Model save
Browse files- README.md +77 -0
- all_results.json +9 -0
- generation_config.json +12 -0
- train_results.json +9 -0
- trainer_state.json +826 -0
README.md
ADDED
|
@@ -0,0 +1,77 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
license: llama3.1
|
| 4 |
+
base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- dpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: llama-3-8b-dpo-full
|
| 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 |
+
# llama-3-8b-dpo-full
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [meta-llama/Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct) on an unknown dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.5582
|
| 22 |
+
- Rewards/chosen: -1.3603
|
| 23 |
+
- Rewards/rejected: -2.0529
|
| 24 |
+
- Rewards/accuracies: 0.7262
|
| 25 |
+
- Rewards/margins: 0.6926
|
| 26 |
+
- Logps/rejected: -600.9839
|
| 27 |
+
- Logps/chosen: -540.5128
|
| 28 |
+
- Logits/rejected: -2.7438
|
| 29 |
+
- Logits/chosen: -2.5853
|
| 30 |
+
|
| 31 |
+
## Model description
|
| 32 |
+
|
| 33 |
+
More information needed
|
| 34 |
+
|
| 35 |
+
## Intended uses & limitations
|
| 36 |
+
|
| 37 |
+
More information needed
|
| 38 |
+
|
| 39 |
+
## Training and evaluation data
|
| 40 |
+
|
| 41 |
+
More information needed
|
| 42 |
+
|
| 43 |
+
## Training procedure
|
| 44 |
+
|
| 45 |
+
### Training hyperparameters
|
| 46 |
+
|
| 47 |
+
The following hyperparameters were used during training:
|
| 48 |
+
- learning_rate: 3e-07
|
| 49 |
+
- train_batch_size: 2
|
| 50 |
+
- eval_batch_size: 4
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 8
|
| 54 |
+
- gradient_accumulation_steps: 8
|
| 55 |
+
- total_train_batch_size: 128
|
| 56 |
+
- total_eval_batch_size: 32
|
| 57 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
+
- lr_scheduler_type: cosine
|
| 59 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
+
- num_epochs: 1
|
| 61 |
+
|
| 62 |
+
### Training results
|
| 63 |
+
|
| 64 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 65 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 66 |
+
| 0.6341 | 0.2094 | 100 | 0.6223 | -0.6021 | -0.8779 | 0.7103 | 0.2758 | -483.4830 | -464.6949 | -2.6189 | -2.4353 |
|
| 67 |
+
| 0.5887 | 0.4187 | 200 | 0.5796 | -1.0505 | -1.5993 | 0.7143 | 0.5488 | -555.6263 | -509.5346 | -2.6508 | -2.4854 |
|
| 68 |
+
| 0.5667 | 0.6281 | 300 | 0.5653 | -1.0427 | -1.6191 | 0.7222 | 0.5764 | -557.6055 | -508.7539 | -2.6684 | -2.5120 |
|
| 69 |
+
| 0.5803 | 0.8375 | 400 | 0.5582 | -1.3603 | -2.0529 | 0.7262 | 0.6926 | -600.9839 | -540.5128 | -2.7438 | -2.5853 |
|
| 70 |
+
|
| 71 |
+
|
| 72 |
+
### Framework versions
|
| 73 |
+
|
| 74 |
+
- Transformers 4.45.1
|
| 75 |
+
- Pytorch 2.4.0+cu121
|
| 76 |
+
- Datasets 3.0.0
|
| 77 |
+
- Tokenizers 0.20.0
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.998691442030882,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5939158873488068,
|
| 5 |
+
"train_runtime": 3918.7641,
|
| 6 |
+
"train_samples": 61135,
|
| 7 |
+
"train_samples_per_second": 15.601,
|
| 8 |
+
"train_steps_per_second": 0.122
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 128000,
|
| 3 |
+
"do_sample": true,
|
| 4 |
+
"eos_token_id": [
|
| 5 |
+
128001,
|
| 6 |
+
128008,
|
| 7 |
+
128009
|
| 8 |
+
],
|
| 9 |
+
"temperature": 0.6,
|
| 10 |
+
"top_p": 0.9,
|
| 11 |
+
"transformers_version": "4.45.1"
|
| 12 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.998691442030882,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5939158873488068,
|
| 5 |
+
"train_runtime": 3918.7641,
|
| 6 |
+
"train_samples": 61135,
|
| 7 |
+
"train_samples_per_second": 15.601,
|
| 8 |
+
"train_steps_per_second": 0.122
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,826 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.998691442030882,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 477,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.002093692750588851,
|
| 13 |
+
"grad_norm": 5.844382227449463,
|
| 14 |
+
"learning_rate": 6.25e-09,
|
| 15 |
+
"logits/chosen": -1.6512014865875244,
|
| 16 |
+
"logits/rejected": -1.4355601072311401,
|
| 17 |
+
"logps/chosen": -375.0550842285156,
|
| 18 |
+
"logps/rejected": -355.35711669921875,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.0,
|
| 21 |
+
"rewards/chosen": 0.0,
|
| 22 |
+
"rewards/margins": 0.0,
|
| 23 |
+
"rewards/rejected": 0.0,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.02093692750588851,
|
| 28 |
+
"grad_norm": 5.235821845579244,
|
| 29 |
+
"learning_rate": 6.25e-08,
|
| 30 |
+
