Model save
Browse files
README.md
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 3 |
+
library_name: peft
|
| 4 |
+
license: apache-2.0
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- dpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-dpo-qlora-gpt4-5e-6
|
| 11 |
+
results: []
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
+
|
| 17 |
+
# zephyr-dpo-qlora-gpt4-5e-6
|
| 18 |
+
|
| 19 |
+
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.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.9371
|
| 22 |
+
- Rewards/chosen: -6.0411
|
| 23 |
+
- Rewards/rejected: -6.9811
|
| 24 |
+
- Rewards/accuracies: 0.6450
|
| 25 |
+
- Rewards/margins: 0.9400
|
| 26 |
+
- Rewards/margins Max: 4.7437
|
| 27 |
+
- Rewards/margins Min: -3.0368
|
| 28 |
+
- Rewards/margins Std: 2.6550
|
| 29 |
+
- Logps/rejected: -956.6877
|
| 30 |
+
- Logps/chosen: -888.7039
|
| 31 |
+
- Logits/rejected: -1.4320
|
| 32 |
+
- Logits/chosen: -1.4909
|
| 33 |
+
|
| 34 |
+
## Model description
|
| 35 |
+
|
| 36 |
+
More information needed
|
| 37 |
+
|
| 38 |
+
## Intended uses & limitations
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Training and evaluation data
|
| 43 |
+
|
| 44 |
+
More information needed
|
| 45 |
+
|
| 46 |
+
## Training procedure
|
| 47 |
+
|
| 48 |
+
### Training hyperparameters
|
| 49 |
+
|
| 50 |
+
The following hyperparameters were used during training:
|
| 51 |
+
- learning_rate: 5e-06
|
| 52 |
+
- train_batch_size: 4
|
| 53 |
+
- eval_batch_size: 8
|
| 54 |
+
- seed: 42
|
| 55 |
+
- distributed_type: multi-GPU
|
| 56 |
+
- num_devices: 2
|
| 57 |
+
- gradient_accumulation_steps: 2
|
| 58 |
+
- total_train_batch_size: 16
|
| 59 |
+
- total_eval_batch_size: 16
|
| 60 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 61 |
+
- lr_scheduler_type: cosine
|
| 62 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 63 |
+
- num_epochs: 1
|
| 64 |
+
|
| 65 |
+
### Training results
|
| 66 |
+
|
| 67 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Rewards/margins Max | Rewards/margins Min | Rewards/margins Std | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 68 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:-------------------:|:-------------------:|:-------------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 69 |
+
| 0.3843 | 0.28 | 100 | 0.6754 | -0.8222 | -0.9456 | 0.6190 | 0.1235 | 0.8275 | -0.5849 | 0.4661 | -353.1436 | -366.8085 | -2.4707 | -2.5001 |
|
| 70 |
+
| 0.1345 | 0.56 | 200 | 0.9115 | -6.6830 | -7.3423 | 0.6230 | 0.6594 | 4.1300 | -2.9206 | 2.3615 | -992.8121 | -952.8915 | -1.3494 | -1.4037 |
|
| 71 |
+
| 0.0467 | 0.85 | 300 | 0.9371 | -6.0411 | -6.9811 | 0.6450 | 0.9400 | 4.7437 | -3.0368 | 2.6550 | -956.6877 | -888.7039 | -1.4320 | -1.4909 |
|
| 72 |
+
|
| 73 |
+
|
| 74 |
+
### Framework versions
|
| 75 |
+
|
| 76 |
+
- PEFT 0.7.1
|
| 77 |
+
- Transformers 4.39.0.dev0
|
| 78 |
+
- Pytorch 2.1.2+cu121
|
| 79 |
+
- Datasets 2.14.6
|
| 80 |
+
- Tokenizers 0.15.2
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 671150064
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5ca25177e09688d083d9a7090f82048f51b4951b11edc7a99ba9ee07dddf23bb
|
| 3 |
size 671150064
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.24729210323011372,
|
| 4 |
+
"train_runtime": 4030.3896,
|
| 5 |
+
"train_samples": 5678,
|
| 6 |
+
"train_samples_per_second": 1.409,
|
| 7 |
+
"train_steps_per_second": 0.088
|
| 8 |
+
}
|
runs/Jul28_07-09-30_notebook-deployment-48-7d9b6c99-khd85/events.out.tfevents.1722150668.notebook-deployment-48-7d9b6c99-khd85.1186377.0
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d856c33d5decd4f3ff40781d5cc8789a2f65437572cb6585ff3bb04e6beaaf7c
|
| 3 |
+
size 39947
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.24729210323011372,
|
| 4 |
+
"train_runtime": 4030.3896,
|
| 5 |
+
"train_samples": 5678,
|
| 6 |
+
"train_samples_per_second": 1.409,
|
| 7 |
+
"train_steps_per_second": 0.088
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,735 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 355,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0,
|
| 13 |
+
"grad_norm": 2.1560881180741176,
|
| 14 |
+
"learning_rate": 1.3888888888888888e-07,
|
| 15 |
+
"logits/chosen": -2.804708957672119,
|
| 16 |
+
"logits/rejected": -2.8150453567504883,
|
| 17 |
+
"logps/chosen": -217.97438049316406,
|
| 18 |
+
"logps/rejected": -216.58865356445312,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.0,
