
|
| 28 |
+
- Logits/chosen: 2.7595
|
| 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 |
+
| 0.6803 | 0.2092 | 100 | 0.6741 | -0.1830 | -0.2238 | 0.6055 | 0.0408 | -23.4969 | -19.2551 | 5.1619 | 4.6722 |
|
| 66 |
+
| 0.6333 | 0.4184 | 200 | 0.6318 | -0.2607 | -0.4263 | 0.7070 | 0.1656 | -43.7487 | -27.0254 | 2.0618 | 2.2059 |
|
| 67 |
+
| 0.6202 | 0.6276 | 300 | 0.6256 | -0.1960 | -0.3896 | 0.7188 | 0.1936 | -40.0823 | -20.5594 | 3.0356 | 2.9998 |
|
| 68 |
+
| 0.6195 | 0.8368 | 400 | 0.6219 | -0.2287 | -0.4214 | 0.7188 | 0.1928 | -43.2637 | -23.8237 | 2.7050 | 2.7595 |
|
| 69 |
+
|
| 70 |
+
|
| 71 |
+
### Framework versions
|
| 72 |
+
|
| 73 |
+
- Transformers 4.44.0.dev0
|
| 74 |
+
- Pytorch 2.1.2
|
| 75 |
+
- Datasets 2.20.0
|
| 76 |
+
- 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": 0.6403242514223234,
|
| 5 |
+
"train_runtime": 11960.3252,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 5.111,
|
| 8 |
+
"train_steps_per_second": 0.04
|
| 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": 0.6403242514223234,
|
| 5 |
+
"train_runtime": 11960.3252,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 5.111,
|
| 8 |
+
"train_steps_per_second": 0.04
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,811 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 478,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.02092050209205021,
|
| 13 |
+
"grad_norm": 0.07870777149790437,
|
| 14 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 15 |
+
"logits/chosen": -2.6306002140045166,
|
| 16 |
+
"logits/rejected": -2.576826572418213,
|
| 17 |
+
"logps/chosen": -1.0156770944595337,
|
| 18 |
+
"logps/rejected": -1.187302827835083,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.48124998807907104,
|
| 21 |
+
"rewards/chosen": 4.107605946046533e-06,
|
| 22 |
+
"rewards/margins": 6.403818588296417e-06,
|
| 23 |
+
"rewards/rejected": -2.296213096997235e-06,
|
| 24 |
+
"step": 10
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.04184100418410042,
|
| 28 |
+
"grad_norm": 0.07901324239389455,
|
| 29 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 30 |
+
"logits/chosen": -2.6465299129486084,
|
| 31 |
+
"logits/rejected": -2.615328311920166,
|
| 32 |
+
"logps/chosen": -1.072858452796936,
|
| 33 |
+
"logps/rejected": -1.1568002700805664,
|
| 34 |
+
"loss": 0.6931,
|
| 35 |
+
"rewards/accuracies": 0.581250011920929,
|
| 36 |
+
"rewards/chosen": 5.4746556997997686e-05,
|
| 37 |
+
"rewards/margins": 2.0796986063942313e-05,
|
| 38 |
+
"rewards/rejected": 3.394957457203418e-05,
|
| 39 |
+
"step": 20
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06276150627615062,
|
| 43 |
+
"grad_norm": 0.06023523282455774,
|
| 44 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 45 |
+
"logits/chosen": -2.6788735389709473,
|
| 46 |
+
"logits/rejected": -2.601842164993286,
|
| 47 |
+
"logps/chosen": -0.9440663456916809,
|
| 48 |
+
"logps/rejected": -1.1466290950775146,
|
| 49 |
+
"loss": 0.6931,
|
| 50 |
+
"rewards/accuracies": 0.543749988079071,
|
| 51 |
+
"rewards/chosen": 0.0002553145168349147,
|
| 52 |
+
"rewards/margins": 6.28855632385239e-05,
|
| 53 |
+
"rewards/rejected": 0.00019242893904447556,
|
| 54 |
+
"step": 30
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.08368200836820083,
|
| 58 |
+
"grad_norm": 0.07304770397797064,
|
| 59 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 60 |
+
"logits/chosen": -2.655829906463623,
|
| 61 |
+
"logits/rejected": -2.574197292327881,
|
| 62 |
+
"logps/chosen": -0.9297592043876648,
|
| 63 |
+
"logps/rejected": -1.1075788736343384,
|
| 64 |
+
"loss": 0.6929,
|
| 65 |
+
"rewards/accuracies": 0.581250011920929,
|
| 66 |
+
"rewards/chosen": 0.0005508430185727775,
|
| 67 |
+
"rewards/margins": 7.169989112298936e-05,
|
| 68 |
+
"rewards/rejected": 0.0004791432002093643,
|
| 69 |
+
"step": 40
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.10460251046025104,
|
| 73 |
+
"grad_norm": 0.10718963835004064,
|
| 74 |
+
"learning_rate": 4.999733114418725e-07,
|
| 75 |
+
"logits/chosen": -2.5783586502075195,
|
| 76 |
+
"logits/rejected": -2.5689890384674072,
|
| 77 |
+
"logps/chosen": -1.1152536869049072,
|
| 78 |
+
"logps/rejected": -1.2834819555282593,
|
| 79 |
+
"loss": 0.6926,
|
| 80 |
+
"rewards/accuracies": 0.6875,
|
| 81 |
+
"rewards/chosen": 0.00013837238657288253,
|
| 82 |
+