"logits/chosen": -1.7515054941177368,
|
| 31 |
+
"logits/rejected": -2.0137522220611572,
|
| 32 |
+
"logps/chosen": -417.4104919433594,
|
| 33 |
+
"logps/rejected": -399.6319274902344,
|
| 34 |
+
"loss": 0.693,
|
| 35 |
+
"rewards/accuracies": 0.4305555522441864,
|
| 36 |
+
"rewards/chosen": -0.0020523108541965485,
|
| 37 |
+
"rewards/margins": -0.00028108435799367726,
|
| 38 |
+
"rewards/rejected": -0.0017712265253067017,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.04187385501177702,
|
| 43 |
+
"grad_norm": 4.969390451186454,
|
| 44 |
+
"learning_rate": 1.25e-07,
|
| 45 |
+
"logits/chosen": -1.6114051342010498,
|
| 46 |
+
"logits/rejected": -2.0293774604797363,
|
| 47 |
+
"logps/chosen": -404.6941833496094,
|
| 48 |
+
"logps/rejected": -367.32257080078125,
|
| 49 |
+
"loss": 0.6929,
|
| 50 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 51 |
+
"rewards/chosen": -0.0030800788663327694,
|
| 52 |
+
"rewards/margins": -2.9044622351648286e-05,
|
| 53 |
+
"rewards/rejected": -0.00305103394202888,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.06281078251766553,
|
| 58 |
+
"grad_norm": 5.046869414952022,
|
| 59 |
+
"learning_rate": 1.875e-07,
|
| 60 |
+
"logits/chosen": -1.7840598821640015,
|
| 61 |
+
"logits/rejected": -2.01045298576355,
|
| 62 |
+
"logps/chosen": -380.27178955078125,
|
| 63 |
+
"logps/rejected": -359.34735107421875,
|
| 64 |
+
"loss": 0.6913,
|
| 65 |
+
"rewards/accuracies": 0.550000011920929,
|
| 66 |
+
"rewards/chosen": -0.013285043649375439,
|
| 67 |
+
"rewards/margins": 0.0011252568801864982,
|
| 68 |
+
"rewards/rejected": -0.01441030204296112,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.08374771002355404,
|
| 73 |
+
"grad_norm": 4.698944247342457,
|
| 74 |
+
"learning_rate": 2.5e-07,
|
| 75 |
+
"logits/chosen": -1.6258041858673096,
|
| 76 |
+
"logits/rejected": -1.7915970087051392,
|
| 77 |
+
"logps/chosen": -407.2330017089844,
|
| 78 |
+
"logps/rejected": -409.3331298828125,
|
| 79 |
+
"loss": 0.6861,
|
| 80 |
+
"rewards/accuracies": 0.668749988079071,
|
| 81 |
+
"rewards/chosen": -0.038303375244140625,
|
| 82 |
+
"rewards/margins": 0.012813824228942394,
|
| 83 |
+
"rewards/rejected": -0.05111720412969589,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.10468463752944256,
|
| 88 |
+
"grad_norm": 5.305945732445058,
|
| 89 |
+
"learning_rate": 2.9998391212614154e-07,
|
| 90 |
+
"logits/chosen": -1.5322725772857666,
|
| 91 |
+
"logits/rejected": -1.7322235107421875,
|
| 92 |
+
"logps/chosen": -404.21575927734375,
|
| 93 |
+
"logps/rejected": -431.61004638671875,
|
| 94 |
+
"loss": 0.6751,
|
| 95 |
+
"rewards/accuracies": 0.625,
|
| 96 |
+
"rewards/chosen": -0.09563098847866058,
|
| 97 |
+
"rewards/margins": 0.032842181622982025,
|
| 98 |
+
"rewards/rejected": -0.1284731775522232,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.12562156503533106,
|
| 103 |
+
"grad_norm": 5.972525123376799,
|
| 104 |
+
"learning_rate": 2.9942119880575817e-07,
|
| 105 |
+
"logits/chosen": -1.8156770467758179,
|
| 106 |
+
"logits/rejected": -1.956964135169983,
|
| 107 |
+
"logps/chosen": -413.17218017578125,
|
| 108 |
+
"logps/rejected": -401.9190979003906,
|
| 109 |
+
"loss": 0.662,
|
| 110 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 111 |
+
"rewards/chosen": -0.22033607959747314,
|
| 112 |
+
"rewards/margins": 0.05222884565591812,
|
| 113 |
+
"rewards/rejected": -0.27256491780281067,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.14655849254121958,
|
| 118 |
+
"grad_norm": 8.237654744616636,
|
| 119 |
+
"learning_rate": 2.980575393956869e-07,
|
| 120 |
+
"logits/chosen": -1.9767135381698608,
|
| 121 |
+
"logits/rejected": -2.0496904850006104,
|
| 122 |
+
"logps/chosen": -442.13055419921875,
|
| 123 |
+
"logps/rejected": -466.5682678222656,
|
| 124 |
+
"loss": 0.6501,
|
| 125 |
+
"rewards/accuracies": 0.606249988079071,
|
| 126 |
+
"rewards/chosen": -0.4638500213623047,
|
| 127 |
+
"rewards/margins": 0.14070633053779602,
|
| 128 |
+
"rewards/rejected": -0.6045563817024231,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.16749542004710807,
|
| 133 |
+
"grad_norm": 7.104867946653516,
|
| 134 |
+
"learning_rate": 2.9590024355266256e-07,
|
| 135 |
+
"logits/chosen": -1.9303478002548218,
|
| 136 |
+
"logits/rejected": -2.1261794567108154,
|
| 137 |
+
"logps/chosen": -456.86334228515625,
|
| 138 |
+
"logps/rejected": -468.8231506347656,
|
| 139 |
+
"loss": 0.6315,
|
| 140 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 141 |
+
"rewards/chosen": -0.610644519329071,
|
| 142 |
+
"rewards/margins": 0.14992806315422058,
|
| 143 |