|
| 21 |
+
"rewards/chosen": 0.0,
|
| 22 |
+
"rewards/margins": 0.0,
|
| 23 |
+
"rewards/margins_max": 0.0,
|
| 24 |
+
"rewards/margins_min": 0.0,
|
| 25 |
+
"rewards/margins_std": 0.0,
|
| 26 |
+
"rewards/rejected": 0.0,
|
| 27 |
+
"step": 1
|
| 28 |
+
},
|
| 29 |
+
{
|
| 30 |
+
"epoch": 0.03,
|
| 31 |
+
"grad_norm": 10.20672067373361,
|
| 32 |
+
"learning_rate": 1.3888888888888892e-06,
|
| 33 |
+
"logits/chosen": -2.8844592571258545,
|
| 34 |
+
"logits/rejected": -2.79921555519104,
|
| 35 |
+
"logps/chosen": -366.6466064453125,
|
| 36 |
+
"logps/rejected": -275.4883728027344,
|
| 37 |
+
"loss": 0.6927,
|
| 38 |
+
"rewards/accuracies": 0.5138888955116272,
|
| 39 |
+
"rewards/chosen": 0.0008819872164167464,
|
| 40 |
+
"rewards/margins": 0.0012480419827625155,
|
| 41 |
+
"rewards/margins_max": 0.005487875547260046,
|
| 42 |
+
"rewards/margins_min": -0.0020528410095721483,
|
| 43 |
+
"rewards/margins_std": 0.003371911123394966,
|
| 44 |
+
"rewards/rejected": -0.00036605476634576917,
|
| 45 |
+
"step": 10
|
| 46 |
+
},
|
| 47 |
+
{
|
| 48 |
+
"epoch": 0.06,
|
| 49 |
+
"grad_norm": 2.007156324220208,
|
| 50 |
+
"learning_rate": 2.7777777777777783e-06,
|
| 51 |
+
"logits/chosen": -2.7444252967834473,
|
| 52 |
+
"logits/rejected": -2.6909613609313965,
|
| 53 |
+
"logps/chosen": -328.3125915527344,
|
| 54 |
+
"logps/rejected": -217.0628204345703,
|
| 55 |
+
"loss": 0.6867,
|
| 56 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 57 |
+
"rewards/chosen": 0.008345814421772957,
|
| 58 |
+
"rewards/margins": 0.01153595745563507,
|
| 59 |
+
"rewards/margins_max": 0.026045221835374832,
|
| 60 |
+
"rewards/margins_min": 0.0008641376043669879,
|
| 61 |
+
"rewards/margins_std": 0.011482259258627892,
|
| 62 |
+
"rewards/rejected": -0.003190141636878252,
|
| 63 |
+
"step": 20
|
| 64 |
+
},
|
| 65 |
+
{
|
| 66 |
+
"epoch": 0.08,
|
| 67 |
+
"grad_norm": 2.158164459848784,
|
| 68 |
+
"learning_rate": 4.166666666666667e-06,
|
| 69 |
+
"logits/chosen": -2.825068712234497,
|
| 70 |
+
"logits/rejected": -2.7449638843536377,
|
| 71 |
+
"logps/chosen": -325.515869140625,
|
| 72 |
+
"logps/rejected": -234.0052490234375,
|
| 73 |
+
"loss": 0.6655,
|
| 74 |
+
"rewards/accuracies": 0.800000011920929,
|
| 75 |
+
"rewards/chosen": 0.0376933328807354,
|
| 76 |
+
"rewards/margins": 0.0479232482612133,
|
| 77 |
+
"rewards/margins_max": 0.11641480773687363,
|
| 78 |
+
"rewards/margins_min": 0.0012140764156356454,
|
| 79 |
+
"rewards/margins_std": 0.05392361804842949,
|
| 80 |
+
"rewards/rejected": -0.010229915380477905,
|
| 81 |
+
"step": 30
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 0.11,
|
| 85 |
+
"grad_norm": 1.7675389406697828,
|
| 86 |
+
"learning_rate": 4.998060489154965e-06,
|
| 87 |
+
"logits/chosen": -2.7818994522094727,
|
| 88 |
+
"logits/rejected": -2.7307980060577393,
|
| 89 |
+
"logps/chosen": -276.9420471191406,
|
| 90 |
+
"logps/rejected": -230.1887969970703,
|
| 91 |
+
"loss": 0.6326,
|
| 92 |
+
"rewards/accuracies": 0.925000011920929,
|
| 93 |
+
"rewards/chosen": 0.08895337581634521,
|
| 94 |
+
"rewards/margins": 0.11473529040813446,
|
| 95 |
+
"rewards/margins_max": 0.2537270188331604,
|
| 96 |
+
"rewards/margins_min": 0.01772489957511425,
|
| 97 |
+
"rewards/margins_std": 0.10599502176046371,
|
| 98 |
+
"rewards/rejected": -0.025781899690628052,
|
| 99 |
+
"step": 40
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.14,
|
| 103 |
+
"grad_norm": 2.021631325803066,
|
| 104 |
+
"learning_rate": 4.976275538042932e-06,
|
| 105 |
+
"logits/chosen": -2.7466044425964355,
|
| 106 |
+
"logits/rejected": -2.6687169075012207,
|
| 107 |
+
"logps/chosen": -304.30511474609375,
|
| 108 |
+
"logps/rejected": -243.2073211669922,
|
| 109 |
+
"loss": 0.589,
|
| 110 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 111 |
+
"rewards/chosen": 0.13480214774608612,
|
| 112 |
+
"rewards/margins": 0.22513285279273987,
|
| 113 |
+
"rewards/margins_max": 0.46415281295776367,
|
| 114 |
+
"rewards/margins_min": 0.0476544126868248,
|
| 115 |
+
"rewards/margins_std": 0.18787814676761627,
|
| 116 |
+
"rewards/rejected": -0.09033068269491196,
|
| 117 |
+
"step": 50
|
| 118 |
+
},
|
| 119 |
+
{
|
| 120 |
+
"epoch": 0.17,
|
| 121 |
+
"grad_norm": 1.9422427549351036,
|
| 122 |
+
"learning_rate": 4.93049306999712e-06,
|
| 123 |
+
"logits/chosen": -2.6410093307495117,
|
| 124 |
+
"logits/rejected": -2.60492205619812,