"rewards/margins": 0.0015539798187091947,
|
| 83 |
+
"rewards/rejected": -0.0014156072866171598,
|
| 84 |
+
"step": 50
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.12552301255230125,
|
| 88 |
+
"grad_norm": 0.746842093919948,
|
| 89 |
+
"learning_rate": 4.990398100856366e-07,
|
| 90 |
+
"logits/chosen": -2.4468166828155518,
|
| 91 |
+
"logits/rejected": -2.4004383087158203,
|
| 92 |
+
"logps/chosen": -1.0850234031677246,
|
| 93 |
+
"logps/rejected": -1.3491606712341309,
|
| 94 |
+
"loss": 0.6921,
|
| 95 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 96 |
+
"rewards/chosen": -0.0004565033013932407,
|
| 97 |
+
"rewards/margins": 0.00173471518792212,
|
| 98 |
+
"rewards/rejected": -0.0021912185475230217,
|
| 99 |
+
"step": 60
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.14644351464435146,
|
| 103 |
+
"grad_norm": 0.4410404107128935,
|
| 104 |
+
"learning_rate": 4.967775735898179e-07,
|
| 105 |
+
"logits/chosen": -1.8864914178848267,
|
| 106 |
+
"logits/rejected": -1.7537386417388916,
|
| 107 |
+
"logps/chosen": -1.3654557466506958,
|
| 108 |
+
"logps/rejected": -1.9325025081634521,
|
| 109 |
+
"loss": 0.6915,
|
| 110 |
+
"rewards/accuracies": 0.699999988079071,
|
| 111 |
+
"rewards/chosen": -0.0032336972653865814,
|
| 112 |
+
"rewards/margins": 0.004393851850181818,
|
| 113 |
+
"rewards/rejected": -0.007627549115568399,
|
| 114 |
+
"step": 70
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.16736401673640167,
|
| 118 |
+
"grad_norm": 1.1545608141293162,
|
| 119 |
+
"learning_rate": 4.931986719649298e-07,
|
| 120 |
+
"logits/chosen": 0.8977311253547668,
|
| 121 |
+
"logits/rejected": 0.8506044149398804,
|
| 122 |
+
"logps/chosen": -1.9253437519073486,
|
| 123 |
+
"logps/rejected": -2.216801643371582,
|
| 124 |
+
"loss": 0.6897,
|
| 125 |
+
"rewards/accuracies": 0.574999988079071,
|
| 126 |
+
"rewards/chosen": -0.008814895525574684,
|
| 127 |
+
"rewards/margins": 0.0028276063967496157,
|
| 128 |
+
"rewards/rejected": -0.01164250262081623,
|
| 129 |
+
"step": 80
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.18828451882845187,
|
| 133 |
+
"grad_norm": 0.6816222180173854,
|
| 134 |
+
"learning_rate": 4.883222001996351e-07,
|
| 135 |
+
"logits/chosen": 2.8068461418151855,
|
| 136 |
+
"logits/rejected": 3.1994242668151855,
|
| 137 |
+
"logps/chosen": -3.0255942344665527,
|
| 138 |
+
"logps/rejected": -4.264664649963379,
|
| 139 |
+
"loss": 0.6871,
|
| 140 |
+
"rewards/accuracies": 0.581250011920929,
|
| 141 |
+
"rewards/chosen": -0.019984012469649315,
|
| 142 |
+
"rewards/margins": 0.011252423748373985,
|
| 143 |
+
"rewards/rejected": -0.0312364362180233,
|
| 144 |
+
"step": 90
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.20920502092050208,
|
| 148 |
+
"grad_norm": 2.826730795468728,
|
| 149 |
+
"learning_rate": 4.821741763807186e-07,
|
| 150 |
+
"logits/chosen": 3.5266506671905518,
|
| 151 |
+
"logits/rejected": 4.590703010559082,
|
| 152 |
+
"logps/chosen": -8.408990859985352,
|
| 153 |
+
"logps/rejected": -11.962701797485352,
|
| 154 |
+
"loss": 0.6803,
|
| 155 |
+
"rewards/accuracies": 0.668749988079071,
|
| 156 |
+
"rewards/chosen": -0.07384248822927475,
|
| 157 |
+
"rewards/margins": 0.0347314216196537,
|
| 158 |
+
"rewards/rejected": -0.10857391357421875,
|
| 159 |
+
"step": 100
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.20920502092050208,
|
| 163 |
+
"eval_logits/chosen": 4.672222137451172,
|
| 164 |
+
"eval_logits/rejected": 5.161917209625244,
|
| 165 |
+
"eval_logps/chosen": -19.255094528198242,
|
| 166 |
+
"eval_logps/rejected": -23.496896743774414,
|
| 167 |
+
"eval_loss": 0.6740710735321045,
|
| 168 |
+
"eval_rewards/accuracies": 0.60546875,
|
| 169 |
+
"eval_rewards/chosen": -0.18297159671783447,
|
| 170 |
+
"eval_rewards/margins": 0.04080774635076523,
|
| 171 |
+
"eval_rewards/rejected": -0.2237793505191803,
|
| 172 |
+
"eval_runtime": 101.5736,
|
| 173 |
+
"eval_samples_per_second": 19.69,
|
| 174 |
+
"eval_steps_per_second": 0.315,
|
| 175 |
+
"step": 100
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.2301255230125523,
|
| 179 |
+
"grad_norm": 3.6684021455052416,
|
| 180 |
+
"learning_rate": 4.747874028753375e-07,
|
| 181 |
+
"logits/chosen": 4.655301094055176,
|
| 182 |
+
"logits/rejected": 5.133517265319824,
|
| 183 |
+
"logps/chosen": -21.937213897705078,
|
| 184 |
+