+
"rewards/rejected": -0.7605725526809692,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.1884323475529966,
|
| 148 |
+
"grad_norm": 7.380030502862527,
|
| 149 |
+
"learning_rate": 2.929608750821129e-07,
|
| 150 |
+
"logits/chosen": -1.861717939376831,
|
| 151 |
+
"logits/rejected": -2.3124728202819824,
|
| 152 |
+
"logps/chosen": -473.9281311035156,
|
| 153 |
+
"logps/rejected": -462.75018310546875,
|
| 154 |
+
"loss": 0.6356,
|
| 155 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 156 |
+
"rewards/chosen": -0.72138911485672,
|
| 157 |
+
"rewards/margins": 0.17633597552776337,
|
| 158 |
+
"rewards/rejected": -0.8977251052856445,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.2093692750588851,
|
| 163 |
+
"grad_norm": 7.1928607065555585,
|
| 164 |
+
"learning_rate": 2.892551899524109e-07,
|
| 165 |
+
"logits/chosen": -1.8947665691375732,
|
| 166 |
+
"logits/rejected": -2.0735626220703125,
|
| 167 |
+
"logps/chosen": -474.9530334472656,
|
| 168 |
+
"logps/rejected": -498.10223388671875,
|
| 169 |
+
"loss": 0.6341,
|
| 170 |
+
"rewards/accuracies": 0.699999988079071,
|
| 171 |
+
"rewards/chosen": -0.6549180746078491,
|
| 172 |
+
"rewards/margins": 0.23183219134807587,
|
| 173 |
+
"rewards/rejected": -0.886750340461731,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.2093692750588851,
|
| 178 |
+
"eval_logits/chosen": -2.435333490371704,
|
| 179 |
+
"eval_logits/rejected": -2.6189088821411133,
|
| 180 |
+
"eval_logps/chosen": -464.6949462890625,
|
| 181 |
+
"eval_logps/rejected": -483.48297119140625,
|
| 182 |
+
"eval_loss": 0.6222504377365112,
|
| 183 |
+
"eval_rewards/accuracies": 0.7103174328804016,
|
| 184 |
+
"eval_rewards/chosen": -0.6020920872688293,
|
| 185 |
+
"eval_rewards/margins": 0.2758086025714874,
|
| 186 |
+
"eval_rewards/rejected": -0.8779006600379944,
|
| 187 |
+
"eval_runtime": 40.059,
|
| 188 |
+
"eval_samples_per_second": 49.926,
|
| 189 |
+
"eval_steps_per_second": 1.573,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.23030620256477363,
|
| 194 |
+
"grad_norm": 7.415850428843577,
|
| 195 |
+
"learning_rate": 2.848030518377739e-07,
|
| 196 |
+
"logits/chosen": -1.8223705291748047,
|
| 197 |
+
"logits/rejected": -2.2238452434539795,
|
| 198 |
+
"logps/chosen": -484.059814453125,
|
| 199 |
+
"logps/rejected": -462.3560485839844,
|
| 200 |
+
"loss": 0.6067,
|
| 201 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 202 |
+
"rewards/chosen": -0.7122976779937744,
|
| 203 |
+
"rewards/margins": 0.2089848816394806,
|
| 204 |
+
"rewards/rejected": -0.9212825894355774,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.2512431300706621,
|
| 209 |
+
"grad_norm": 10.357041495108014,
|
| 210 |
+
"learning_rate": 2.7962832564252725e-07,
|
| 211 |
+
"logits/chosen": -2.0539345741271973,
|
| 212 |
+
"logits/rejected": -2.1838014125823975,
|
| 213 |
+
"logps/chosen": -490.2127990722656,
|
| 214 |
+
"logps/rejected": -524.0123901367188,
|
| 215 |
+
"loss": 0.6168,
|
| 216 |
+
"rewards/accuracies": 0.668749988079071,
|
| 217 |
+
"rewards/chosen": -0.9230595827102661,
|
| 218 |
+
"rewards/margins": 0.2751563489437103,
|
| 219 |
+
"rewards/rejected": -1.1982159614562988,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.2721800575765506,
|
| 224 |
+
"grad_norm": 8.727812082209544,
|
| 225 |
+
"learning_rate": 2.737587495774764e-07,
|
| 226 |
+
"logits/chosen": -2.0211193561553955,
|
| 227 |
+
"logits/rejected": -2.091749906539917,
|
| 228 |
+
"logps/chosen": -476.74884033203125,
|
| 229 |
+
"logps/rejected": -513.7708740234375,
|
| 230 |
+
"loss": 0.6219,
|
| 231 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 232 |
+
"rewards/chosen": -0.8179013133049011,
|
| 233 |
+
"rewards/margins": 0.2984716296195984,
|
| 234 |
+
"rewards/rejected": -1.116373062133789,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.29311698508243916,
|
| 239 |
+
"grad_norm": 10.936927213853266,
|
| 240 |
+
"learning_rate": 2.672257864741005e-07,
|
| 241 |
+
"logits/chosen": -2.028709888458252,
|
| 242 |
+
"logits/rejected": -2.1645443439483643,
|
| 243 |
+
"logps/chosen": -472.2229919433594,
|
| 244 |
+
"logps/rejected": -513.4671020507812,
|
| 245 |
+
"loss": 0.5967,
|
| 246 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 247 |
+
"rewards/chosen": -0.7414780855178833,
|
| 248 |
+
"rewards/margins": 0.44093647599220276,
|
| 249 |
+
"rewards/rejected": -1.1824145317077637,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.31405391258832765,
|
| 254 |
+
"grad_norm": 11.027705905016305,
|
| 255 |
+
"learning_rate": 2.6006445513357056e-07,