|
| 125 |
+
"logps/chosen": -327.55523681640625,
|
| 126 |
+
"logps/rejected": -283.3590393066406,
|
| 127 |
+
"loss": 0.5384,
|
| 128 |
+
"rewards/accuracies": 0.949999988079071,
|
| 129 |
+
"rewards/chosen": 0.17225147783756256,
|
| 130 |
+
"rewards/margins": 0.3661409616470337,
|
| 131 |
+
"rewards/margins_max": 0.6910489201545715,
|
| 132 |
+
"rewards/margins_min": 0.1382865607738495,
|
| 133 |
+
"rewards/margins_std": 0.2512677311897278,
|
| 134 |
+
"rewards/rejected": -0.19388948380947113,
|
| 135 |
+
"step": 60
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.2,
|
| 139 |
+
"grad_norm": 1.7976861864205922,
|
| 140 |
+
"learning_rate": 4.861156761634014e-06,
|
| 141 |
+
"logits/chosen": -2.649125576019287,
|
| 142 |
+
"logits/rejected": -2.5835797786712646,
|
| 143 |
+
"logps/chosen": -343.18719482421875,
|
| 144 |
+
"logps/rejected": -268.24822998046875,
|
| 145 |
+
"loss": 0.4917,
|
| 146 |
+
"rewards/accuracies": 0.949999988079071,
|
| 147 |
+
"rewards/chosen": 0.19151589274406433,
|
| 148 |
+
"rewards/margins": 0.5065498352050781,
|
| 149 |
+
"rewards/margins_max": 0.9112545251846313,
|
| 150 |
+
"rewards/margins_min": 0.11504770815372467,
|
| 151 |
+
"rewards/margins_std": 0.35929277539253235,
|
| 152 |
+
"rewards/rejected": -0.3150339722633362,
|
| 153 |
+
"step": 70
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"epoch": 0.23,
|
| 157 |
+
"grad_norm": 3.0636121326493675,
|
| 158 |
+
"learning_rate": 4.7689385491773934e-06,
|
| 159 |
+
"logits/chosen": -2.6317451000213623,
|
| 160 |
+
"logits/rejected": -2.5868935585021973,
|
| 161 |
+
"logps/chosen": -316.9160461425781,
|
| 162 |
+
"logps/rejected": -335.57354736328125,
|
| 163 |
+
"loss": 0.456,
|
| 164 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 165 |
+
"rewards/chosen": 0.1493242084980011,
|
| 166 |
+
"rewards/margins": 0.6301286220550537,
|
| 167 |
+
"rewards/margins_max": 1.0892175436019897,
|
| 168 |
+
"rewards/margins_min": 0.18430864810943604,
|
| 169 |
+
"rewards/margins_std": 0.39657825231552124,
|
| 170 |
+
"rewards/rejected": -0.480804443359375,
|
| 171 |
+
"step": 80
|
| 172 |
+
},
|
| 173 |
+
{
|
| 174 |
+
"epoch": 0.25,
|
| 175 |
+
"grad_norm": 4.331028594376021,
|
| 176 |
+
"learning_rate": 4.654732116743193e-06,
|
| 177 |
+
"logits/chosen": -2.5327608585357666,
|
| 178 |
+
"logits/rejected": -2.4845287799835205,
|
| 179 |
+
"logps/chosen": -329.42620849609375,
|
| 180 |
+
"logps/rejected": -258.33453369140625,
|
| 181 |
+
"loss": 0.4162,
|
| 182 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 183 |
+
"rewards/chosen": 0.09532937407493591,
|
| 184 |
+
"rewards/margins": 0.688909649848938,
|
| 185 |
+
"rewards/margins_max": 1.1923458576202393,
|
| 186 |
+
"rewards/margins_min": 0.23197071254253387,
|
| 187 |
+
"rewards/margins_std": 0.4409102499485016,
|
| 188 |
+
"rewards/rejected": -0.5935802459716797,
|
| 189 |
+
"step": 90
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"epoch": 0.28,
|
| 193 |
+
"grad_norm": 4.295716285529308,
|
| 194 |
+
"learning_rate": 4.5196442356717526e-06,
|
| 195 |
+
"logits/chosen": -2.578570604324341,
|
| 196 |
+
"logits/rejected": -2.512251377105713,
|
| 197 |
+
"logps/chosen": -341.71697998046875,
|
| 198 |
+
"logps/rejected": -342.2934875488281,
|
| 199 |
+
"loss": 0.3843,
|
| 200 |
+
"rewards/accuracies": 0.925000011920929,
|
| 201 |
+
"rewards/chosen": 0.04628221318125725,
|
| 202 |
+
"rewards/margins": 0.8309832811355591,
|
| 203 |
+
"rewards/margins_max": 1.4554868936538696,
|
| 204 |
+
"rewards/margins_min": 0.29576951265335083,
|
| 205 |
+
"rewards/margins_std": 0.5407849550247192,
|
| 206 |
+
"rewards/rejected": -0.7847011089324951,
|
| 207 |
+
"step": 100
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 0.28,
|
| 211 |
+
"eval_logits/chosen": -2.500088691711426,
|
| 212 |
+
"eval_logits/rejected": -2.47070574760437,
|
| 213 |
+
"eval_logps/chosen": -366.80853271484375,
|
| 214 |
+
"eval_logps/rejected": -353.1435546875,
|
| 215 |
+
"eval_loss": 0.6753506064414978,
|
| 216 |
+
"eval_rewards/accuracies": 0.6190000176429749,
|
| 217 |
+
"eval_rewards/chosen": -0.8221511244773865,
|
| 218 |
+
"eval_rewards/margins": 0.1234951838850975,
|
| 219 |
+
"eval_rewards/margins_max": 0.8275310397148132,
|
| 220 |
+
"eval_rewards/margins_min": -0.5849089622497559,
|
| 221 |
+
"eval_rewards/margins_std": 0.4660783112049103,
|
| 222 |
+
"eval_rewards/rejected": -0.9456462860107422,