"logps/rejected": -27.49126625061035,
|
| 185 |
+
"loss": 0.6637,
|
| 186 |
+
"rewards/accuracies": 0.668749988079071,
|
| 187 |
+
"rewards/chosen": -0.20875534415245056,
|
| 188 |
+
"rewards/margins": 0.05406813696026802,
|
| 189 |
+
"rewards/rejected": -0.2628234922885895,
|
| 190 |
+
"step": 110
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.2510460251046025,
|
| 194 |
+
"grad_norm": 3.7695575614366703,
|
| 195 |
+
"learning_rate": 4.662012913161997e-07,
|
| 196 |
+
"logits/chosen": 3.5462818145751953,
|
| 197 |
+
"logits/rejected": 4.2834906578063965,
|
| 198 |
+
"logps/chosen": -20.185375213623047,
|
| 199 |
+
"logps/rejected": -30.037878036499023,
|
| 200 |
+
"loss": 0.6586,
|
| 201 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 202 |
+
"rewards/chosen": -0.19164811074733734,
|
| 203 |
+
"rewards/margins": 0.09708912670612335,
|
| 204 |
+
"rewards/rejected": -0.2887372672557831,
|
| 205 |
+
"step": 120
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.2719665271966527,
|
| 209 |
+
"grad_norm": 5.06646121428358,
|
| 210 |
+
"learning_rate": 4.5646165232345103e-07,
|
| 211 |
+
"logits/chosen": 3.5566489696502686,
|
| 212 |
+
"logits/rejected": 3.932034969329834,
|
| 213 |
+
"logps/chosen": -26.674022674560547,
|
| 214 |
+
"logps/rejected": -33.86052703857422,
|
| 215 |
+
"loss": 0.6538,
|
| 216 |
+
"rewards/accuracies": 0.59375,
|
| 217 |
+
"rewards/chosen": -0.25690948963165283,
|
| 218 |
+
"rewards/margins": 0.0707937479019165,
|
| 219 |
+
"rewards/rejected": -0.32770323753356934,
|
| 220 |
+
"step": 130
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.2928870292887029,
|
| 224 |
+
"grad_norm": 6.761925793876663,
|
| 225 |
+
"learning_rate": 4.456204510851956e-07,
|
| 226 |
+
"logits/chosen": 3.1402018070220947,
|
| 227 |
+
"logits/rejected": 3.708683490753174,
|
| 228 |
+
"logps/chosen": -21.81682777404785,
|
| 229 |
+
"logps/rejected": -37.612449645996094,
|
| 230 |
+
"loss": 0.6431,
|
| 231 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 232 |
+
"rewards/chosen": -0.20800963044166565,
|
| 233 |
+
"rewards/margins": 0.15619885921478271,
|
| 234 |
+
"rewards/rejected": -0.36420848965644836,
|
| 235 |
+
"step": 140
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.3138075313807531,
|
| 239 |
+
"grad_norm": 8.735857187801553,
|
| 240 |
+
"learning_rate": 4.337355301007335e-07,
|
| 241 |
+
"logits/chosen": 3.099961519241333,
|
| 242 |
+
"logits/rejected": 3.81238055229187,
|
| 243 |
+
"logps/chosen": -23.575672149658203,
|
| 244 |
+
"logps/rejected": -36.89662551879883,
|
| 245 |
+
"loss": 0.6395,
|
| 246 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 247 |
+
"rewards/chosen": -0.2260146588087082,
|
| 248 |
+
"rewards/margins": 0.1314857304096222,
|
| 249 |
+
"rewards/rejected": -0.3575003743171692,
|
| 250 |
+
"step": 150
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.33472803347280333,
|
| 254 |
+
"grad_norm": 6.923549888626696,
|
| 255 |
+
"learning_rate": 4.2087030056579986e-07,
|
| 256 |
+
"logits/chosen": 3.2462105751037598,
|
| 257 |
+
"logits/rejected": 3.581221103668213,
|
| 258 |
+
"logps/chosen": -21.311235427856445,
|
| 259 |
+
"logps/rejected": -36.60784149169922,
|
| 260 |
+
"loss": 0.6369,
|
| 261 |
+
"rewards/accuracies": 0.731249988079071,
|
| 262 |
+
"rewards/chosen": -0.2027742564678192,
|
| 263 |
+
"rewards/margins": 0.151877760887146,
|
| 264 |
+
"rewards/rejected": -0.3546520173549652,
|
| 265 |
+
"step": 160
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.35564853556485354,
|
| 269 |
+
"grad_norm": 7.240479894508596,
|
| 270 |
+
"learning_rate": 4.070934040463998e-07,
|
| 271 |
+
"logits/chosen": 3.023836612701416,
|
| 272 |
+
"logits/rejected": 3.3588485717773438,
|
| 273 |
+
"logps/chosen": -23.696680068969727,
|
| 274 |
+
"logps/rejected": -36.59798049926758,
|
| 275 |
+
"loss": 0.6339,
|
| 276 |
+
"rewards/accuracies": 0.675000011920929,
|
| 277 |
+
"rewards/chosen": -0.22723570466041565,
|
| 278 |
+
"rewards/margins": 0.12730170786380768,
|
| 279 |
+
"rewards/rejected": -0.3545374274253845,
|
| 280 |
+
"step": 170
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.37656903765690375,
|
| 284 |
+
"grad_norm": 9.14137324995646,
|
| 285 |
+
"learning_rate": 3.9247834624635404e-07,
|
| 286 |
+
"logits/chosen": 2.783609390258789,
|
| 287 |
+
"logits/rejected": 3.172799587249756,
|
| 288 |