|
| 256 |
+
"logits/chosen": -1.9190442562103271,
|
| 257 |
+
"logits/rejected": -2.1971538066864014,
|
| 258 |
+
"logps/chosen": -533.9457397460938,
|
| 259 |
+
"logps/rejected": -540.30322265625,
|
| 260 |
+
"loss": 0.6027,
|
| 261 |
+
"rewards/accuracies": 0.6875,
|
| 262 |
+
"rewards/chosen": -1.019513726234436,
|
| 263 |
+
"rewards/margins": 0.4124508500099182,
|
| 264 |
+
"rewards/rejected": -1.431964635848999,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.33499084009421615,
|
| 269 |
+
"grad_norm": 14.081707059974091,
|
| 270 |
+
"learning_rate": 2.523131426146173e-07,
|
| 271 |
+
"logits/chosen": -2.0433707237243652,
|
| 272 |
+
"logits/rejected": -2.262349843978882,
|
| 273 |
+
"logps/chosen": -504.19873046875,
|
| 274 |
+
"logps/rejected": -551.8690185546875,
|
| 275 |
+
"loss": 0.5571,
|
| 276 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 277 |
+
"rewards/chosen": -1.0314991474151611,
|
| 278 |
+
"rewards/margins": 0.5809075832366943,
|
| 279 |
+
"rewards/rejected": -1.6124067306518555,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.3559277676001047,
|
| 284 |
+
"grad_norm": 11.826987567285538,
|
| 285 |
+
"learning_rate": 2.440133984664454e-07,
|
| 286 |
+
"logits/chosen": -1.9292476177215576,
|
| 287 |
+
"logits/rejected": -2.1319830417633057,
|
| 288 |
+
"logps/chosen": -521.266845703125,
|
| 289 |
+
"logps/rejected": -494.2020568847656,
|
| 290 |
+
"loss": 0.5976,
|
| 291 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 292 |
+
"rewards/chosen": -1.065916657447815,
|
| 293 |
+
"rewards/margins": 0.2420383244752884,
|
| 294 |
+
"rewards/rejected": -1.307955026626587,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.3768646951059932,
|
| 299 |
+
"grad_norm": 9.632333751579734,
|
| 300 |
+
"learning_rate": 2.3520971200967334e-07,
|
| 301 |
+
"logits/chosen": -2.046644449234009,
|
| 302 |
+
"logits/rejected": -2.234806537628174,
|
| 303 |
+
"logps/chosen": -473.70770263671875,
|
| 304 |
+
"logps/rejected": -492.889404296875,
|
| 305 |
+
"loss": 0.5926,
|
| 306 |
+
"rewards/accuracies": 0.6875,
|
| 307 |
+
"rewards/chosen": -0.6946296095848083,
|
| 308 |
+
"rewards/margins": 0.37669837474823,
|
| 309 |
+
"rewards/rejected": -1.0713279247283936,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.39780162261188173,
|
| 314 |
+
"grad_norm": 11.984832910221622,
|
| 315 |
+
"learning_rate": 2.2594927385914543e-07,
|
| 316 |
+
"logits/chosen": -1.9993702173233032,
|
| 317 |
+
"logits/rejected": -2.236987590789795,
|
| 318 |
+
"logps/chosen": -489.000244140625,
|
| 319 |
+
"logps/rejected": -519.3321533203125,
|
| 320 |
+
"loss": 0.5755,
|
| 321 |
+
"rewards/accuracies": 0.768750011920929,
|
| 322 |
+
"rewards/chosen": -1.0067808628082275,
|
| 323 |
+
"rewards/margins": 0.5224589705467224,
|
| 324 |
+
"rewards/rejected": -1.5292397737503052,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.4187385501177702,
|
| 329 |
+
"grad_norm": 12.816716120801315,
|
| 330 |
+
"learning_rate": 2.162817229669295e-07,
|
| 331 |
+
"logits/chosen": -2.1309211254119873,
|
| 332 |
+
"logits/rejected": -2.3286731243133545,
|
| 333 |
+
"logps/chosen": -511.89959716796875,
|
| 334 |
+
"logps/rejected": -556.114990234375,
|
| 335 |
+
"loss": 0.5887,
|
| 336 |
+
"rewards/accuracies": 0.71875,
|
| 337 |
+
"rewards/chosen": -1.1402111053466797,
|
| 338 |
+
"rewards/margins": 0.55492103099823,
|
| 339 |
+
"rewards/rejected": -1.6951322555541992,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.4187385501177702,
|
| 344 |
+
"eval_logits/chosen": -2.485398769378662,
|
| 345 |
+
"eval_logits/rejected": -2.6507983207702637,
|
| 346 |
+
"eval_logps/chosen": -509.5345764160156,
|
| 347 |
+
"eval_logps/rejected": -555.6263427734375,
|
| 348 |
+
"eval_loss": 0.5795689821243286,
|
| 349 |
+
"eval_rewards/accuracies": 0.7142857313156128,
|
| 350 |
+
"eval_rewards/chosen": -1.0504883527755737,
|
| 351 |
+
"eval_rewards/margins": 0.5488466620445251,
|
| 352 |
+
"eval_rewards/rejected": -1.599334955215454,
|
| 353 |
+
"eval_runtime": 38.586,
|
| 354 |
+
"eval_samples_per_second": 51.832,
|
| 355 |
+
"eval_steps_per_second": 1.633,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.4396754776236587,
|
| 360 |
+
"grad_norm": 12.909762723332605,
|
| 361 |
+
"learning_rate": 2.0625888054143427e-07,
|
| 362 |
+
"logits/chosen": -2.1252217292785645,
|
| 363 |
+
"logits/rejected": -2.271873950958252,
|
| 364 |
+
"logps/chosen": -487.28045654296875,
|
| 365 |
+
"logps/rejected": -531.2611083984375,
|
| 366 |
+
"loss": 0.5759,
|