|
| 223 |
+
"eval_runtime": 429.3559,
|
| 224 |
+
"eval_samples_per_second": 4.658,
|
| 225 |
+
"eval_steps_per_second": 0.291,
|
| 226 |
+
"step": 100
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"epoch": 0.31,
|
| 230 |
+
"grad_norm": 3.6576814376248663,
|
| 231 |
+
"learning_rate": 4.364984038837727e-06,
|
| 232 |
+
"logits/chosen": -2.486511707305908,
|
| 233 |
+
"logits/rejected": -2.4051737785339355,
|
| 234 |
+
"logps/chosen": -395.6208801269531,
|
| 235 |
+
"logps/rejected": -412.60369873046875,
|
| 236 |
+
"loss": 0.2943,
|
| 237 |
+
"rewards/accuracies": 0.987500011920929,
|
| 238 |
+
"rewards/chosen": -0.04258393868803978,
|
| 239 |
+
"rewards/margins": 1.210031509399414,
|
| 240 |
+
"rewards/margins_max": 1.81746506690979,
|
| 241 |
+
"rewards/margins_min": 0.560971736907959,
|
| 242 |
+
"rewards/margins_std": 0.5801268815994263,
|
| 243 |
+
"rewards/rejected": -1.2526154518127441,
|
| 244 |
+
"step": 110
|
| 245 |
+
},
|
| 246 |
+
{
|
| 247 |
+
"epoch": 0.34,
|
| 248 |
+
"grad_norm": 4.445946405194985,
|
| 249 |
+
"learning_rate": 4.192250333880045e-06,
|
| 250 |
+
"logits/chosen": -2.3081154823303223,
|
| 251 |
+
"logits/rejected": -2.2312095165252686,
|
| 252 |
+
"logps/chosen": -416.22509765625,
|
| 253 |
+
"logps/rejected": -439.51318359375,
|
| 254 |
+
"loss": 0.262,
|
| 255 |
+
"rewards/accuracies": 0.949999988079071,
|
| 256 |
+
"rewards/chosen": -0.21479575335979462,
|
| 257 |
+
"rewards/margins": 1.514960527420044,
|
| 258 |
+
"rewards/margins_max": 2.5221104621887207,
|
| 259 |
+
"rewards/margins_min": 0.6053955554962158,
|
| 260 |
+
"rewards/margins_std": 0.8898022770881653,
|
| 261 |
+
"rewards/rejected": -1.7297561168670654,
|
| 262 |
+
"step": 120
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"epoch": 0.37,
|
| 266 |
+
"grad_norm": 5.213009355206989,
|
| 267 |
+
"learning_rate": 4.0031170782990214e-06,
|
| 268 |
+
"logits/chosen": -2.059696912765503,
|
| 269 |
+
"logits/rejected": -1.9685862064361572,
|
| 270 |
+
"logps/chosen": -424.81292724609375,
|
| 271 |
+
"logps/rejected": -523.0667114257812,
|
| 272 |
+
"loss": 0.1912,
|
| 273 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 274 |
+
"rewards/chosen": -0.1958635300397873,
|
| 275 |
+
"rewards/margins": 2.025548219680786,
|
| 276 |
+
"rewards/margins_max": 3.0465281009674072,
|
| 277 |
+
"rewards/margins_min": 0.8110028505325317,
|
| 278 |
+
"rewards/margins_std": 0.9938048124313354,
|
| 279 |
+
"rewards/rejected": -2.221411943435669,
|
| 280 |
+
"step": 130
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.39,
|
| 284 |
+
"grad_norm": 13.540271738049956,
|
| 285 |
+
"learning_rate": 3.7994171571810756e-06,
|
| 286 |
+
"logits/chosen": -1.8918193578720093,
|
| 287 |
+
"logits/rejected": -1.8057823181152344,
|
| 288 |
+
"logps/chosen": -401.6873474121094,
|
| 289 |
+
"logps/rejected": -562.7242431640625,
|
| 290 |
+
"loss": 0.179,
|
| 291 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 292 |
+
"rewards/chosen": -0.28990477323532104,
|
| 293 |
+
"rewards/margins": 2.6347408294677734,
|
| 294 |
+
"rewards/margins_max": 3.8110084533691406,
|
| 295 |
+
"rewards/margins_min": 1.1343365907669067,
|
| 296 |
+
"rewards/margins_std": 1.2381587028503418,
|
| 297 |
+
"rewards/rejected": -2.9246459007263184,
|
| 298 |
+
"step": 140
|
| 299 |
+
},
|
| 300 |
+
{
|
| 301 |
+
"epoch": 0.42,
|
| 302 |
+
"grad_norm": 7.258937456966438,
|
| 303 |
+
"learning_rate": 3.5831246207606597e-06,
|
| 304 |
+
"logits/chosen": -1.8240772485733032,
|
| 305 |
+
"logits/rejected": -1.6942760944366455,
|
| 306 |
+
"logps/chosen": -383.18658447265625,
|
| 307 |
+
"logps/rejected": -533.005859375,
|
| 308 |
+
"loss": 0.1794,
|
| 309 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 310 |
+
"rewards/chosen": -0.6033543348312378,
|
| 311 |
+
"rewards/margins": 2.5889029502868652,
|
| 312 |
+
"rewards/margins_max": 3.9328532218933105,
|
| 313 |
+
"rewards/margins_min": 0.870226263999939,
|
| 314 |
+
"rewards/margins_std": 1.4039599895477295,
|
| 315 |
+
"rewards/rejected": -3.1922574043273926,
|
| 316 |
+
"step": 150
|
| 317 |
+
},
|
| 318 |
+
{
|
| 319 |
+
"epoch": 0.45,
|
| 320 |
+
"grad_norm": 3.520516887340939,
|
| 321 |
+
"learning_rate": 3.3563355539546795e-06,
|
| 322 |
+
"logits/chosen": -1.6218408346176147,
|
| 323 |
+
"logits/rejected": -1.5248044729232788,
|
| 324 |
+
"logps/chosen": -405.44110107421875,
|
| 325 |
+
