Model save
Browse files- README.md +80 -0
- all_results.json +21 -0
- eval_results.json +16 -0
- generation_config.json +7 -0
- model.safetensors +1 -1
- runs/Jan26_08-44-51_main1/events.out.tfevents.1706259102.main1.71884.0 +2 -2
- runs/Jan26_08-44-51_main1/events.out.tfevents.1706276579.main1.71884.1 +3 -0
- train_results.json +8 -0
- trainer_state.json +1518 -0
README.md
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: ondevicellm/tinyllama_mole_sft_ultrachat_ep3
|
| 3 |
+
tags:
|
| 4 |
+
- trl
|
| 5 |
+
- dpo
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
model-index:
|
| 8 |
+
- name: tinyllama_mole_dpo_ep3
|
| 9 |
+
results: []
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 13 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 14 |
+
|
| 15 |
+
# tinyllama_mole_dpo_ep3
|
| 16 |
+
|
| 17 |
+
This model is a fine-tuned version of [ondevicellm/tinyllama_mole_sft_ultrachat_ep3](https://huggingface.co/ondevicellm/tinyllama_mole_sft_ultrachat_ep3) on the None dataset.
|
| 18 |
+
It achieves the following results on the evaluation set:
|
| 19 |
+
- Loss: 0.6285
|
| 20 |
+
- Rewards/chosen: -0.3050
|
| 21 |
+
- Rewards/rejected: -0.5353
|
| 22 |
+
- Rewards/accuracies: 0.6806
|
| 23 |
+
- Rewards/margins: 0.2302
|
| 24 |
+
- Logps/rejected: -354.2071
|
| 25 |
+
- Logps/chosen: -373.1399
|
| 26 |
+
- Logits/rejected: -1.6731
|
| 27 |
+
- Logits/chosen: -1.8041
|
| 28 |
+
|
| 29 |
+
## Model description
|
| 30 |
+
|
| 31 |
+
More information needed
|
| 32 |
+
|
| 33 |
+
## Intended uses & limitations
|
| 34 |
+
|
| 35 |
+
More information needed
|
| 36 |
+
|
| 37 |
+
## Training and evaluation data
|
| 38 |
+
|
| 39 |
+
More information needed
|
| 40 |
+
|
| 41 |
+
## Training procedure
|
| 42 |
+
|
| 43 |
+
### Training hyperparameters
|
| 44 |
+
|
| 45 |
+
The following hyperparameters were used during training:
|
| 46 |
+
- learning_rate: 5e-07
|
| 47 |
+
- train_batch_size: 8
|
| 48 |
+
- eval_batch_size: 8
|
| 49 |
+
- seed: 42
|
| 50 |
+
- distributed_type: multi-GPU
|
| 51 |
+
- num_devices: 4
|
| 52 |
+
- gradient_accumulation_steps: 2
|
| 53 |
+
- total_train_batch_size: 64
|
| 54 |
+
- total_eval_batch_size: 32
|
| 55 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 56 |
+
- lr_scheduler_type: cosine
|
| 57 |
+
- lr_scheduler_warmup_steps: 100
|
| 58 |
+
- num_epochs: 1
|
| 59 |
+
|
| 60 |
+
### Training results
|
| 61 |
+
|
| 62 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 63 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 64 |
+
| 0.6896 | 0.1 | 100 | 0.6899 | 0.0064 | -0.0013 | 0.6448 | 0.0076 | -300.8089 | -342.0017 | -1.7574 | -1.8918 |
|
| 65 |
+
| 0.6762 | 0.21 | 200 | 0.6756 | -0.0293 | -0.0716 | 0.6627 | 0.0423 | -307.8423 | -345.5688 | -1.7501 | -1.8839 |
|
| 66 |
+
| 0.6499 | 0.31 | 300 | 0.6587 | -0.0875 | -0.1813 | 0.6687 | 0.0938 | -318.8118 | -351.3895 | -1.7358 | -1.8688 |
|
| 67 |
+
| 0.6374 | 0.42 | 400 | 0.6451 | -0.1726 | -0.3218 | 0.6746 | 0.1493 | -332.8632 | -359.8953 | -1.7164 | -1.8482 |
|
| 68 |
+
| 0.6348 | 0.52 | 500 | 0.6377 | -0.2696 | -0.4550 | 0.6647 | 0.1854 | -346.1808 | -369.6013 | -1.6884 | -1.8208 |
|
| 69 |
+
| 0.6308 | 0.63 | 600 | 0.6333 | -0.2783 | -0.4815 | 0.6726 | 0.2032 | -348.8291 | -370.4673 | -1.6965 | -1.8269 |
|
| 70 |
+
| 0.62 | 0.73 | 700 | 0.6312 | -0.2323 | -0.4505 | 0.6806 | 0.2182 | -345.7306 | -365.8656 | -1.6841 | -1.8149 |
|
| 71 |
+
| 0.6055 | 0.84 | 800 | 0.6287 | -0.2877 | -0.5169 | 0.6865 | 0.2292 | -352.3697 | -371.4099 | -1.6793 | -1.8099 |
|
| 72 |
+
| 0.6357 | 0.94 | 900 | 0.6285 | -0.3050 | -0.5353 | 0.6806 | 0.2302 | -354.2071 | -373.1399 | -1.6731 | -1.8041 |
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
### Framework versions
|
| 76 |
+
|
| 77 |
+
- Transformers 4.37.0
|
| 78 |
+
- Pytorch 2.1.2+cu118
|
| 79 |
+
- Datasets 2.16.1
|
| 80 |
+
- Tokenizers 0.15.0
|
all_results.json
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"eval_logits/chosen": -1.8041099309921265,
|
| 4 |
+
"eval_logits/rejected": -1.6730901002883911,
|
| 5 |
+
"eval_logps/chosen": -373.1398620605469,
|
| 6 |
+
"eval_logps/rejected": -354.2070617675781,
|
| 7 |
+
"eval_loss": 0.6285176873207092,
|
| 8 |
+
"eval_rewards/accuracies": 0.6805555820465088,
|
| 9 |
+
"eval_rewards/chosen": -0.30501797795295715,
|
| 10 |
+
"eval_rewards/margins": 0.23024281859397888,
|
| 11 |
+
"eval_rewards/rejected": -0.535260796546936,
|
| 12 |
+
"eval_runtime": 337.0045,
|
| 13 |
+
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 5.935,
|
| 15 |
+
"eval_steps_per_second": 0.187,
|
| 16 |
+
"train_loss": 0.6456358959537526,