| 367 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 368 |
+
"rewards/chosen": -1.0537116527557373,
|
| 369 |
+
"rewards/margins": 0.4694788455963135,
|
| 370 |
+
"rewards/rejected": -1.5231904983520508,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.46061240512954726,
|
| 375 |
+
"grad_norm": 14.088630852147947,
|
| 376 |
+
"learning_rate": 1.9593447226892386e-07,
|
| 377 |
+
"logits/chosen": -2.2382607460021973,
|
| 378 |
+
"logits/rejected": -2.3763673305511475,
|
| 379 |
+
"logps/chosen": -526.3925170898438,
|
| 380 |
+
"logps/rejected": -573.842529296875,
|
| 381 |
+
"loss": 0.5762,
|
| 382 |
+
"rewards/accuracies": 0.75,
|
| 383 |
+
"rewards/chosen": -1.2542815208435059,
|
| 384 |
+
"rewards/margins": 0.5530741810798645,
|
| 385 |
+
"rewards/rejected": -1.807355523109436,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.48154933263543576,
|
| 390 |
+
"grad_norm": 15.369265950204866,
|
| 391 |
+
"learning_rate": 1.853638403264141e-07,
|
| 392 |
+
"logits/chosen": -2.082986354827881,
|
| 393 |
+
"logits/rejected": -2.2771239280700684,
|
| 394 |
+
"logps/chosen": -481.3154296875,
|
| 395 |
+
"logps/rejected": -538.543212890625,
|
| 396 |
+
"loss": 0.5646,
|
| 397 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 398 |
+
"rewards/chosen": -0.9790234565734863,
|
| 399 |
+
"rewards/margins": 0.5569753050804138,
|
| 400 |
+
"rewards/rejected": -1.535998821258545,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.5024862601413242,
|
| 405 |
+
"grad_norm": 14.740154999219994,
|
| 406 |
+
"learning_rate": 1.7460364672965327e-07,
|
| 407 |
+
"logits/chosen": -2.0640745162963867,
|
| 408 |
+
"logits/rejected": -2.2337191104888916,
|
| 409 |
+
"logps/chosen": -490.3001403808594,
|
| 410 |
+
"logps/rejected": -569.6817626953125,
|
| 411 |
+
"loss": 0.5559,
|
| 412 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 413 |
+
"rewards/chosen": -1.325878381729126,
|
| 414 |
+
"rewards/margins": 0.620599627494812,
|
| 415 |
+
"rewards/rejected": -1.9464781284332275,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5234231876472127,
|
| 420 |
+
"grad_norm": 11.893999213827813,
|
| 421 |
+
"learning_rate": 1.6371156960634017e-07,
|
| 422 |
+
"logits/chosen": -2.005420684814453,
|
| 423 |
+
"logits/rejected": -2.3055458068847656,
|
| 424 |
+
"logps/chosen": -528.0526733398438,
|
| 425 |
+
"logps/rejected": -554.3103637695312,
|
| 426 |
+
"loss": 0.5972,
|
| 427 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 428 |
+
"rewards/chosen": -1.2345831394195557,
|
| 429 |
+
"rewards/margins": 0.4900982975959778,
|
| 430 |
+
"rewards/rejected": -1.7246814966201782,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.5443601151531012,
|
| 435 |
+
"grad_norm": 12.09341713044732,
|
| 436 |
+
"learning_rate": 1.527459940226516e-07,
|
| 437 |
+
"logits/chosen": -2.0832793712615967,
|
| 438 |
+
"logits/rejected": -2.202723741531372,
|
| 439 |
+
"logps/chosen": -532.1043701171875,
|
| 440 |
+
"logps/rejected": -550.4468994140625,
|
| 441 |
+
"loss": 0.5794,
|
| 442 |
+
"rewards/accuracies": 0.699999988079071,
|
| 443 |
+
"rewards/chosen": -1.092650294303894,
|
| 444 |
+
"rewards/margins": 0.4655452370643616,
|
| 445 |
+
"rewards/rejected": -1.5581955909729004,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.5652970426589898,
|
| 450 |
+
"grad_norm": 13.578869062104298,
|
| 451 |
+
"learning_rate": 1.4176569902035086e-07,
|
| 452 |
+
"logits/chosen": -2.2014851570129395,
|
| 453 |
+
"logits/rejected": -2.173820972442627,
|
| 454 |
+
"logps/chosen": -536.9981689453125,
|
| 455 |
+
"logps/rejected": -623.5565185546875,
|
| 456 |
+
"loss": 0.5642,
|
| 457 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 458 |
+
"rewards/chosen": -1.3192862272262573,
|
| 459 |
+
"rewards/margins": 0.5946845412254333,
|
| 460 |
+
"rewards/rejected": -1.913970708847046,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.5862339701648783,
|
| 465 |
+
"grad_norm": 11.79408971766642,
|
| 466 |
+
"learning_rate": 1.3082954254205928e-07,
|
| 467 |
+
"logits/chosen": -2.1410746574401855,
|
| 468 |
+
"logits/rejected": -2.3836779594421387,
|
| 469 |
+
"logps/chosen": -525.254638671875,
|
| 470 |
+
"logps/rejected": -562.7789306640625,
|
| 471 |
+
"loss": 0.5766,
|
| 472 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 473 |
+
"rewards/chosen": -1.3711321353912354,
|
| 474 |
+
"rewards/margins": 0.5404887199401855,
|
| 475 |
+
"rewards/rejected": -1.9116207361221313,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.6071708976707668,
|
| 480 |
+
"grad_norm": 11.223086706265748,
|
| 481 |
+