"logps/rejected": -626.1888427734375,
|
| 326 |
+
"loss": 0.1195,
|
| 327 |
+
"rewards/accuracies": 0.987500011920929,
|
| 328 |
+
"rewards/chosen": -0.5993050336837769,
|
| 329 |
+
"rewards/margins": 3.3186564445495605,
|
| 330 |
+
"rewards/margins_max": 4.61026668548584,
|
| 331 |
+
"rewards/margins_min": 1.6127792596817017,
|
| 332 |
+
"rewards/margins_std": 1.3561828136444092,
|
| 333 |
+
"rewards/rejected": -3.917961835861206,
|
| 334 |
+
"step": 160
|
| 335 |
+
},
|
| 336 |
+
{
|
| 337 |
+
"epoch": 0.48,
|
| 338 |
+
"grad_norm": 4.463403878446236,
|
| 339 |
+
"learning_rate": 3.121247763262235e-06,
|
| 340 |
+
"logits/chosen": -1.5979077816009521,
|
| 341 |
+
"logits/rejected": -1.4862514734268188,
|
| 342 |
+
"logps/chosen": -569.4307861328125,
|
| 343 |
+
"logps/rejected": -859.6317138671875,
|
| 344 |
+
"loss": 0.1156,
|
| 345 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 346 |
+
"rewards/chosen": -1.9402544498443604,
|
| 347 |
+
"rewards/margins": 3.6933372020721436,
|
| 348 |
+
"rewards/margins_max": 5.283831596374512,
|
| 349 |
+
"rewards/margins_min": 1.5569112300872803,
|
| 350 |
+
"rewards/margins_std": 1.7149642705917358,
|
| 351 |
+
"rewards/rejected": -5.6335906982421875,
|
| 352 |
+
"step": 170
|
| 353 |
+
},
|
| 354 |
+
{
|
| 355 |
+
"epoch": 0.51,
|
| 356 |
+
"grad_norm": 28.94814953201888,
|
| 357 |
+
"learning_rate": 2.8801394778833475e-06,
|
| 358 |
+
"logits/chosen": -1.4478727579116821,
|
| 359 |
+
"logits/rejected": -1.3353261947631836,
|
| 360 |
+
"logps/chosen": -677.9209594726562,
|
| 361 |
+
"logps/rejected": -972.3616943359375,
|
| 362 |
+
"loss": 0.1518,
|
| 363 |
+
"rewards/accuracies": 0.9375,
|
| 364 |
+
"rewards/chosen": -3.2994747161865234,
|
| 365 |
+
"rewards/margins": 3.4859726428985596,
|
| 366 |
+
"rewards/margins_max": 5.324586868286133,
|
| 367 |
+
"rewards/margins_min": 1.277085781097412,
|
| 368 |
+
"rewards/margins_std": 1.7865108251571655,
|
| 369 |
+
"rewards/rejected": -6.785447597503662,
|
| 370 |
+
"step": 180
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"epoch": 0.54,
|
| 374 |
+
"grad_norm": 1.6663635534583416,
|
| 375 |
+
"learning_rate": 2.6353472714635443e-06,
|
| 376 |
+
"logits/chosen": -1.4343066215515137,
|
| 377 |
+
"logits/rejected": -1.2333465814590454,
|
| 378 |
+
"logps/chosen": -710.7890014648438,
|
| 379 |
+
"logps/rejected": -980.9490966796875,
|
| 380 |
+
"loss": 0.1188,
|
| 381 |
+
"rewards/accuracies": 0.987500011920929,
|
| 382 |
+
"rewards/chosen": -3.5141806602478027,
|
| 383 |
+
"rewards/margins": 3.9161250591278076,
|
| 384 |
+
"rewards/margins_max": 5.42164421081543,
|
| 385 |
+
"rewards/margins_min": 1.7522633075714111,
|
| 386 |
+
"rewards/margins_std": 1.6805788278579712,
|
| 387 |
+
"rewards/rejected": -7.430306434631348,
|
| 388 |
+
"step": 190
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.56,
|
| 392 |
+
"grad_norm": 3.6276038453521187,
|
| 393 |
+
"learning_rate": 2.3892434184240536e-06,
|
| 394 |
+
"logits/chosen": -1.5112391710281372,
|
| 395 |
+
"logits/rejected": -1.3403419256210327,
|
| 396 |
+
"logps/chosen": -812.4998779296875,
|
| 397 |
+
"logps/rejected": -1101.1927490234375,
|
| 398 |
+
"loss": 0.1345,
|
| 399 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 400 |
+
"rewards/chosen": -4.135049343109131,
|
| 401 |
+
"rewards/margins": 4.216553688049316,
|
| 402 |
+
"rewards/margins_max": 6.037824630737305,
|
| 403 |
+
"rewards/margins_min": 1.9130373001098633,
|
| 404 |
+
"rewards/margins_std": 1.9190574884414673,
|
| 405 |
+
"rewards/rejected": -8.351602554321289,
|
| 406 |
+
"step": 200
|
| 407 |
+
},
|
| 408 |
+
{
|
| 409 |
+
"epoch": 0.56,
|
| 410 |
+
"eval_logits/chosen": -1.4037277698516846,
|
| 411 |
+
"eval_logits/rejected": -1.349421739578247,
|
| 412 |
+
"eval_logps/chosen": -952.8915405273438,
|
| 413 |
+
"eval_logps/rejected": -992.8120727539062,
|
| 414 |
+
"eval_loss": 0.9115286469459534,
|
| 415 |
+
"eval_rewards/accuracies": 0.6230000257492065,
|
| 416 |
+
"eval_rewards/chosen": -6.682980537414551,
|
| 417 |
+
"eval_rewards/margins": 0.659350574016571,
|
| 418 |
+
"eval_rewards/margins_max": 4.129995822906494,
|
| 419 |
+
"eval_rewards/margins_min": -2.920621633529663,
|
| 420 |
+
"eval_rewards/margins_std": 2.361518383026123,
|
| 421 |
+
"eval_rewards/rejected": -7.342331886291504,
|
| 422 |
+
"eval_runtime": 428.8033,
|
| 423 |
+
"eval_samples_per_second": 4.664,
|
| 424 |
+
"eval_steps_per_second": 0.292,