+
"logps/chosen": -22.1041259765625,
|
| 289 |
+
"logps/rejected": -37.87432098388672,
|
| 290 |
+
"loss": 0.6295,
|
| 291 |
+
"rewards/accuracies": 0.699999988079071,
|
| 292 |
+
"rewards/chosen": -0.2107868194580078,
|
| 293 |
+
"rewards/margins": 0.15683028101921082,
|
| 294 |
+
"rewards/rejected": -0.36761707067489624,
|
| 295 |
+
"step": 180
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.39748953974895396,
|
| 299 |
+
"grad_norm": 9.772844987285321,
|
| 300 |
+
"learning_rate": 3.7710310482256523e-07,
|
| 301 |
+
"logits/chosen": 2.702998399734497,
|
| 302 |
+
"logits/rejected": 2.8405003547668457,
|
| 303 |
+
"logps/chosen": -26.399200439453125,
|
| 304 |
+
"logps/rejected": -40.94459915161133,
|
| 305 |
+
"loss": 0.6258,
|
| 306 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 307 |
+
"rewards/chosen": -0.25330352783203125,
|
| 308 |
+
"rewards/margins": 0.14452621340751648,
|
| 309 |
+
"rewards/rejected": -0.3978297710418701,
|
| 310 |
+
"step": 190
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.41841004184100417,
|
| 314 |
+
"grad_norm": 8.573703981557731,
|
| 315 |
+
"learning_rate": 3.610497133404795e-07,
|
| 316 |
+
"logits/chosen": 2.867384433746338,
|
| 317 |
+
"logits/rejected": 3.402123212814331,
|
| 318 |
+
"logps/chosen": -25.27474021911621,
|
| 319 |
+
"logps/rejected": -39.822689056396484,
|
| 320 |
+
"loss": 0.6333,
|
| 321 |
+
"rewards/accuracies": 0.71875,
|
| 322 |
+
"rewards/chosen": -0.24253730475902557,
|
| 323 |
+
"rewards/margins": 0.14454875886440277,
|
| 324 |
+
"rewards/rejected": -0.38708609342575073,
|
| 325 |
+
"step": 200
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.41841004184100417,
|
| 329 |
+
"eval_logits/chosen": 2.205869197845459,
|
| 330 |
+
"eval_logits/rejected": 2.0618207454681396,
|
| 331 |
+
"eval_logps/chosen": -27.025375366210938,
|
| 332 |
+
"eval_logps/rejected": -43.748695373535156,
|
| 333 |
+
"eval_loss": 0.6317887902259827,
|
| 334 |
+
"eval_rewards/accuracies": 0.70703125,
|
| 335 |
+
"eval_rewards/chosen": -0.2606744170188904,
|
| 336 |
+
"eval_rewards/margins": 0.16562291979789734,
|
| 337 |
+
"eval_rewards/rejected": -0.4262973666191101,
|
| 338 |
+
"eval_runtime": 102.633,
|
| 339 |
+
"eval_samples_per_second": 19.487,
|
| 340 |
+
"eval_steps_per_second": 0.312,
|
| 341 |
+
"step": 200
|
| 342 |
+
},
|
| 343 |
+
{
|
| 344 |
+
"epoch": 0.4393305439330544,
|
| 345 |
+
"grad_norm": 10.103691000737252,
|
| 346 |
+
"learning_rate": 3.4440382358952115e-07,
|
| 347 |
+
"logits/chosen": 2.632490634918213,
|
| 348 |
+
"logits/rejected": 2.8943495750427246,
|
| 349 |
+
"logps/chosen": -26.576526641845703,
|
| 350 |
+
"logps/rejected": -39.573448181152344,
|
| 351 |
+
"loss": 0.6294,
|
| 352 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 353 |
+
"rewards/chosen": -0.2549188733100891,
|
| 354 |
+
"rewards/margins": 0.12898248434066772,
|
| 355 |
+
"rewards/rejected": -0.38390129804611206,
|
| 356 |
+
"step": 210
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.4602510460251046,
|
| 360 |
+
"grad_norm": 8.68088284973258,
|
| 361 |
+
"learning_rate": 3.272542485937368e-07,
|
| 362 |
+
"logits/chosen": 2.8408799171447754,
|
| 363 |
+
"logits/rejected": 3.1735482215881348,
|
| 364 |
+
"logps/chosen": -28.81484031677246,
|
| 365 |
+
"logps/rejected": -44.7733268737793,
|
| 366 |
+
"loss": 0.6277,
|
| 367 |
+
"rewards/accuracies": 0.65625,
|
| 368 |
+
"rewards/chosen": -0.2777389883995056,
|
| 369 |
+
"rewards/margins": 0.15828576683998108,
|
| 370 |
+
"rewards/rejected": -0.43602481484413147,
|
| 371 |
+
"step": 220
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.4811715481171548,
|
| 375 |
+
"grad_norm": 9.381367466525985,
|
| 376 |
+
"learning_rate": 3.096924887558854e-07,
|
| 377 |
+
"logits/chosen": 2.756274700164795,
|
| 378 |
+
"logits/rejected": 3.042253017425537,
|
| 379 |
+
"logps/chosen": -20.681415557861328,
|
| 380 |
+
"logps/rejected": -41.724674224853516,
|
| 381 |
+
"loss": 0.6228,
|
| 382 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 383 |
+
"rewards/chosen": -0.195968359708786,
|
| 384 |
+
"rewards/margins": 0.20941869914531708,
|
| 385 |
+
"rewards/rejected": -0.4053870737552643,
|
| 386 |
+
"step": 230
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.502092050209205,
|
| 390 |
+
"grad_norm": 9.698557591113461,
|
| 391 |
+