"learning_rate": 1.1999614593359334e-07,
|
| 482 |
+
"logits/chosen": -2.239353895187378,
|
| 483 |
+
"logits/rejected": -2.298745632171631,
|
| 484 |
+
"logps/chosen": -540.8309936523438,
|
| 485 |
+
"logps/rejected": -590.9267578125,
|
| 486 |
+
"loss": 0.5838,
|
| 487 |
+
"rewards/accuracies": 0.699999988079071,
|
| 488 |
+
"rewards/chosen": -1.269129991531372,
|
| 489 |
+
"rewards/margins": 0.5602521300315857,
|
| 490 |
+
"rewards/rejected": -1.829382300376892,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6281078251766553,
|
| 495 |
+
"grad_norm": 11.552990511047495,
|
| 496 |
+
"learning_rate": 1.0932357971453743e-07,
|
| 497 |
+
"logits/chosen": -2.2803478240966797,
|
| 498 |
+
"logits/rejected": -2.3654093742370605,
|
| 499 |
+
"logps/chosen": -488.808837890625,
|
| 500 |
+
"logps/rejected": -528.6336059570312,
|
| 501 |
+
"loss": 0.5667,
|
| 502 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 503 |
+
"rewards/chosen": -1.0937414169311523,
|
| 504 |
+
"rewards/margins": 0.4998832643032074,
|
| 505 |
+
"rewards/rejected": -1.5936245918273926,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.6281078251766553,
|
| 510 |
+
"eval_logits/chosen": -2.511992931365967,
|
| 511 |
+
"eval_logits/rejected": -2.6684465408325195,
|
| 512 |
+
"eval_logps/chosen": -508.75390625,
|
| 513 |
+
"eval_logps/rejected": -557.6055297851562,
|
| 514 |
+
"eval_loss": 0.5652720928192139,
|
| 515 |
+
"eval_rewards/accuracies": 0.7222222089767456,
|
| 516 |
+
"eval_rewards/chosen": -1.0426820516586304,
|
| 517 |
+
"eval_rewards/margins": 0.5764438509941101,
|
| 518 |
+
"eval_rewards/rejected": -1.6191262006759644,
|
| 519 |
+
"eval_runtime": 38.5252,
|
| 520 |
+
"eval_samples_per_second": 51.914,
|
| 521 |
+
"eval_steps_per_second": 1.635,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.6490447526825438,
|
| 526 |
+
"grad_norm": 11.573976352456944,
|
| 527 |
+
"learning_rate": 9.886905230142431e-08,
|
| 528 |
+
"logits/chosen": -2.1670825481414795,
|
| 529 |
+
"logits/rejected": -2.3857502937316895,
|
| 530 |
+
"logps/chosen": -506.4759216308594,
|
| 531 |
+
"logps/rejected": -566.917724609375,
|
| 532 |
+
"loss": 0.5538,
|
| 533 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 534 |
+
"rewards/chosen": -1.1441192626953125,
|
| 535 |
+
"rewards/margins": 0.5939659476280212,
|
| 536 |
+
"rewards/rejected": -1.738085389137268,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.6699816801884323,
|
| 541 |
+
"grad_norm": 11.811853616414988,
|
| 542 |
+
"learning_rate": 8.868860335206678e-08,
|
| 543 |
+
"logits/chosen": -2.2445249557495117,
|
| 544 |
+
"logits/rejected": -2.408531665802002,
|
| 545 |
+
"logps/chosen": -516.3126220703125,
|
| 546 |
+
"logps/rejected": -578.1045532226562,
|
| 547 |
+
"loss": 0.5687,
|
| 548 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 549 |
+
"rewards/chosen": -1.3060988187789917,
|
| 550 |
+
"rewards/margins": 0.6043392419815063,
|
| 551 |
+
"rewards/rejected": -1.9104379415512085,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.6909186076943209,
|
| 556 |
+
"grad_norm": 15.067885639931706,
|
| 557 |
+
"learning_rate": 7.883680337481599e-08,
|
| 558 |
+
"logits/chosen": -2.2785134315490723,
|
| 559 |
+
"logits/rejected": -2.36972713470459,
|
| 560 |
+
"logps/chosen": -516.9320678710938,
|
| 561 |
+
"logps/rejected": -572.5419311523438,
|
| 562 |
+
"loss": 0.5766,
|
| 563 |
+
"rewards/accuracies": 0.75,
|
| 564 |
+
"rewards/chosen": -1.1142256259918213,
|
| 565 |
+
"rewards/margins": 0.5552430748939514,
|
| 566 |
+
"rewards/rejected": -1.669468641281128,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.7118555352002094,
|
| 571 |
+
"grad_norm": 11.201673342190341,
|
| 572 |
+
"learning_rate": 6.936646121293653e-08,
|
| 573 |
+
"logits/chosen": -2.2320239543914795,
|
| 574 |
+
"logits/rejected": -2.4628634452819824,
|
| 575 |
+
"logps/chosen": -464.08502197265625,
|
| 576 |
+
"logps/rejected": -495.380859375,
|
| 577 |
+
"loss": 0.576,
|
| 578 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 579 |
+
"rewards/chosen": -1.0917171239852905,
|
| 580 |
+
"rewards/margins": 0.5427553653717041,
|
| 581 |
+
"rewards/rejected": -1.6344726085662842,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.7327924627060979,
|
| 586 |
+
"grad_norm": 11.655689256355245,
|
| 587 |
+
"learning_rate": 6.032834097207888e-08,
|
| 588 |
+
"logits/chosen": -2.152733325958252,
|
| 589 |
+
"logits/rejected": -2.451592445373535,
|
| 590 |
+
"logps/chosen": -545.7279663085938,
|
| 591 |
+
"logps/rejected": -572.2804565429688,
|
| 592 |
+