|
| 17 |
+
"train_runtime": 17140.5232,
|
| 18 |
+
"train_samples": 61135,
|
| 19 |
+
"train_samples_per_second": 3.567,
|
| 20 |
+
"train_steps_per_second": 0.056
|
| 21 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"eval_logits/chosen": -1.8041099309921265,
|
| 4 |
+
"eval_logits/rejected": -1.6730901002883911,
|
| 5 |
+
"eval_logps/chosen": -373.1398620605469,
|
| 6 |
+
"eval_logps/rejected": -354.2070617675781,
|
| 7 |
+
"eval_loss": 0.6285176873207092,
|
| 8 |
+
"eval_rewards/accuracies": 0.6805555820465088,
|
| 9 |
+
"eval_rewards/chosen": -0.30501797795295715,
|
| 10 |
+
"eval_rewards/margins": 0.23024281859397888,
|
| 11 |
+
"eval_rewards/rejected": -0.535260796546936,
|
| 12 |
+
"eval_runtime": 337.0045,
|
| 13 |
+
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 5.935,
|
| 15 |
+
"eval_steps_per_second": 0.187
|
| 16 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"transformers_version": "4.37.0",
|
| 6 |
+
"use_cache": false
|
| 7 |
+
}
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 2223960880
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:31732c1d87bb760b754f5718db58cb66cb3423c26baf7a1f0855ab58a2d9d493
|
| 3 |
size 2223960880
|
runs/Jan26_08-44-51_main1/events.out.tfevents.1706259102.main1.71884.0
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 sha256:477da43a49b5ad27fe6bf8a92038c819502ebab2ee0ae51907ecae0421544580
|
| 3 |
+
size 72954
|
runs/Jan26_08-44-51_main1/events.out.tfevents.1706276579.main1.71884.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ff51a67ac05583c8159b204d729a64cd493901238fa240da1c0a36688ca5e16b
|
| 3 |
+
size 828
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.6456358959537526,
|
| 4 |
+
"train_runtime": 17140.5232,
|
| 5 |
+
"train_samples": 61135,
|
| 6 |
+
"train_samples_per_second": 3.567,
|
| 7 |
+
"train_steps_per_second": 0.056
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1518 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 0.6285176873207092,
|
| 3 |
+
"best_model_checkpoint": "data/tinyllama_mole_dpo_ep3/checkpoint-900",
|
| 4 |
+
"epoch": 0.9994767137624281,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 955,
|
| 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 |
+
"learning_rate": 5e-09,
|
| 14 |
+
"logits/chosen": -1.790444016456604,
|
| 15 |
+
"logits/rejected": -1.7375602722167969,
|
| 16 |
+
"logps/chosen": -288.11163330078125,
|
| 17 |
+
"logps/rejected": -270.18121337890625,
|
| 18 |
+
"loss": 0.6931,
|
| 19 |
+
"rewards/accuracies": 0.0,
|
| 20 |
+
"rewards/chosen": 0.0,
|
| 21 |
+
"rewards/margins": 0.0,
|
| 22 |
+
"rewards/rejected": 0.0,
|
| 23 |
+
"step": 1
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.01,
|
| 27 |
+
"learning_rate": 5e-08,
|
| 28 |
+
"logits/chosen": -1.9029791355133057,
|
| 29 |
+
"logits/rejected": -1.6527897119522095,
|
| 30 |
+
"logps/chosen": -349.9918518066406,
|
| 31 |
+
"logps/rejected": -281.4112243652344,
|
| 32 |
+
"loss": 0.693,
|
| 33 |
+
"rewards/accuracies": 0.5069444179534912,
|
| 34 |
+
"rewards/chosen": 0.0003158848558086902,
|
| 35 |
+
"rewards/margins": 0.0004107448039576411,
|
| 36 |
+
"rewards/rejected": -9.485996997682378e-05,
|
| 37 |
+
"step": 10
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.02,
|
| 41 |
+
"learning_rate": 1e-07,
|
| 42 |
+
"logits/chosen": -1.773256540298462,
|
| 43 |
+
"logits/rejected": -1.7059326171875,
|
| 44 |
+
"logps/chosen": -306.29168701171875,
|
| 45 |
+
"logps/rejected": -274.650634765625,
|
| 46 |
+
"loss": 0.693,
|
| 47 |
+
"rewards/accuracies": 0.5,
|
| 48 |
+
"rewards/chosen": 0.0002991842629853636,
|
| 49 |
+
"rewards/margins": 0.00016614615742582828,
|
| 50 |
+
"rewards/rejected": 0.000133038149215281,
|
| 51 |
+
"step": 20
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.03,
|
| 55 |
+
"learning_rate": 1.5e-07,
|
| 56 |
+
"logits/chosen": -1.8527847528457642,
|
| 57 |
+
"logits/rejected": -1.707421898841858,
|
| 58 |
+
"logps/chosen": -339.32342529296875,
|
| 59 |
+
"logps/rejected": -297.5170593261719,
|
| 60 |
+
"loss": 0.693,
|
| 61 |
+
"rewards/accuracies": 0.512499988079071,
|
| 62 |
+
"rewards/chosen": -0.0001883753720903769,
|
| 63 |
+
"rewards/margins": -0.00028402512543834746,
|
| 64 |
+
"rewards/rejected": 9.56498843152076e-05,
|
| 65 |
+
"step": 30
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.04,
|
| 69 |
+
"learning_rate": 2e-07,
|
| 70 |
+
"logits/chosen": -1.869114875793457,
|
| 71 |
+
"logits/rejected": -1.7149085998535156,
|
| 72 |
+
"logps/chosen": -329.600830078125,
|
| 73 |
+
"logps/rejected": -274.0981140136719,
|
| 74 |
+
"loss": 0.693,
|