|
| 425 |
+
"step": 200
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"epoch": 0.59,
|
| 429 |
+
"grad_norm": 10.435489666521553,
|
| 430 |
+
"learning_rate": 2.1442129043167877e-06,
|
| 431 |
+
"logits/chosen": -1.39103102684021,
|
| 432 |
+
"logits/rejected": -1.270121455192566,
|
| 433 |
+
"logps/chosen": -762.1986694335938,
|
| 434 |
+
"logps/rejected": -1084.1378173828125,
|
| 435 |
+
"loss": 0.1299,
|
| 436 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 437 |
+
"rewards/chosen": -4.0687384605407715,
|
| 438 |
+
"rewards/margins": 4.200930118560791,
|
| 439 |
+
"rewards/margins_max": 6.173401355743408,
|
| 440 |
+
"rewards/margins_min": 1.691877007484436,
|
| 441 |
+
"rewards/margins_std": 2.071268320083618,
|
| 442 |
+
"rewards/rejected": -8.269669532775879,
|
| 443 |
+
"step": 210
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 0.62,
|
| 447 |
+
"grad_norm": 5.123588367601374,
|
| 448 |
+
"learning_rate": 1.9026303129961049e-06,
|
| 449 |
+
"logits/chosen": -1.5539300441741943,
|
| 450 |
+
"logits/rejected": -1.3544563055038452,
|
| 451 |
+
"logps/chosen": -777.8438720703125,
|
| 452 |
+
"logps/rejected": -1124.8797607421875,
|
| 453 |
+
"loss": 0.0627,
|
| 454 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 455 |
+
"rewards/chosen": -3.7083256244659424,
|
| 456 |
+
"rewards/margins": 4.7932538986206055,
|
| 457 |
+
"rewards/margins_max": 6.416988372802734,
|
| 458 |
+
"rewards/margins_min": 2.6955573558807373,
|
| 459 |
+
"rewards/margins_std": 1.6872974634170532,
|
| 460 |
+
"rewards/rejected": -8.501578330993652,
|
| 461 |
+
"step": 220
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.65,
|
| 465 |
+
"grad_norm": 8.383127396811519,
|
| 466 |
+
"learning_rate": 1.66683681459314e-06,
|
| 467 |
+
"logits/chosen": -1.5905059576034546,
|
| 468 |
+
"logits/rejected": -1.3816958665847778,
|
| 469 |
+
"logps/chosen": -772.0947265625,
|
| 470 |
+
"logps/rejected": -1100.8663330078125,
|
| 471 |
+
"loss": 0.0918,
|
| 472 |
+
"rewards/accuracies": 0.987500011920929,
|
| 473 |
+
"rewards/chosen": -3.6928131580352783,
|
| 474 |
+
"rewards/margins": 4.681826591491699,
|
| 475 |
+
"rewards/margins_max": 6.284341335296631,
|
| 476 |
+
"rewards/margins_min": 2.5316035747528076,
|
| 477 |
+
"rewards/margins_std": 1.6845054626464844,
|
| 478 |
+
"rewards/rejected": -8.374638557434082,
|
| 479 |
+
"step": 230
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.68,
|
| 483 |
+
"grad_norm": 4.5829488188390926,
|
| 484 |
+
"learning_rate": 1.4391174773015836e-06,
|
| 485 |
+
"logits/chosen": -1.5085848569869995,
|
| 486 |
+
"logits/rejected": -1.3704946041107178,
|
| 487 |
+
"logps/chosen": -735.5416870117188,
|
| 488 |
+
"logps/rejected": -1119.05029296875,
|
| 489 |
+
"loss": 0.131,
|
| 490 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 491 |
+
"rewards/chosen": -3.923596143722534,
|
| 492 |
+
"rewards/margins": 4.414877414703369,
|
| 493 |
+
"rewards/margins_max": 6.264102935791016,
|
| 494 |
+
"rewards/margins_min": 2.1018009185791016,
|
| 495 |
+
"rewards/margins_std": 1.8714433908462524,
|
| 496 |
+
"rewards/rejected": -8.338472366333008,
|
| 497 |
+
"step": 240
|
| 498 |
+
},
|
| 499 |
+
{
|
| 500 |
+
"epoch": 0.7,
|
| 501 |
+
"grad_norm": 7.92768529410158,
|
| 502 |
+
"learning_rate": 1.2216791228457778e-06,
|
| 503 |
+
"logits/chosen": -1.5116816759109497,
|
| 504 |
+
"logits/rejected": -1.3140555620193481,
|
| 505 |
+
"logps/chosen": -774.6199951171875,
|
| 506 |
+
"logps/rejected": -1130.3253173828125,
|
| 507 |
+
"loss": 0.0892,
|
| 508 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 509 |
+
"rewards/chosen": -4.107260704040527,
|
| 510 |
+
"rewards/margins": 4.6648783683776855,
|
| 511 |
+
"rewards/margins_max": 6.308585166931152,
|
| 512 |
+
"rewards/margins_min": 2.563100814819336,
|
| 513 |
+
"rewards/margins_std": 1.706753134727478,
|
| 514 |
+
"rewards/rejected": -8.772139549255371,
|
| 515 |
+
"step": 250
|
| 516 |
+
},
|
| 517 |
+
{
|
| 518 |
+
"epoch": 0.73,
|
| 519 |
+
"grad_norm": 5.013402333461111,
|
| 520 |
+
"learning_rate": 1.0166289402331391e-06,
|
| 521 |
+
"logits/chosen": -1.6035802364349365,
|
| 522 |
+
"logits/rejected": -1.4094022512435913,
|
| 523 |
+
"logps/chosen": -718.6414794921875,
|
| 524 |
+
"logps/rejected": -1079.9404296875,
|
| 525 |
+
"loss": 0.077,
|
| 526 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 527 |
+