"learning_rate": 2.9181224366319943e-07,
|
| 392 |
+
"logits/chosen": 2.9523372650146484,
|
| 393 |
+
"logits/rejected": 3.568504810333252,
|
| 394 |
+
"logps/chosen": -25.638051986694336,
|
| 395 |
+
"logps/rejected": -40.016014099121094,
|
| 396 |
+
"loss": 0.6288,
|
| 397 |
+
"rewards/accuracies": 0.668749988079071,
|
| 398 |
+
"rewards/chosen": -0.24575933814048767,
|
| 399 |
+
"rewards/margins": 0.14226052165031433,
|
| 400 |
+
"rewards/rejected": -0.3880198299884796,
|
| 401 |
+
"step": 240
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.5230125523012552,
|
| 405 |
+
"grad_norm": 7.996022666238754,
|
| 406 |
+
"learning_rate": 2.7370891215954565e-07,
|
| 407 |
+
"logits/chosen": 2.842616558074951,
|
| 408 |
+
"logits/rejected": 3.386730670928955,
|
| 409 |
+
"logps/chosen": -16.753459930419922,
|
| 410 |
+
"logps/rejected": -32.727378845214844,
|
| 411 |
+
"loss": 0.6271,
|
| 412 |
+
"rewards/accuracies": 0.706250011920929,
|
| 413 |
+
"rewards/chosen": -0.157147616147995,
|
| 414 |
+
"rewards/margins": 0.15919244289398193,
|
| 415 |
+
"rewards/rejected": -0.31634002923965454,
|
| 416 |
+
"step": 250
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5439330543933054,
|
| 420 |
+
"grad_norm": 9.49367392875526,
|
| 421 |
+
"learning_rate": 2.55479083351317e-07,
|
| 422 |
+
"logits/chosen": 2.8096323013305664,
|
| 423 |
+
"logits/rejected": 2.8222105503082275,
|
| 424 |
+
"logps/chosen": -26.005813598632812,
|
| 425 |
+
"logps/rejected": -45.71054458618164,
|
| 426 |
+
"loss": 0.6241,
|
| 427 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 428 |
+
"rewards/chosen": -0.24941372871398926,
|
| 429 |
+
"rewards/margins": 0.19658246636390686,
|
| 430 |
+
"rewards/rejected": -0.4459961950778961,
|
| 431 |
+
"step": 260
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.5648535564853556,
|
| 435 |
+
"grad_norm": 14.638817153423537,
|
| 436 |
+
"learning_rate": 2.3722002126275822e-07,
|
| 437 |
+
"logits/chosen": 2.8568336963653564,
|
| 438 |
+
"logits/rejected": 3.19636869430542,
|
| 439 |
+
"logps/chosen": -22.6005916595459,
|
| 440 |
+
"logps/rejected": -40.90196990966797,
|
| 441 |
+
"loss": 0.6224,
|
| 442 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 443 |
+
"rewards/chosen": -0.21595272421836853,
|
| 444 |
+
"rewards/margins": 0.1823262870311737,
|
| 445 |
+
"rewards/rejected": -0.3982790410518646,
|
| 446 |
+
"step": 270
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.5857740585774058,
|
| 450 |
+
"grad_norm": 10.46787470104777,
|
| 451 |
+
"learning_rate": 2.19029145890313e-07,
|
| 452 |
+
"logits/chosen": 2.681917428970337,
|
| 453 |
+
"logits/rejected": 2.849825382232666,
|
| 454 |
+
"logps/chosen": -21.89919090270996,
|
| 455 |
+
"logps/rejected": -37.612815856933594,
|
| 456 |
+
"loss": 0.6265,
|
| 457 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 458 |
+
"rewards/chosen": -0.2091476172208786,
|
| 459 |
+
"rewards/margins": 0.1555161476135254,
|
| 460 |
+
"rewards/rejected": -0.3646637797355652,
|
| 461 |
+
"step": 280
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.606694560669456,
|
| 465 |
+
"grad_norm": 7.74921044050114,
|
| 466 |
+
"learning_rate": 2.0100351342479216e-07,
|
| 467 |
+
"logits/chosen": 2.8385300636291504,
|
| 468 |
+
"logits/rejected": 2.8731513023376465,
|
| 469 |
+
"logps/chosen": -24.38838768005371,
|
| 470 |
+
"logps/rejected": -42.81266784667969,
|
| 471 |
+
"loss": 0.6231,
|
| 472 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 473 |
+
"rewards/chosen": -0.23373980820178986,
|
| 474 |
+
"rewards/margins": 0.18318690359592438,
|
| 475 |
+
"rewards/rejected": -0.41692671179771423,
|
| 476 |
+
"step": 290
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.6276150627615062,
|
| 480 |
+
"grad_norm": 8.46554493519494,
|
| 481 |
+
"learning_rate": 1.8323929841460178e-07,
|
| 482 |
+
"logits/chosen": 3.3211822509765625,
|
| 483 |
+
"logits/rejected": 3.4708034992218018,
|
| 484 |
+
"logps/chosen": -22.359912872314453,
|
| 485 |
+
"logps/rejected": -38.96550369262695,
|
| 486 |
+
"loss": 0.6202,
|
| 487 |
+
"rewards/accuracies": 0.699999988079071,
|
| 488 |
+
"rewards/chosen": -0.21357199549674988,
|
| 489 |
+
"rewards/margins": 0.16511210799217224,
|
| 490 |
+
"rewards/rejected": -0.3786841034889221,
|
| 491 |
+
"step": 300
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6276150627615062,