| 75 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 76 |
+
"rewards/chosen": 0.0005581147270277143,
|
| 77 |
+
"rewards/margins": 0.000760135124437511,
|
| 78 |
+
"rewards/rejected": -0.00020202035375405103,
|
| 79 |
+
"step": 40
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.05,
|
| 83 |
+
"learning_rate": 2.5e-07,
|
| 84 |
+
"logits/chosen": -1.8775875568389893,
|
| 85 |
+
"logits/rejected": -1.7631309032440186,
|
| 86 |
+
"logps/chosen": -344.46002197265625,
|
| 87 |
+
"logps/rejected": -307.79693603515625,
|
| 88 |
+
"loss": 0.6928,
|
| 89 |
+
"rewards/accuracies": 0.5,
|
| 90 |
+
"rewards/chosen": 0.0002629683876875788,
|
| 91 |
+
"rewards/margins": 0.0003627274709288031,
|
| 92 |
+
"rewards/rejected": -9.97590395854786e-05,
|
| 93 |
+
"step": 50
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.06,
|
| 97 |
+
"learning_rate": 3e-07,
|
| 98 |
+
"logits/chosen": -1.8707891702651978,
|
| 99 |
+
"logits/rejected": -1.711287498474121,
|
| 100 |
+
"logps/chosen": -345.78326416015625,
|
| 101 |
+
"logps/rejected": -318.0264892578125,
|
| 102 |
+
"loss": 0.6926,
|
| 103 |
+
"rewards/accuracies": 0.574999988079071,
|
| 104 |
+
"rewards/chosen": 0.0013746457407251,
|
| 105 |
+
"rewards/margins": 0.0012839403934776783,
|
| 106 |
+
"rewards/rejected": 9.070520900422707e-05,
|
| 107 |
+
"step": 60
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.07,
|
| 111 |
+
"learning_rate": 3.5e-07,
|
| 112 |
+
"logits/chosen": -1.997075080871582,
|
| 113 |
+
"logits/rejected": -1.8358447551727295,
|
| 114 |
+
"logps/chosen": -379.5538024902344,
|
| 115 |
+
"logps/rejected": -332.5546875,
|
| 116 |
+
"loss": 0.6926,
|
| 117 |
+
"rewards/accuracies": 0.612500011920929,
|
| 118 |
+
"rewards/chosen": 0.001883229473605752,
|
| 119 |
+
"rewards/margins": 0.0018280971562489867,
|
| 120 |
+
"rewards/rejected": 5.51322809769772e-05,
|
| 121 |
+
"step": 70
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.08,
|
| 125 |
+
"learning_rate": 4e-07,
|
| 126 |
+
"logits/chosen": -1.7132848501205444,
|
| 127 |
+
"logits/rejected": -1.5544992685317993,
|
| 128 |
+
"logps/chosen": -352.10699462890625,
|
| 129 |
+
"logps/rejected": -292.5279846191406,
|
| 130 |
+
"loss": 0.6914,
|
| 131 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 132 |
+
"rewards/chosen": 0.0026874816976487637,
|
| 133 |
+
"rewards/margins": 0.0036670640110969543,
|
| 134 |
+
"rewards/rejected": -0.000979582779109478,
|
| 135 |
+
"step": 80
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.09,
|
| 139 |
+
"learning_rate": 4.5e-07,
|
| 140 |
+
"logits/chosen": -1.9277913570404053,
|
| 141 |
+
"logits/rejected": -1.8120664358139038,
|
| 142 |
+
"logps/chosen": -345.2822570800781,
|
| 143 |
+
"logps/rejected": -309.4122314453125,
|
| 144 |
+
"loss": 0.6909,
|
| 145 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 146 |
+
"rewards/chosen": 0.00368087668903172,
|
| 147 |
+
"rewards/margins": 0.004829054698348045,
|
| 148 |
+
"rewards/rejected": -0.0011481784749776125,
|
| 149 |
+
"step": 90
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.1,
|
| 153 |
+
"learning_rate": 5e-07,
|
| 154 |
+
"logits/chosen": -1.9077354669570923,
|
| 155 |
+
"logits/rejected": -1.8217405080795288,
|
| 156 |
+
"logps/chosen": -364.0276184082031,
|
| 157 |
+
"logps/rejected": -340.09906005859375,
|
| 158 |
+
"loss": 0.6896,
|
| 159 |
+
"rewards/accuracies": 0.699999988079071,
|
| 160 |
+
"rewards/chosen": 0.00508699519559741,
|
| 161 |
+
"rewards/margins": 0.007315085269510746,
|
| 162 |
+
"rewards/rejected": -0.002228089142590761,
|
| 163 |
+
"step": 100
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.1,
|
| 167 |
+
"eval_logits/chosen": -1.8918097019195557,
|
| 168 |
+
"eval_logits/rejected": -1.7573894262313843,
|
| 169 |
+
"eval_logps/chosen": -342.0017395019531,
|
| 170 |
+
"eval_logps/rejected": -300.80889892578125,
|
| 171 |
+
"eval_loss": 0.689919114112854,
|
| 172 |
+
"eval_rewards/accuracies": 0.6448412537574768,
|
| 173 |
+
"eval_rewards/chosen": 0.006363342050462961,
|
| 174 |
+
"eval_rewards/margins": 0.007642398122698069,
|
| 175 |
+
"eval_rewards/rejected": -0.0012790567707270384,
|
| 176 |
+
"eval_runtime": 339.8787,
|
| 177 |
+
"eval_samples_per_second": 5.884,
|
| 178 |
+
"eval_steps_per_second": 0.185,
|
| 179 |
+
"step": 100
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 0.12,
|
| 183 |
+
"learning_rate": 4.998312558730158e-07,
|
| 184 |
+
"logits/chosen": -1.8893086910247803,
|
| 185 |
+
"logits/rejected": -1.6524394750595093,
|
| 186 |
+
"logps/chosen": -312.1143493652344,
|
| 187 |
+
"logps/rejected": -249.64523315429688,
|
| 188 |
+