"loss": 0.569,
|
| 593 |
+
"rewards/accuracies": 0.6875,
|
| 594 |
+
"rewards/chosen": -1.1281747817993164,
|
| 595 |
+
"rewards/margins": 0.5794798731803894,
|
| 596 |
+
"rewards/rejected": -1.7076547145843506,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.7537293902119864,
|
| 601 |
+
"grad_norm": 13.61262579360644,
|
| 602 |
+
"learning_rate": 5.177088990820725e-08,
|
| 603 |
+
"logits/chosen": -2.3029279708862305,
|
| 604 |
+
"logits/rejected": -2.4399266242980957,
|
| 605 |
+
"logps/chosen": -488.69921875,
|
| 606 |
+
"logps/rejected": -547.46142578125,
|
| 607 |
+
"loss": 0.572,
|
| 608 |
+
"rewards/accuracies": 0.75,
|
| 609 |
+
"rewards/chosen": -1.1252793073654175,
|
| 610 |
+
"rewards/margins": 0.5917247533798218,
|
| 611 |
+
"rewards/rejected": -1.7170041799545288,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.7746663177178749,
|
| 616 |
+
"grad_norm": 12.249704332676693,
|
| 617 |
+
"learning_rate": 4.373997873459449e-08,
|
| 618 |
+
"logits/chosen": -2.2016916275024414,
|
| 619 |
+
"logits/rejected": -2.460952043533325,
|
| 620 |
+
"logps/chosen": -532.7511596679688,
|
| 621 |
+
"logps/rejected": -570.0086669921875,
|
| 622 |
+
"loss": 0.5523,
|
| 623 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 624 |
+
"rewards/chosen": -1.1311514377593994,
|
| 625 |
+
"rewards/margins": 0.6326709985733032,
|
| 626 |
+
"rewards/rejected": -1.7638225555419922,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.7956032452237635,
|
| 631 |
+
"grad_norm": 16.95867889347482,
|
| 632 |
+
"learning_rate": 3.627865573992087e-08,
|
| 633 |
+
"logits/chosen": -2.2143394947052,
|
| 634 |
+
"logits/rejected": -2.3504748344421387,
|
| 635 |
+
"logps/chosen": -514.0362548828125,
|
| 636 |
+
"logps/rejected": -542.30712890625,
|
| 637 |
+
"loss": 0.5731,
|
| 638 |
+
"rewards/accuracies": 0.6875,
|
| 639 |
+
"rewards/chosen": -1.1771132946014404,
|
| 640 |
+
"rewards/margins": 0.5136756300926208,
|
| 641 |
+
"rewards/rejected": -1.690788984298706,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.816540172729652,
|
| 646 |
+
"grad_norm": 12.40383909871666,
|
| 647 |
+
"learning_rate": 2.942691603548416e-08,
|
| 648 |
+
"logits/chosen": -2.1582541465759277,
|
| 649 |
+
"logits/rejected": -2.3233611583709717,
|
| 650 |
+
"logps/chosen": -540.1260375976562,
|
| 651 |
+
"logps/rejected": -588.2706909179688,
|
| 652 |
+
"loss": 0.5499,
|
| 653 |
+
"rewards/accuracies": 0.699999988079071,
|
| 654 |
+
"rewards/chosen": -1.2853848934173584,
|
| 655 |
+
"rewards/margins": 0.5986013412475586,
|
| 656 |
+
"rewards/rejected": -1.883986473083496,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.8374771002355405,
|
| 661 |
+
"grad_norm": 15.729235077567395,
|
| 662 |
+
"learning_rate": 2.322148716843081e-08,
|
| 663 |
+
"logits/chosen": -2.103240966796875,
|
| 664 |
+
"logits/rejected": -2.3556132316589355,
|
| 665 |
+
"logps/chosen": -578.9059448242188,
|
| 666 |
+
"logps/rejected": -590.7369384765625,
|
| 667 |
+
"loss": 0.5803,
|
| 668 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 669 |
+
"rewards/chosen": -1.4074345827102661,
|
| 670 |
+
"rewards/margins": 0.5764046907424927,
|
| 671 |
+
"rewards/rejected": -1.9838392734527588,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.8374771002355405,
|
| 676 |
+
"eval_logits/chosen": -2.5853030681610107,
|
| 677 |
+
"eval_logits/rejected": -2.7438056468963623,
|
| 678 |
+
"eval_logps/chosen": -540.5127563476562,
|
| 679 |
+
"eval_logps/rejected": -600.98388671875,
|
| 680 |
+
"eval_loss": 0.5581567287445068,
|
| 681 |
+
"eval_rewards/accuracies": 0.726190447807312,
|
| 682 |
+
"eval_rewards/chosen": -1.3602701425552368,
|
| 683 |
+
"eval_rewards/margins": 0.6926405429840088,
|
| 684 |
+
"eval_rewards/rejected": -2.052910804748535,
|
| 685 |
+
"eval_runtime": 38.5105,
|
| 686 |
+
"eval_samples_per_second": 51.934,
|
| 687 |
+
"eval_steps_per_second": 1.636,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.8584140277414289,
|
| 692 |
+
"grad_norm": 13.568875613159559,
|
| 693 |
+
"learning_rate": 1.7695632250191e-08,
|
| 694 |
+
"logits/chosen": -2.2451727390289307,
|
| 695 |
+
"logits/rejected": -2.3778178691864014,
|
| 696 |
+
"logps/chosen": -545.51318359375,
|
| 697 |
+
"logps/rejected": -575.8936157226562,
|
| 698 |
+
"loss": 0.5625,
|
| 699 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 700 |
+
"rewards/chosen": -1.4706048965454102,
|
| 701 |
+
"rewards/margins": 0.4073447287082672,
|
| 702 |
+
"rewards/rejected": -1.8779499530792236,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.8793509552473174,