|
| 495 |
+
"eval_logits/chosen": 2.999788284301758,
|
| 496 |
+
"eval_logits/rejected": 3.0356078147888184,
|
| 497 |
+
"eval_logps/chosen": -20.559364318847656,
|
| 498 |
+
"eval_logps/rejected": -40.0822639465332,
|
| 499 |
+
"eval_loss": 0.6255786418914795,
|
| 500 |
+
"eval_rewards/accuracies": 0.71875,
|
| 501 |
+
"eval_rewards/chosen": -0.19601428508758545,
|
| 502 |
+
"eval_rewards/margins": 0.1936187446117401,
|
| 503 |
+
"eval_rewards/rejected": -0.38963305950164795,
|
| 504 |
+
"eval_runtime": 102.0512,
|
| 505 |
+
"eval_samples_per_second": 19.598,
|
| 506 |
+
"eval_steps_per_second": 0.314,
|
| 507 |
+
"step": 300
|
| 508 |
+
},
|
| 509 |
+
{
|
| 510 |
+
"epoch": 0.6485355648535565,
|
| 511 |
+
"grad_norm": 9.417564911888258,
|
| 512 |
+
"learning_rate": 1.6583128063291573e-07,
|
| 513 |
+
"logits/chosen": 2.8539745807647705,
|
| 514 |
+
"logits/rejected": 3.0914368629455566,
|
| 515 |
+
"logps/chosen": -20.804203033447266,
|
| 516 |
+
"logps/rejected": -41.28700637817383,
|
| 517 |
+
"loss": 0.621,
|
| 518 |
+
"rewards/accuracies": 0.75,
|
| 519 |
+
"rewards/chosen": -0.1979689598083496,
|
| 520 |
+
"rewards/margins": 0.20332905650138855,
|
| 521 |
+
"rewards/rejected": -0.40129804611206055,
|
| 522 |
+
"step": 310
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.6694560669456067,
|
| 526 |
+
"grad_norm": 8.668181751551456,
|
| 527 |
+
"learning_rate": 1.488723393865766e-07,
|
| 528 |
+
"logits/chosen": 2.763029098510742,
|
| 529 |
+
"logits/rejected": 3.5786728858947754,
|
| 530 |
+
"logps/chosen": -25.161691665649414,
|
| 531 |
+
"logps/rejected": -41.82555389404297,
|
| 532 |
+
"loss": 0.6159,
|
| 533 |
+
"rewards/accuracies": 0.706250011920929,
|
| 534 |
+
"rewards/chosen": -0.24154594540596008,
|
| 535 |
+
"rewards/margins": 0.16512060165405273,
|
| 536 |
+
"rewards/rejected": -0.4066665768623352,
|
| 537 |
+
"step": 320
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.6903765690376569,
|
| 541 |
+
"grad_norm": 9.135541705526716,
|
| 542 |
+
"learning_rate": 1.3245295796480788e-07,
|
| 543 |
+
"logits/chosen": 2.6453309059143066,
|
| 544 |
+
"logits/rejected": 3.213120937347412,
|
| 545 |
+
"logps/chosen": -28.319080352783203,
|
| 546 |
+
"logps/rejected": -45.0162467956543,
|
| 547 |
+
"loss": 0.6209,
|
| 548 |
+
"rewards/accuracies": 0.65625,
|
| 549 |
+
"rewards/chosen": -0.27260464429855347,
|
| 550 |
+
"rewards/margins": 0.16600076854228973,
|
| 551 |
+
"rewards/rejected": -0.438605397939682,
|
| 552 |
+
"step": 330
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.7112970711297071,
|
| 556 |
+
"grad_norm": 9.485727128957436,
|
| 557 |
+
"learning_rate": 1.1666074087171627e-07,
|
| 558 |
+
"logits/chosen": 3.553588390350342,
|
| 559 |
+
"logits/rejected": 3.5263543128967285,
|
| 560 |
+
"logps/chosen": -23.71384048461914,
|
| 561 |
+
"logps/rejected": -40.1160774230957,
|
| 562 |
+
"loss": 0.6268,
|
| 563 |
+
"rewards/accuracies": 0.675000011920929,
|
| 564 |
+
"rewards/chosen": -0.22761042416095734,
|
| 565 |
+
"rewards/margins": 0.16276967525482178,
|
| 566 |
+
"rewards/rejected": -0.3903800845146179,
|
| 567 |
+
"step": 340
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.7322175732217573,
|
| 571 |
+
"grad_norm": 9.579729502767362,
|
| 572 |
+
"learning_rate": 1.0157994641835734e-07,
|
| 573 |
+
"logits/chosen": 3.2801671028137207,
|
| 574 |
+
"logits/rejected": 3.6302173137664795,
|
| 575 |
+
"logps/chosen": -23.160703659057617,
|
| 576 |
+
"logps/rejected": -39.75481414794922,
|
| 577 |
+
"loss": 0.6148,
|
| 578 |
+
"rewards/accuracies": 0.706250011920929,
|
| 579 |
+
"rewards/chosen": -0.22147643566131592,
|
| 580 |
+
"rewards/margins": 0.1643747091293335,
|
| 581 |
+
"rewards/rejected": -0.3858511745929718,
|
| 582 |
+
"step": 350
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.7531380753138075,
|
| 586 |
+
"grad_norm": 9.56792620122,
|
| 587 |
+
"learning_rate": 8.729103716819111e-08,
|
| 588 |
+
"logits/chosen": 2.997981548309326,
|
| 589 |
+
"logits/rejected": 3.285050868988037,
|
| 590 |
+
"logps/chosen": -25.180973052978516,
|
| 591 |
+
"logps/rejected": -41.596805572509766,
|
| 592 |
+
"loss": 0.6213,
|
| 593 |
+
"rewards/accuracies": 0.668749988079071,
|
| 594 |
+
"rewards/chosen": -0.2415219247341156,
|
| 595 |
+
"rewards/margins": 0.16281357407569885,
|
| 596 |
+