"loss": 0.6887,
|
| 189 |
+
"rewards/accuracies": 0.643750011920929,
|
| 190 |
+
"rewards/chosen": 0.005943284835666418,
|
| 191 |
+
"rewards/margins": 0.008504782803356647,
|
| 192 |
+
"rewards/rejected": -0.002561498200520873,
|
| 193 |
+
"step": 110
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 0.13,
|
| 197 |
+
"learning_rate": 4.993252512887069e-07,
|
| 198 |
+
"logits/chosen": -1.8319896459579468,
|
| 199 |
+
"logits/rejected": -1.6496975421905518,
|
| 200 |
+
"logps/chosen": -358.3703308105469,
|
| 201 |
+
"logps/rejected": -289.1592102050781,
|
| 202 |
+
"loss": 0.6873,
|
| 203 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 204 |
+
"rewards/chosen": 0.009710123762488365,
|
| 205 |
+
"rewards/margins": 0.017875777557492256,
|
| 206 |
+
"rewards/rejected": -0.008165654726326466,
|
| 207 |
+
"step": 120
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 0.14,
|
| 211 |
+
"learning_rate": 4.984826693294873e-07,
|
| 212 |
+
"logits/chosen": -1.9560505151748657,
|
| 213 |
+
"logits/rejected": -1.7518360614776611,
|
| 214 |
+
"logps/chosen": -377.28082275390625,
|
| 215 |
+
"logps/rejected": -308.9075927734375,
|
| 216 |
+
"loss": 0.6866,
|
| 217 |
+
"rewards/accuracies": 0.612500011920929,
|
| 218 |
+
"rewards/chosen": 0.009400544688105583,
|
| 219 |
+
"rewards/margins": 0.014766154810786247,
|
| 220 |
+
"rewards/rejected": -0.005365610122680664,
|
| 221 |
+
"step": 130
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.15,
|
| 225 |
+
"learning_rate": 4.973046474414144e-07,
|
| 226 |
+
"logits/chosen": -1.7957839965820312,
|
| 227 |
+
"logits/rejected": -1.6959806680679321,
|
| 228 |
+
"logps/chosen": -283.5320129394531,
|
| 229 |
+
"logps/rejected": -259.69097900390625,
|
| 230 |
+
"loss": 0.6851,
|
| 231 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 232 |
+
"rewards/chosen": 0.00773229356855154,
|
| 233 |
+
"rewards/margins": 0.01577392965555191,
|
| 234 |
+
"rewards/rejected": -0.00804163608700037,
|
| 235 |
+
"step": 140
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.16,
|
| 239 |
+
"learning_rate": 4.957927758986888e-07,
|
| 240 |
+
"logits/chosen": -1.8573449850082397,
|
| 241 |
+
"logits/rejected": -1.7665218114852905,
|
| 242 |
+
"logps/chosen": -333.4697265625,
|
| 243 |
+
"logps/rejected": -334.0047912597656,
|
| 244 |
+
"loss": 0.6864,
|
| 245 |
+
"rewards/accuracies": 0.625,
|
| 246 |
+
"rewards/chosen": -0.00011548856127774343,
|
| 247 |
+
"rewards/margins": 0.005317127797752619,
|
| 248 |
+
"rewards/rejected": -0.005432615987956524,
|
| 249 |
+
"step": 150
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.17,
|
| 253 |
+
"learning_rate": 4.939490956568589e-07,
|
| 254 |
+
"logits/chosen": -1.9507248401641846,
|
| 255 |
+
"logits/rejected": -1.7802870273590088,
|
| 256 |
+
"logps/chosen": -352.24407958984375,
|
| 257 |
+
"logps/rejected": -330.3514099121094,
|
| 258 |
+
"loss": 0.6815,
|
| 259 |
+
"rewards/accuracies": 0.668749988079071,
|
| 260 |
+
"rewards/chosen": 0.0028652914334088564,
|
| 261 |
+
"rewards/margins": 0.018842989578843117,
|
| 262 |
+
"rewards/rejected": -0.01597769930958748,
|
| 263 |
+
"step": 160
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"epoch": 0.18,
|
| 267 |
+
"learning_rate": 4.917760955976277e-07,
|
| 268 |
+
"logits/chosen": -1.9207957983016968,
|
| 269 |
+
"logits/rejected": -1.7402126789093018,
|
| 270 |
+
"logps/chosen": -322.8128662109375,
|
| 271 |
+
"logps/rejected": -280.6071472167969,
|
| 272 |
+
"loss": 0.6807,
|
| 273 |
+
"rewards/accuracies": 0.65625,
|
| 274 |
+
"rewards/chosen": 0.006607011891901493,
|
| 275 |
+
"rewards/margins": 0.029789209365844727,
|
| 276 |
+
"rewards/rejected": -0.023182198405265808,
|
| 277 |
+
"step": 170
|
| 278 |
+
},
|
| 279 |
+
{
|
| 280 |
+
"epoch": 0.19,
|
| 281 |
+
"learning_rate": 4.892767091689785e-07,
|
| 282 |
+
"logits/chosen": -1.8039575815200806,
|
| 283 |
+
"logits/rejected": -1.6338218450546265,
|
| 284 |
+
"logps/chosen": -298.98541259765625,
|
| 285 |
+
"logps/rejected": -268.31658935546875,
|
| 286 |
+
"loss": 0.6781,
|
| 287 |
+
"rewards/accuracies": 0.668749988079071,
|
| 288 |
+
"rewards/chosen": -0.010788476094603539,
|
| 289 |
+
"rewards/margins": 0.033482082188129425,
|
| 290 |
+
"rewards/rejected": -0.044270556420087814,
|
| 291 |
+
"step": 180
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"epoch": 0.2,
|
| 295 |
+
"learning_rate": 4.864543104251586e-07,
|
| 296 |
+
"logits/chosen": -1.9226608276367188,
|
| 297 |
+
"logits/rejected": -1.7317349910736084,
|
| 298 |
+
"logps/chosen": -342.0778503417969,
|
| 299 |
+
"logps/rejected": -300.3584899902344,