|
| 707 |
+
"grad_norm": 15.557207200171755,
|
| 708 |
+
"learning_rate": 1.2878971655412513e-08,
|
| 709 |
+
"logits/chosen": -2.2278409004211426,
|
| 710 |
+
"logits/rejected": -2.379141330718994,
|
| 711 |
+
"logps/chosen": -526.79248046875,
|
| 712 |
+
"logps/rejected": -568.4296875,
|
| 713 |
+
"loss": 0.5456,
|
| 714 |
+
"rewards/accuracies": 0.6875,
|
| 715 |
+
"rewards/chosen": -1.2908048629760742,
|
| 716 |
+
"rewards/margins": 0.5515668392181396,
|
| 717 |
+
"rewards/rejected": -1.8423715829849243,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.9002878827532059,
|
| 722 |
+
"grad_norm": 13.028076908432435,
|
| 723 |
+
"learning_rate": 8.79732424714541e-09,
|
| 724 |
+
"logits/chosen": -2.252420663833618,
|
| 725 |
+
"logits/rejected": -2.4495749473571777,
|
| 726 |
+
"logps/chosen": -547.81787109375,
|
| 727 |
+
"logps/rejected": -600.2841186523438,
|
| 728 |
+
"loss": 0.5432,
|
| 729 |
+
"rewards/accuracies": 0.731249988079071,
|
| 730 |
+
"rewards/chosen": -1.333070993423462,
|
| 731 |
+
"rewards/margins": 0.7184783220291138,
|
| 732 |
+
"rewards/rejected": -2.051549196243286,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9212248102590945,
|
| 737 |
+
"grad_norm": 11.81802801299209,
|
| 738 |
+
"learning_rate": 5.472568979361852e-09,
|
| 739 |
+
"logits/chosen": -2.3049111366271973,
|
| 740 |
+
"logits/rejected": -2.2651658058166504,
|
| 741 |
+
"logps/chosen": -557.698486328125,
|
| 742 |
+
"logps/rejected": -628.7259521484375,
|
| 743 |
+
"loss": 0.5597,
|
| 744 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 745 |
+
"rewards/chosen": -1.3383581638336182,
|
| 746 |
+
"rewards/margins": 0.5372647047042847,
|
| 747 |
+
"rewards/rejected": -1.8756227493286133,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.942161737764983,
|
| 752 |
+
"grad_norm": 13.690789391511412,
|
| 753 |
+
"learning_rate": 2.922527618666465e-09,
|
| 754 |
+
"logits/chosen": -2.1817500591278076,
|
| 755 |
+
"logits/rejected": -2.4527294635772705,
|
| 756 |
+
"logps/chosen": -550.814697265625,
|
| 757 |
+
"logps/rejected": -563.7845458984375,
|
| 758 |
+
"loss": 0.5551,
|
| 759 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 760 |
+
"rewards/chosen": -1.2616031169891357,
|
| 761 |
+
"rewards/margins": 0.5693789720535278,
|
| 762 |
+
"rewards/rejected": -1.8309822082519531,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.9630986652708715,
|
| 767 |
+
"grad_norm": 16.036131164059057,
|
| 768 |
+
"learning_rate": 1.1608692138469378e-09,
|
| 769 |
+
"logits/chosen": -2.184504508972168,
|
| 770 |
+
"logits/rejected": -2.432163715362549,
|
| 771 |
+
"logps/chosen": -579.7196044921875,
|
| 772 |
+
"logps/rejected": -595.66162109375,
|
| 773 |
+
"loss": 0.5728,
|
| 774 |
+
"rewards/accuracies": 0.699999988079071,
|
| 775 |
+
"rewards/chosen": -1.4352037906646729,
|
| 776 |
+
"rewards/margins": 0.5179115533828735,
|
| 777 |
+
"rewards/rejected": -1.9531152248382568,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.98403559277676,
|
| 782 |
+
"grad_norm": 10.779013615465365,
|
| 783 |
+
"learning_rate": 1.9703682533901976e-10,
|
| 784 |
+
"logits/chosen": -2.1344995498657227,
|
| 785 |
+
"logits/rejected": -2.3982386589050293,
|
| 786 |
+
"logps/chosen": -520.6542358398438,
|
| 787 |
+
"logps/rejected": -542.9959716796875,
|
| 788 |
+
"loss": 0.5528,
|
| 789 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 790 |
+
"rewards/chosen": -1.2908748388290405,
|
| 791 |
+
"rewards/margins": 0.5777503252029419,
|
| 792 |
+
"rewards/rejected": -1.8686254024505615,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.998691442030882,
|
| 797 |
+
"step": 477,
|
| 798 |
+
"total_flos": 0.0,
|
| 799 |
+
"train_loss": 0.5939158873488068,
|
| 800 |
+
"train_runtime": 3918.7641,
|
| 801 |
+
"train_samples_per_second": 15.601,
|
| 802 |
+
"train_steps_per_second": 0.122
|
| 803 |
+
}
|
| 804 |
+
],
|
| 805 |
+
"logging_steps": 10,
|
| 806 |
+
"max_steps": 477,
|
| 807 |
+
"num_input_tokens_seen": 0,
|
| 808 |
+
"num_train_epochs": 1,
|
| 809 |
+
"save_steps": 100,
|
| 810 |
+
"stateful_callbacks": {
|
| 811 |
+
"TrainerControl": {
|
| 812 |
+
"args": {
|
| 813 |
+
"should_epoch_stop": false,
|
| 814 |
+
"should_evaluate": false,
|
| 815 |
+
"should_log": false,
|
| 816 |
+
"should_save": true,
|
| 817 |
+
"should_training_stop": true
|
| 818 |
+
},
|
| 819 |
+
"attributes": {}
|
| 820 |
+
}
|
| 821 |
+
},
|
| 822 |
+
"total_flos": 0.0,
|
| 823 |
+
"train_batch_size": 2,
|
| 824 |
+
"trial_name": null,
|
| 825 |
+
"trial_params": null
|
| 826 |
+
}
|