"rewards/rejected": -0.40433549880981445,
|
| 597 |
+
"step": 360
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.7740585774058577,
|
| 601 |
+
"grad_norm": 9.46320162833878,
|
| 602 |
+
"learning_rate": 7.387025063449081e-08,
|
| 603 |
+
"logits/chosen": 2.827638864517212,
|
| 604 |
+
"logits/rejected": 3.168362855911255,
|
| 605 |
+
"logps/chosen": -20.08965301513672,
|
| 606 |
+
"logps/rejected": -41.27653121948242,
|
| 607 |
+
"loss": 0.6151,
|
| 608 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 609 |
+
"rewards/chosen": -0.1905418187379837,
|
| 610 |
+
"rewards/margins": 0.21073243021965027,
|
| 611 |
+
"rewards/rejected": -0.40127426385879517,
|
| 612 |
+
"step": 370
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.7949790794979079,
|
| 616 |
+
"grad_norm": 11.287573566387135,
|
| 617 |
+
"learning_rate": 6.138919252022435e-08,
|
| 618 |
+
"logits/chosen": 2.806427001953125,
|
| 619 |
+
"logits/rejected": 2.9862289428710938,
|
| 620 |
+
"logps/chosen": -26.241718292236328,
|
| 621 |
+
"logps/rejected": -40.23133850097656,
|
| 622 |
+
"loss": 0.6247,
|
| 623 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 624 |
+
"rewards/chosen": -0.25202974677085876,
|
| 625 |
+
"rewards/margins": 0.1393345296382904,
|
| 626 |
+
"rewards/rejected": -0.39136427640914917,
|
| 627 |
+
"step": 380
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.8158995815899581,
|
| 631 |
+
"grad_norm": 13.141931276110663,
|
| 632 |
+
"learning_rate": 4.991445467064689e-08,
|
| 633 |
+
"logits/chosen": 3.050666093826294,
|
| 634 |
+
"logits/rejected": 3.1723926067352295,
|
| 635 |
+
"logps/chosen": -28.155725479125977,
|
| 636 |
+
"logps/rejected": -46.16926193237305,
|
| 637 |
+
"loss": 0.6197,
|
| 638 |
+
"rewards/accuracies": 0.706250011920929,
|
| 639 |
+
"rewards/chosen": -0.2713962495326996,
|
| 640 |
+
"rewards/margins": 0.17873115837574005,
|
| 641 |
+
"rewards/rejected": -0.45012742280960083,
|
| 642 |
+
"step": 390
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.8368200836820083,
|
| 646 |
+
"grad_norm": 7.223640139685516,
|
| 647 |
+
"learning_rate": 3.9507259776993954e-08,
|
| 648 |
+
"logits/chosen": 2.904824733734131,
|
| 649 |
+
"logits/rejected": 2.7292914390563965,
|
| 650 |
+
"logps/chosen": -28.01041603088379,
|
| 651 |
+
"logps/rejected": -43.947017669677734,
|
| 652 |
+
"loss": 0.6195,
|
| 653 |
+
"rewards/accuracies": 0.699999988079071,
|
| 654 |
+
"rewards/chosen": -0.27024754881858826,
|
| 655 |
+
"rewards/margins": 0.1586446762084961,
|
| 656 |
+
"rewards/rejected": -0.42889222502708435,
|
| 657 |
+
"step": 400
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.8368200836820083,
|
| 661 |
+
"eval_logits/chosen": 2.7595362663269043,
|
| 662 |
+
"eval_logits/rejected": 2.7050070762634277,
|
| 663 |
+
"eval_logps/chosen": -23.82366180419922,
|
| 664 |
+
"eval_logps/rejected": -43.26374435424805,
|
| 665 |
+
"eval_loss": 0.6218914985656738,
|
| 666 |
+
"eval_rewards/accuracies": 0.71875,
|
| 667 |
+
"eval_rewards/chosen": -0.2286572903394699,
|
| 668 |
+
"eval_rewards/margins": 0.19279056787490845,
|
| 669 |
+
"eval_rewards/rejected": -0.42144784331321716,
|
| 670 |
+
"eval_runtime": 101.9605,
|
| 671 |
+
"eval_samples_per_second": 19.615,
|
| 672 |
+
"eval_steps_per_second": 0.314,
|
| 673 |
+
"step": 400
|
| 674 |
+
},
|
| 675 |
+
{
|
| 676 |
+
"epoch": 0.8577405857740585,
|
| 677 |
+
"grad_norm": 9.798220212045168,
|
| 678 |
+
"learning_rate": 3.022313472693447e-08,
|
| 679 |
+
"logits/chosen": 2.254669427871704,
|
| 680 |
+
"logits/rejected": 2.458228588104248,
|
| 681 |
+
"logps/chosen": -25.69635009765625,
|
| 682 |
+
"logps/rejected": -43.98773956298828,
|
| 683 |
+
"loss": 0.6191,
|
| 684 |
+
"rewards/accuracies": 0.706250011920929,
|
| 685 |
+
"rewards/chosen": -0.24772438406944275,
|
| 686 |
+
"rewards/margins": 0.18161071836948395,
|
| 687 |
+
"rewards/rejected": -0.42933517694473267,
|
| 688 |
+
"step": 410
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.8786610878661087,
|
| 692 |
+
"grad_norm": 9.457071183698215,
|
| 693 |
+
"learning_rate": 2.2111614344599684e-08,
|
| 694 |
+
"logits/chosen": 2.807976484298706,
|
| 695 |
+
"logits/rejected": 3.5379626750946045,
|
| 696 |
+
"logps/chosen": -23.7137393951416,
|
| 697 |
+
"logps/rejected": -40.27167510986328,
|
| 698 |
+
"loss": 0.6151,
|
| 699 |
+