|
| 300 |
+
"loss": 0.6758,
|
| 301 |
+
"rewards/accuracies": 0.675000011920929,
|
| 302 |
+
"rewards/chosen": -0.010154007002711296,
|
| 303 |
+
"rewards/margins": 0.03691417723894119,
|
| 304 |
+
"rewards/rejected": -0.04706818610429764,
|
| 305 |
+
"step": 190
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 0.21,
|
| 309 |
+
"learning_rate": 4.833127094718643e-07,
|
| 310 |
+
"logits/chosen": -1.7831227779388428,
|
| 311 |
+
"logits/rejected": -1.6590824127197266,
|
| 312 |
+
"logps/chosen": -343.6825866699219,
|
| 313 |
+
"logps/rejected": -319.09796142578125,
|
| 314 |
+
"loss": 0.6762,
|
| 315 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 316 |
+
"rewards/chosen": -0.012453018687665462,
|
| 317 |
+
"rewards/margins": 0.04117157310247421,
|
| 318 |
+
"rewards/rejected": -0.05362458899617195,
|
| 319 |
+
"step": 200
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"epoch": 0.21,
|
| 323 |
+
"eval_logits/chosen": -1.883855938911438,
|
| 324 |
+
"eval_logits/rejected": -1.7501325607299805,
|
| 325 |
+
"eval_logps/chosen": -345.56878662109375,
|
| 326 |
+
"eval_logps/rejected": -307.84234619140625,
|
| 327 |
+
"eval_loss": 0.6756463646888733,
|
| 328 |
+
"eval_rewards/accuracies": 0.6626983880996704,
|
| 329 |
+
"eval_rewards/chosen": -0.029307426884770393,
|
| 330 |
+
"eval_rewards/margins": 0.04230639338493347,
|
| 331 |
+
"eval_rewards/rejected": -0.07161381840705872,
|
| 332 |
+
"eval_runtime": 342.3556,
|
| 333 |
+
"eval_samples_per_second": 5.842,
|
| 334 |
+
"eval_steps_per_second": 0.184,
|
| 335 |
+
"step": 200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 0.22,
|
| 339 |
+
"learning_rate": 4.79856147322777e-07,
|
| 340 |
+
"logits/chosen": -1.9185651540756226,
|
| 341 |
+
"logits/rejected": -1.6988731622695923,
|
| 342 |
+
"logps/chosen": -368.73504638671875,
|
| 343 |
+
"logps/rejected": -302.3857421875,
|
| 344 |
+
"loss": 0.6743,
|
| 345 |
+
"rewards/accuracies": 0.643750011920929,
|
| 346 |
+
"rewards/chosen": -0.036936286836862564,
|
| 347 |
+
"rewards/margins": 0.04566096514463425,
|
| 348 |
+
"rewards/rejected": -0.08259725570678711,
|
| 349 |
+
"step": 210
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 0.23,
|
| 353 |
+
"learning_rate": 4.760892901743944e-07,
|
| 354 |
+
"logits/chosen": -1.9230194091796875,
|
| 355 |
+
"logits/rejected": -1.7622359991073608,
|
| 356 |
+
"logps/chosen": -352.57354736328125,
|
| 357 |
+
"logps/rejected": -324.182373046875,
|
| 358 |
+
"loss": 0.6748,
|
| 359 |
+
"rewards/accuracies": 0.65625,
|
| 360 |
+
"rewards/chosen": -0.03244791179895401,
|
| 361 |
+
"rewards/margins": 0.04869867116212845,
|
| 362 |
+
"rewards/rejected": -0.08114659041166306,
|
| 363 |
+
"step": 220
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 0.24,
|
| 367 |
+
"learning_rate": 4.720172231068844e-07,
|
| 368 |
+
"logits/chosen": -1.8404433727264404,
|
| 369 |
+
"logits/rejected": -1.6695423126220703,
|
| 370 |
+
"logps/chosen": -354.01727294921875,
|
| 371 |
+
"logps/rejected": -295.2720642089844,
|
| 372 |
+
"loss": 0.665,
|
| 373 |
+
"rewards/accuracies": 0.668749988079071,
|
| 374 |
+
"rewards/chosen": -0.050288014113903046,
|
| 375 |
+
"rewards/margins": 0.053208403289318085,
|
| 376 |
+
"rewards/rejected": -0.10349641740322113,
|
| 377 |
+
"step": 230
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 0.25,
|
| 381 |
+
"learning_rate": 4.6764544321946557e-07,
|
| 382 |
+
"logits/chosen": -1.851320505142212,
|
| 383 |
+
"logits/rejected": -1.6485137939453125,
|
| 384 |
+
"logps/chosen": -338.0115051269531,
|
| 385 |
+
"logps/rejected": -272.17596435546875,
|
| 386 |
+
"loss": 0.6675,
|
| 387 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 388 |
+
"rewards/chosen": -0.04714337736368179,
|
| 389 |
+
"rewards/margins": 0.05400116369128227,
|
| 390 |
+
"rewards/rejected": -0.10114455223083496,
|
| 391 |
+
"step": 240
|
| 392 |
+
},
|
| 393 |
+
{
|
| 394 |
+
"epoch": 0.26,
|
| 395 |
+
"learning_rate": 4.6297985220958176e-07,
|
| 396 |
+
"logits/chosen": -1.858806848526001,
|
| 397 |
+
"logits/rejected": -1.7875276803970337,
|
| 398 |
+
"logps/chosen": -323.6578369140625,
|
| 399 |
+
"logps/rejected": -327.8287048339844,
|
| 400 |
+
"loss": 0.6704,
|
| 401 |
+
"rewards/accuracies": 0.668749988079071,
|
| 402 |
+
"rewards/chosen": -0.06438130885362625,
|
| 403 |
+
"rewards/margins": 0.054077375680208206,
|
| 404 |
+
"rewards/rejected": -0.11845867335796356,
|
| 405 |
+
"step": 250
|
| 406 |
+
},
|
| 407 |
+
{
|
| 408 |
+
"epoch": 0.27,
|
| 409 |
+
"learning_rate": 4.580267484058875e-07,
|
| 410 |
+
"logits/chosen": -1.787245512008667,