"rewards/chosen": -3.621960163116455,
|
| 528 |
+
"rewards/margins": 4.5798468589782715,
|
| 529 |
+
"rewards/margins_max": 6.199366569519043,
|
| 530 |
+
"rewards/margins_min": 2.420649290084839,
|
| 531 |
+
"rewards/margins_std": 1.7149250507354736,
|
| 532 |
+
"rewards/rejected": -8.201807022094727,
|
| 533 |
+
"step": 260
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 0.76,
|
| 537 |
+
"grad_norm": 16.373360821165523,
|
| 538 |
+
"learning_rate": 8.259540650444736e-07,
|
| 539 |
+
"logits/chosen": -1.555152177810669,
|
| 540 |
+
"logits/rejected": -1.3883273601531982,
|
| 541 |
+
"logps/chosen": -743.9249877929688,
|
| 542 |
+
"logps/rejected": -1082.8458251953125,
|
| 543 |
+
"loss": 0.0929,
|
| 544 |
+
"rewards/accuracies": 0.949999988079071,
|
| 545 |
+
"rewards/chosen": -3.660247802734375,
|
| 546 |
+
"rewards/margins": 4.5421624183654785,
|
| 547 |
+
"rewards/margins_max": 6.060577392578125,
|
| 548 |
+
"rewards/margins_min": 2.4010677337646484,
|
| 549 |
+
"rewards/margins_std": 1.6667591333389282,
|
| 550 |
+
"rewards/rejected": -8.202409744262695,
|
| 551 |
+
"step": 270
|
| 552 |
+
},
|
| 553 |
+
{
|
| 554 |
+
"epoch": 0.79,
|
| 555 |
+
"grad_norm": 8.099720759346864,
|
| 556 |
+
"learning_rate": 6.515023221586722e-07,
|
| 557 |
+
"logits/chosen": -1.4892373085021973,
|
| 558 |
+
"logits/rejected": -1.3762261867523193,
|
| 559 |
+
"logps/chosen": -695.7076416015625,
|
| 560 |
+
"logps/rejected": -1113.3646240234375,
|
| 561 |
+
"loss": 0.0897,
|
| 562 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 563 |
+
"rewards/chosen": -3.6505286693573,
|
| 564 |
+
"rewards/margins": 4.7476959228515625,
|
| 565 |
+
"rewards/margins_max": 6.505625247955322,
|
| 566 |
+
"rewards/margins_min": 2.4996237754821777,
|
| 567 |
+
"rewards/margins_std": 1.8462820053100586,
|
| 568 |
+
"rewards/rejected": -8.398224830627441,
|
| 569 |
+
"step": 280
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"epoch": 0.82,
|
| 573 |
+
"grad_norm": 18.72504361278716,
|
| 574 |
+
"learning_rate": 4.949643185335288e-07,
|
| 575 |
+
"logits/chosen": -1.5151816606521606,
|
| 576 |
+
"logits/rejected": -1.3798022270202637,
|
| 577 |
+
"logps/chosen": -686.1121826171875,
|
| 578 |
+
"logps/rejected": -1038.52587890625,
|
| 579 |
+
"loss": 0.0905,
|
| 580 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 581 |
+
"rewards/chosen": -3.451200485229492,
|
| 582 |
+
"rewards/margins": 4.272095680236816,
|
| 583 |
+
"rewards/margins_max": 6.312056064605713,
|
| 584 |
+
"rewards/margins_min": 1.573736310005188,
|
| 585 |
+
"rewards/margins_std": 2.1745951175689697,
|
| 586 |
+
"rewards/rejected": -7.72329568862915,
|
| 587 |
+
"step": 290
|
| 588 |
+
},
|
| 589 |
+
{
|
| 590 |
+
"epoch": 0.85,
|
| 591 |
+
"grad_norm": 9.116977457145628,
|
| 592 |
+
"learning_rate": 3.578570595810274e-07,
|
| 593 |
+
"logits/chosen": -1.5946242809295654,
|
| 594 |
+
"logits/rejected": -1.4539456367492676,
|
| 595 |
+
"logps/chosen": -693.6813354492188,
|
| 596 |
+
"logps/rejected": -1122.047119140625,
|
| 597 |
+
"loss": 0.0467,
|
| 598 |
+
"rewards/accuracies": 1.0,
|
| 599 |
+
"rewards/chosen": -3.300471782684326,
|
| 600 |
+
"rewards/margins": 5.025434970855713,
|
| 601 |
+
"rewards/margins_max": 6.420098781585693,
|
| 602 |
+
"rewards/margins_min": 2.8854596614837646,
|
| 603 |
+
"rewards/margins_std": 1.5815579891204834,
|
| 604 |
+
"rewards/rejected": -8.325906753540039,
|
| 605 |
+
"step": 300
|
| 606 |
+
},
|
| 607 |
+
{
|
| 608 |
+
"epoch": 0.85,
|
| 609 |
+
"eval_logits/chosen": -1.4909454584121704,
|
| 610 |
+
"eval_logits/rejected": -1.4320358037948608,
|
| 611 |
+
"eval_logps/chosen": -888.703857421875,
|
| 612 |
+
"eval_logps/rejected": -956.687744140625,
|
| 613 |
+
"eval_loss": 0.9370979070663452,
|
| 614 |
+
"eval_rewards/accuracies": 0.6449999809265137,
|
| 615 |
+
"eval_rewards/chosen": -6.041103839874268,
|
| 616 |
+
"eval_rewards/margins": 0.9399833083152771,
|
| 617 |
+
"eval_rewards/margins_max": 4.743701457977295,
|
| 618 |
+
"eval_rewards/margins_min": -3.0368335247039795,
|
| 619 |
+
"eval_rewards/margins_std": 2.654953718185425,
|
| 620 |
+
"eval_rewards/rejected": -6.981087684631348,
|
| 621 |
+
"eval_runtime": 428.6815,
|
| 622 |
+
"eval_samples_per_second": 4.665,
|
| 623 |
+
"eval_steps_per_second": 0.292,
|
| 624 |
+
"step": 300
|
| 625 |
+
},
|
| 626 |
+
{
|
| 627 |
+
"epoch": 0.87,
|
| 628 |
+
"grad_norm": 17.025809050618562,
|
| 629 |
+