"rewards/accuracies": 0.731249988079071,
|
| 700 |
+
"rewards/chosen": -0.22669024765491486,
|
| 701 |
+
"rewards/margins": 0.16417010128498077,
|
| 702 |
+
"rewards/rejected": -0.390860378742218,
|
| 703 |
+
"step": 420
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.899581589958159,
|
| 707 |
+
"grad_norm": 11.679593436076072,
|
| 708 |
+
"learning_rate": 1.521597710086439e-08,
|
| 709 |
+
"logits/chosen": 2.7207865715026855,
|
| 710 |
+
"logits/rejected": 3.395256757736206,
|
| 711 |
+
"logps/chosen": -21.971521377563477,
|
| 712 |
+
"logps/rejected": -40.920936584472656,
|
| 713 |
+
"loss": 0.6232,
|
| 714 |
+
"rewards/accuracies": 0.699999988079071,
|
| 715 |
+
"rewards/chosen": -0.20926229655742645,
|
| 716 |
+
"rewards/margins": 0.18839535117149353,
|
| 717 |
+
"rewards/rejected": -0.3976576328277588,
|
| 718 |
+
"step": 430
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.9205020920502092,
|
| 722 |
+
"grad_norm": 8.417656516106161,
|
| 723 |
+
"learning_rate": 9.57301420397924e-09,
|
| 724 |
+
"logits/chosen": 3.1824350357055664,
|
| 725 |
+
"logits/rejected": 3.9073855876922607,
|
| 726 |
+
"logps/chosen": -25.478689193725586,
|
| 727 |
+
"logps/rejected": -40.70270538330078,
|
| 728 |
+
"loss": 0.6223,
|
| 729 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 730 |
+
"rewards/chosen": -0.24391062557697296,
|
| 731 |
+
"rewards/margins": 0.15080411732196808,
|
| 732 |
+
"rewards/rejected": -0.3947147727012634,
|
| 733 |
+
"step": 440
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9414225941422594,
|
| 737 |
+
"grad_norm": 8.425685268345296,
|
| 738 |
+
"learning_rate": 5.212833302556258e-09,
|
| 739 |
+
"logits/chosen": 3.0830206871032715,
|
| 740 |
+
"logits/rejected": 3.034968614578247,
|
| 741 |
+
"logps/chosen": -21.253185272216797,
|
| 742 |
+
"logps/rejected": -42.495853424072266,
|
| 743 |
+
"loss": 0.6214,
|
| 744 |
+
"rewards/accuracies": 0.768750011920929,
|
| 745 |
+
"rewards/chosen": -0.20228211581707,
|
| 746 |
+
"rewards/margins": 0.21119949221611023,
|
| 747 |
+
"rewards/rejected": -0.41348162293434143,
|
| 748 |
+
"step": 450
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.9623430962343096,
|
| 752 |
+
"grad_norm": 7.6775915287366585,
|
| 753 |
+
"learning_rate": 2.158697848236607e-09,
|
| 754 |
+
"logits/chosen": 3.344025135040283,
|
| 755 |
+
"logits/rejected": 3.607654571533203,
|
| 756 |
+
"logps/chosen": -26.998760223388672,
|
| 757 |
+
"logps/rejected": -40.048553466796875,
|
| 758 |
+
"loss": 0.6173,
|
| 759 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 760 |
+
"rewards/chosen": -0.26005202531814575,
|
| 761 |
+
"rewards/margins": 0.12917304039001465,
|
| 762 |
+
"rewards/rejected": -0.3892250657081604,
|
| 763 |
+
"step": 460
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.9832635983263598,
|
| 767 |
+
"grad_norm": 9.905482891237243,
|
| 768 |
+
"learning_rate": 4.269029751107489e-10,
|
| 769 |
+
"logits/chosen": 3.274850368499756,
|
| 770 |
+
"logits/rejected": 3.1045289039611816,
|
| 771 |
+
"logps/chosen": -22.779346466064453,
|
| 772 |
+
"logps/rejected": -39.72507095336914,
|
| 773 |
+
"loss": 0.6284,
|
| 774 |
+
"rewards/accuracies": 0.706250011920929,
|
| 775 |
+
"rewards/chosen": -0.217888742685318,
|
| 776 |
+
"rewards/margins": 0.168381929397583,
|
| 777 |
+
"rewards/rejected": -0.386270672082901,
|
| 778 |
+
"step": 470
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 1.0,
|
| 782 |
+
"step": 478,
|
| 783 |
+
"total_flos": 0.0,
|
| 784 |
+
"train_loss": 0.6403242514223234,
|
| 785 |
+
"train_runtime": 11960.3252,
|
| 786 |
+
"train_samples_per_second": 5.111,
|
| 787 |
+
"train_steps_per_second": 0.04
|
| 788 |
+
}
|
| 789 |
+
],
|
| 790 |
+
"logging_steps": 10,
|
| 791 |
+
"max_steps": 478,
|
| 792 |
+
"num_input_tokens_seen": 0,
|
| 793 |
+
"num_train_epochs": 1,
|
| 794 |
+
"save_steps": 100,
|
| 795 |
+
"stateful_callbacks": {
|
| 796 |
+
"TrainerControl": {
|
| 797 |
+
"args": {
|
| 798 |
+
"should_epoch_stop": false,
|
| 799 |
+
"should_evaluate": false,
|
| 800 |
+
"should_log": false,
|
| 801 |
+
"should_save": true,
|
| 802 |
+
"should_training_stop": true
|
| 803 |
+
},
|
| 804 |
+
"attributes": {}
|
| 805 |
+
}
|
| 806 |
+
},
|
| 807 |
+
"total_flos": 0.0,
|
| 808 |
+
"train_batch_size": 8,
|
| 809 |
+
"trial_name": null,
|
| 810 |
+
"trial_params": null
|
| 811 |
+
}
|