|
| 411 |
+
"logits/rejected": -1.6371400356292725,
|
| 412 |
+
"logps/chosen": -346.79132080078125,
|
| 413 |
+
"logps/rejected": -315.600341796875,
|
| 414 |
+
"loss": 0.6655,
|
| 415 |
+
"rewards/accuracies": 0.668749988079071,
|
| 416 |
+
"rewards/chosen": -0.07440805435180664,
|
| 417 |
+
"rewards/margins": 0.07648531347513199,
|
| 418 |
+
"rewards/rejected": -0.15089336037635803,
|
| 419 |
+
"step": 260
|
| 420 |
+
},
|
| 421 |
+
{
|
| 422 |
+
"epoch": 0.28,
|
| 423 |
+
"learning_rate": 4.527928182658005e-07,
|
| 424 |
+
"logits/chosen": -1.9231889247894287,
|
| 425 |
+
"logits/rejected": -1.7725601196289062,
|
| 426 |
+
"logps/chosen": -357.95367431640625,
|
| 427 |
+
"logps/rejected": -305.06793212890625,
|
| 428 |
+
"loss": 0.6624,
|
| 429 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 430 |
+
"rewards/chosen": -0.0633041113615036,
|
| 431 |
+
"rewards/margins": 0.09316254407167435,
|
| 432 |
+
"rewards/rejected": -0.15646666288375854,
|
| 433 |
+
"step": 270
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.29,
|
| 437 |
+
"learning_rate": 4.472851273490984e-07,
|
| 438 |
+
"logits/chosen": -1.9144046306610107,
|
| 439 |
+
"logits/rejected": -1.8413257598876953,
|
| 440 |
+
"logps/chosen": -356.4827575683594,
|
| 441 |
+
"logps/rejected": -339.57952880859375,
|
| 442 |
+
"loss": 0.6647,
|
| 443 |
+
"rewards/accuracies": 0.706250011920929,
|
| 444 |
+
"rewards/chosen": -0.07623454183340073,
|
| 445 |
+
"rewards/margins": 0.06355991959571838,
|
| 446 |
+
"rewards/rejected": -0.1397944837808609,
|
| 447 |
+
"step": 280
|
| 448 |
+
},
|
| 449 |
+
{
|
| 450 |
+
"epoch": 0.3,
|
| 451 |
+
"learning_rate": 4.415111107797445e-07,
|
| 452 |
+
"logits/chosen": -1.7934792041778564,
|
| 453 |
+
"logits/rejected": -1.7078644037246704,
|
| 454 |
+
"logps/chosen": -343.297607421875,
|
| 455 |
+
"logps/rejected": -307.45904541015625,
|
| 456 |
+
"loss": 0.6574,
|
| 457 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 458 |
+
"rewards/chosen": -0.09327341616153717,
|
| 459 |
+
"rewards/margins": 0.06554602831602097,
|
| 460 |
+
"rewards/rejected": -0.15881945192813873,
|
| 461 |
+
"step": 290
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.31,
|
| 465 |
+
"learning_rate": 4.3547856320882036e-07,
|
| 466 |
+
"logits/chosen": -1.8048231601715088,
|
| 467 |
+
"logits/rejected": -1.560224175453186,
|
| 468 |
+
"logps/chosen": -344.6669921875,
|
| 469 |
+
"logps/rejected": -290.4841613769531,
|
| 470 |
+
"loss": 0.6499,
|
| 471 |
+
"rewards/accuracies": 0.65625,
|
| 472 |
+
"rewards/chosen": -0.07834906131029129,
|
| 473 |
+
"rewards/margins": 0.09436166286468506,
|
| 474 |
+
"rewards/rejected": -0.17271070182323456,
|
| 475 |
+
"step": 300
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 0.31,
|
| 479 |
+
"eval_logits/chosen": -1.868820071220398,
|
| 480 |
+
"eval_logits/rejected": -1.7357724905014038,
|
| 481 |
+
"eval_logps/chosen": -351.38946533203125,
|
| 482 |
+
"eval_logps/rejected": -318.8117980957031,
|
| 483 |
+
"eval_loss": 0.6586803793907166,
|
| 484 |
+
"eval_rewards/accuracies": 0.6686508059501648,
|
| 485 |
+
"eval_rewards/chosen": -0.08751402050256729,
|
| 486 |
+
"eval_rewards/margins": 0.09379409998655319,
|
| 487 |
+
"eval_rewards/rejected": -0.18130813539028168,
|
| 488 |
+
"eval_runtime": 336.7839,
|
| 489 |
+
"eval_samples_per_second": 5.939,
|
| 490 |
+
"eval_steps_per_second": 0.187,
|
| 491 |
+
"step": 300
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.32,
|
| 495 |
+
"learning_rate": 4.291956282921128e-07,
|
| 496 |
+
"logits/chosen": -1.7621396780014038,
|
| 497 |
+
"logits/rejected": -1.651000738143921,
|
| 498 |
+
"logps/chosen": -348.511474609375,
|
| 499 |
+
"logps/rejected": -330.1929931640625,
|
| 500 |
+
"loss": 0.6514,
|
| 501 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 502 |
+
"rewards/chosen": -0.10528180748224258,
|
| 503 |
+
"rewards/margins": 0.11313805729150772,
|
| 504 |
+
"rewards/rejected": -0.2184198647737503,
|
| 505 |
+
"step": 310
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 0.33,
|
| 509 |
+
"learning_rate": 4.2267078769656105e-07,
|
| 510 |
+
"logits/chosen": -1.8093763589859009,
|
| 511 |
+
"logits/rejected": -1.733646035194397,
|
| 512 |
+
"logps/chosen": -339.96807861328125,
|
| 513 |
+
"logps/rejected": -329.6910705566406,
|
| 514 |
+
"loss": 0.6485,
|
| 515 |
+
"rewards/accuracies": 0.606249988079071,
|
| 516 |
+
"rewards/chosen": -0.10823850333690643,
|
| 517 |
+
"rewards/margins": 0.10142701864242554,
|
| 518 |
+
"rewards/rejected": -0.20966553688049316,
|
| 519 |
+
"step": 320