"learning_rate": 2.4150924791035037e-07,
|
| 630 |
+
"logits/chosen": -1.5680278539657593,
|
| 631 |
+
"logits/rejected": -1.3463890552520752,
|
| 632 |
+
"logps/chosen": -778.5208740234375,
|
| 633 |
+
"logps/rejected": -1089.8924560546875,
|
| 634 |
+
"loss": 0.1036,
|
| 635 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 636 |
+
"rewards/chosen": -4.222385406494141,
|
| 637 |
+
"rewards/margins": 4.524166584014893,
|
| 638 |
+
"rewards/margins_max": 6.391068458557129,
|
| 639 |
+
"rewards/margins_min": 2.287738084793091,
|
| 640 |
+
"rewards/margins_std": 1.8728872537612915,
|
| 641 |
+
"rewards/rejected": -8.746552467346191,
|
| 642 |
+
"step": 310
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.9,
|
| 646 |
+
"grad_norm": 6.614213043020825,
|
| 647 |
+
"learning_rate": 1.4704840690808658e-07,
|
| 648 |
+
"logits/chosen": -1.5289008617401123,
|
| 649 |
+
"logits/rejected": -1.3871232271194458,
|
| 650 |
+
"logps/chosen": -733.2238159179688,
|
| 651 |
+
"logps/rejected": -1093.3983154296875,
|
| 652 |
+
"loss": 0.0976,
|
| 653 |
+
"rewards/accuracies": 0.949999988079071,
|
| 654 |
+
"rewards/chosen": -3.8258163928985596,
|
| 655 |
+
"rewards/margins": 4.487856388092041,
|
| 656 |
+
"rewards/margins_max": 6.392783164978027,
|
| 657 |
+
"rewards/margins_min": 1.399547815322876,
|
| 658 |
+
"rewards/margins_std": 2.305619955062866,
|
| 659 |
+
"rewards/rejected": -8.313672065734863,
|
| 660 |
+
"step": 320
|
| 661 |
+
},
|
| 662 |
+
{
|
| 663 |
+
"epoch": 0.93,
|
| 664 |
+
"grad_norm": 9.60113340680737,
|
| 665 |
+
"learning_rate": 7.538995394063996e-08,
|
| 666 |
+
"logits/chosen": -1.629804253578186,
|
| 667 |
+
"logits/rejected": -1.444265604019165,
|
| 668 |
+
"logps/chosen": -782.9251708984375,
|
| 669 |
+
"logps/rejected": -1121.7064208984375,
|
| 670 |
+
"loss": 0.1186,
|
| 671 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 672 |
+
"rewards/chosen": -3.826475143432617,
|
| 673 |
+
"rewards/margins": 4.706925392150879,
|
| 674 |
+
"rewards/margins_max": 6.423586368560791,
|
| 675 |
+
"rewards/margins_min": 2.4785187244415283,
|
| 676 |
+
"rewards/margins_std": 1.876365065574646,
|
| 677 |
+
"rewards/rejected": -8.53339958190918,
|
| 678 |
+
"step": 330
|
| 679 |
+
},
|
| 680 |
+
{
|
| 681 |
+
"epoch": 0.96,
|
| 682 |
+
"grad_norm": 18.972390442985663,
|
| 683 |
+
"learning_rate": 2.722832907015971e-08,
|
| 684 |
+
"logits/chosen": -1.5121577978134155,
|
| 685 |
+
"logits/rejected": -1.355576753616333,
|
| 686 |
+
"logps/chosen": -770.6697387695312,
|
| 687 |
+
"logps/rejected": -1123.355712890625,
|
| 688 |
+
"loss": 0.0955,
|
| 689 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 690 |
+
"rewards/chosen": -4.256497383117676,
|
| 691 |
+
"rewards/margins": 4.456811428070068,
|
| 692 |
+
"rewards/margins_max": 6.2082414627075195,
|
| 693 |
+
"rewards/margins_min": 2.056246280670166,
|
| 694 |
+
"rewards/margins_std": 1.880099892616272,
|
| 695 |
+
"rewards/rejected": -8.713308334350586,
|
| 696 |
+
"step": 340
|
| 697 |
+
},
|
| 698 |
+
{
|
| 699 |
+
"epoch": 0.99,
|
| 700 |
+
"grad_norm": 26.738917827064355,
|
| 701 |
+
"learning_rate": 3.030265255329623e-09,
|
| 702 |
+
"logits/chosen": -1.5162473917007446,
|
| 703 |
+
"logits/rejected": -1.3702640533447266,
|
| 704 |
+
"logps/chosen": -736.2762451171875,
|
| 705 |
+
"logps/rejected": -1147.2891845703125,
|
| 706 |
+
"loss": 0.1122,
|
| 707 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 708 |
+
"rewards/chosen": -3.7701263427734375,
|
| 709 |
+
"rewards/margins": 4.834524631500244,
|
| 710 |
+
"rewards/margins_max": 6.308111667633057,
|
| 711 |
+
"rewards/margins_min": 2.398153781890869,
|
| 712 |
+
"rewards/margins_std": 1.7776079177856445,
|
| 713 |
+
"rewards/rejected": -8.604650497436523,
|
| 714 |
+
"step": 350
|
| 715 |
+
},
|
| 716 |
+
{
|
| 717 |
+
"epoch": 1.0,
|
| 718 |
+
"step": 355,
|
| 719 |
+
"total_flos": 0.0,
|
| 720 |
+
"train_loss": 0.24729210323011372,
|
| 721 |
+
"train_runtime": 4030.3896,
|
| 722 |
+
"train_samples_per_second": 1.409,
|
| 723 |
+
"train_steps_per_second": 0.088
|
| 724 |
+
}
|
| 725 |
+
],
|
| 726 |
+
"logging_steps": 10,
|
| 727 |
+
"max_steps": 355,
|
| 728 |
+
"num_input_tokens_seen": 0,
|
| 729 |
+
"num_train_epochs": 1,
|
| 730 |
+
"save_steps": 100,
|
| 731 |
+
"total_flos": 0.0,
|
| 732 |
+
"train_batch_size": 4,
|
| 733 |
+
"trial_name": null,
|
| 734 |
+
"trial_params": null
|
| 735 |
+
}
|