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"epoch": 0.35,
|
| 523 |
+
"learning_rate": 4.159128496504053e-07,
|
| 524 |
+
"logits/chosen": -1.8124803304672241,
|
| 525 |
+
"logits/rejected": -1.656519889831543,
|
| 526 |
+
"logps/chosen": -338.48504638671875,
|
| 527 |
+
"logps/rejected": -287.4867248535156,
|
| 528 |
+
"loss": 0.6577,
|
| 529 |
+
"rewards/accuracies": 0.643750011920929,
|
| 530 |
+
"rewards/chosen": -0.09331468492746353,
|
| 531 |
+
"rewards/margins": 0.10502304136753082,
|
| 532 |
+
"rewards/rejected": -0.19833770394325256,
|
| 533 |
+
"step": 330
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 0.36,
|
| 537 |
+
"learning_rate": 4.0893093705249207e-07,
|
| 538 |
+
"logits/chosen": -1.7872120141983032,
|
| 539 |
+
"logits/rejected": -1.6502597332000732,
|
| 540 |
+
"logps/chosen": -302.8948669433594,
|
| 541 |
+
"logps/rejected": -278.6997985839844,
|
| 542 |
+
"loss": 0.6597,
|
| 543 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 544 |
+
"rewards/chosen": -0.1376679241657257,
|
| 545 |
+
"rewards/margins": 0.06090687960386276,
|
| 546 |
+
"rewards/rejected": -0.19857481122016907,
|
| 547 |
+
"step": 340
|
| 548 |
+
},
|
| 549 |
+
{
|
| 550 |
+
"epoch": 0.37,
|
| 551 |
+
"learning_rate": 4.0173447515678915e-07,
|
| 552 |
+
"logits/chosen": -1.745719313621521,
|
| 553 |
+
"logits/rejected": -1.7379214763641357,
|
| 554 |
+
"logps/chosen": -341.92218017578125,
|
| 555 |
+
"logps/rejected": -346.43121337890625,
|
| 556 |
+
"loss": 0.6531,
|
| 557 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 558 |
+
"rewards/chosen": -0.1528899073600769,
|
| 559 |
+
"rewards/margins": 0.10737234354019165,
|
| 560 |
+
"rewards/rejected": -0.26026225090026855,
|
| 561 |
+
"step": 350
|
| 562 |
+
},
|
| 563 |
+
{
|
| 564 |
+
"epoch": 0.38,
|
| 565 |
+
"learning_rate": 3.9433317884873665e-07,
|
| 566 |
+
"logits/chosen": -1.896888017654419,
|
| 567 |
+
"logits/rejected": -1.7774041891098022,
|
| 568 |
+
"logps/chosen": -345.6350402832031,
|
| 569 |
+
"logps/rejected": -304.7301025390625,
|
| 570 |
+
"loss": 0.6427,
|
| 571 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 572 |
+
"rewards/chosen": -0.18082933127880096,
|
| 573 |
+
"rewards/margins": 0.10347436368465424,
|
| 574 |
+
"rewards/rejected": -0.2843037247657776,
|
| 575 |
+
"step": 360
|
| 576 |
+
},
|
| 577 |
+
{
|
| 578 |
+
"epoch": 0.39,
|
| 579 |
+
"learning_rate": 3.867370395306068e-07,
|
| 580 |
+
"logits/chosen": -1.7397247552871704,
|
| 581 |
+
"logits/rejected": -1.6811482906341553,
|
| 582 |
+
"logps/chosen": -297.6015625,
|
| 583 |
+
"logps/rejected": -314.14093017578125,
|
| 584 |
+
"loss": 0.6415,
|
| 585 |
+
"rewards/accuracies": 0.675000011920929,
|
| 586 |
+
"rewards/chosen": -0.17070330679416656,
|
| 587 |
+
"rewards/margins": 0.12213484942913055,
|
| 588 |
+
"rewards/rejected": -0.2928381562232971,
|
| 589 |
+
"step": 370
|
| 590 |
+
},
|
| 591 |
+
{
|
| 592 |
+
"epoch": 0.4,
|
| 593 |
+
"learning_rate": 3.78956311633581e-07,
|
| 594 |
+
"logits/chosen": -1.9287761449813843,
|
| 595 |
+
"logits/rejected": -1.7968651056289673,
|
| 596 |
+
"logps/chosen": -369.02142333984375,
|
| 597 |
+
"logps/rejected": -335.91986083984375,
|
| 598 |
+
"loss": 0.6426,
|
| 599 |
+
"rewards/accuracies": 0.668749988079071,
|
| 600 |
+
"rewards/chosen": -0.1528002917766571,
|
| 601 |
+
"rewards/margins": 0.15230461955070496,
|
| 602 |
+
"rewards/rejected": -0.30510494112968445,
|
| 603 |
+
"step": 380
|
| 604 |
+
},
|
| 605 |
+
{
|
| 606 |
+
"epoch": 0.41,
|
| 607 |
+
"learning_rate": 3.7100149877474976e-07,
|
| 608 |
+
"logits/chosen": -1.7732282876968384,
|
| 609 |
+
"logits/rejected": -1.6088358163833618,
|
| 610 |
+
"logps/chosen": -340.55804443359375,
|
| 611 |
+
"logps/rejected": -296.6915283203125,
|
| 612 |
+
"loss": 0.6505,
|
| 613 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 614 |
+
"rewards/chosen": -0.1547761708498001,
|
| 615 |
+
"rewards/margins": 0.11781679093837738,
|
| 616 |
+
"rewards/rejected": -0.2725929617881775,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 0.42,
|
| 621 |
+
"learning_rate": 3.6288333957772234e-07,
|
| 622 |
+
"logits/chosen": -1.9192142486572266,
|
| 623 |
+
"logits/rejected": -1.7736551761627197,
|
| 624 |
+
"logps/chosen": -382.8447265625,
|
| 625 |
+
"logps/rejected": -338.8296813964844,
|
| 626 |
+
"loss": 0.6374,
|
| 627 |
+
"rewards/accuracies": 0.668749988079071,
|
| 628 |
+
"rewards/chosen": -0.18712495267391205,
|
| 629 |
+
"rewards/margins": 0.15186870098114014,
|
| 630 |
+
"rewards/rejected": -0.338993638753891,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 0.42,
|
| 635 |
+
"eval_logits/chosen": -1.848189353942871,
|
| 636 |
+
"eval_logits/rejected": -1.7164283990859985,
|
| 637 |
+
"eval_logps/chosen": -359.8952941894531,
|
| 638 |
+
"eval_logps/rejected": -332.86322021484375,
|
| 639 |
+
"eval_loss": 0.6451287865638733,
|
| 640 |
+
"eval_rewards/accuracies": 0.6746031641960144,
|
| 641 |
+
"eval_rewards/chosen": -0.1725723147392273,
|
| 642 |
+
"eval_rewards/margins": 0.14925029873847961,
|
| 643 |
+
"eval_rewards/rejected": -0.3218226134777069,
|
| 644 |
+
"eval_runtime": 336.4318,
|
| 645 |
+
"eval_samples_per_second": 5.945,
|
| 646 |
+
"eval_steps_per_second": 0.187,
|
| 647 |
+
"step": 400
|
| 648 |
+
},
|
| 649 |
+
{
|
| 650 |
+
"epoch": 0.43,
|
| 651 |
+
"learning_rate": 3.5461279317599025e-07,
|
| 652 |
+
"logits/chosen": -1.9535846710205078,
|
| 653 |
+
"logits/rejected": -1.8399330377578735,
|
| 654 |
+
"logps/chosen": -380.81396484375,
|
| 655 |
+
"logps/rejected": -362.853759765625,
|
| 656 |
+
"loss": 0.6402,
|
| 657 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 658 |
+
"rewards/chosen": -0.1376427710056305,
|
| 659 |
+
"rewards/margins": 0.15401354432106018,
|
| 660 |
+
"rewards/rejected": -0.2916563153266907,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.44,
|
| 665 |
+
"learning_rate": 3.4620102441861144e-07,
|
| 666 |
+
"logits/chosen": -1.7898231744766235,
|
| 667 |
+
"logits/rejected": -1.6419872045516968,
|
| 668 |
+
"logps/chosen": -319.6932678222656,
|
| 669 |
+
"logps/rejected": -280.15960693359375,
|
| 670 |
+
"loss": 0.6387,
|
| 671 |
+
"rewards/accuracies": 0.643750011920929,
|
| 672 |
+
"rewards/chosen": -0.14749577641487122,
|
| 673 |
+
"rewards/margins": 0.12767234444618225,
|
| 674 |
+
"rewards/rejected": -0.2751680910587311,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 0.45,
|
| 679 |
+
"learning_rate": 3.376593887981886e-07,
|
| 680 |
+
"logits/chosen": -1.7821381092071533,
|
| 681 |
+
"logits/rejected": -1.6903321743011475,
|
| 682 |
+
"logps/chosen": -329.77142333984375,
|
| 683 |
+
"logps/rejected": -305.03271484375,
|
| 684 |
+
"loss": 0.6447,
|
| 685 |
+
"rewards/accuracies": 0.59375,
|
| 686 |
+
"rewards/chosen": -0.20932936668395996,
|
| 687 |
+
"rewards/margins": 0.10754366964101791,
|
| 688 |
+
"rewards/rejected": -0.3168730139732361,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 0.46,
|
| 693 |
+
"learning_rate": 3.2899941712148813e-07,
|
| 694 |
+
"logits/chosen": -1.7692524194717407,
|
| 695 |
+
"logits/rejected": -1.703125,
|
| 696 |
+
"logps/chosen": -330.9040222167969,
|
| 697 |
+
"logps/rejected": -317.64447021484375,
|
| 698 |
+
"loss": 0.6428,
|
| 699 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 700 |
+
"rewards/chosen": -0.2360539734363556,
|
| 701 |
+
"rewards/margins": 0.1059914231300354,
|
| 702 |
+
"rewards/rejected": -0.342045396566391,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.47,
|
| 707 |
+
"learning_rate": 3.2023279994339236e-07,
|
| 708 |
+
"logits/chosen": -1.6386339664459229,
|
| 709 |
+
"logits/rejected": -1.5296618938446045,
|
| 710 |
+
"logps/chosen": -331.60919189453125,
|
| 711 |
+
"logps/rejected": -312.84686279296875,
|
| 712 |
+
"loss": 0.6366,
|
| 713 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 714 |
+
"rewards/chosen": -0.18304389715194702,
|
| 715 |
+
"rewards/margins": 0.18606719374656677,
|
| 716 |
+
"rewards/rejected": -0.3691111207008362,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 0.48,
|
| 721 |
+
"learning_rate": 3.1137137178519977e-07,
|
| 722 |
+
"logits/chosen": -1.8083422183990479,
|
| 723 |
+
"logits/rejected": -1.666670799255371,
|
| 724 |
+
"logps/chosen": -337.09991455078125,
|
| 725 |
+
"logps/rejected": -324.5543518066406,
|
| 726 |
+
"loss": 0.6313,
|
| 727 |
+
"rewards/accuracies": 0.581250011920929,
|
| 728 |
+
"rewards/chosen": -0.19549410045146942,
|
| 729 |
+
"rewards/margins": 0.13338689506053925,
|
| 730 |
+
"rewards/rejected": -0.32888099551200867,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 0.49,
|
| 735 |
+
"learning_rate": 3.024270951585776e-07,
|
| 736 |
+
"logits/chosen": -1.8487087488174438,
|
| 737 |
+
"logits/rejected": -1.6575820446014404,
|
| 738 |
+
"logps/chosen": -383.4247131347656,
|
| 739 |
+
"logps/rejected": -340.35845947265625,
|
| 740 |
+
"loss": 0.6264,
|
| 741 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 742 |
+
"rewards/chosen": -0.2236436903476715,
|
| 743 |
+
"rewards/margins": 0.2502152621746063,
|
| 744 |
+
"rewards/rejected": -0.4738590121269226,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 0.5,
|
| 749 |
+
"learning_rate": 2.934120444167326e-07,
|
| 750 |
+
"logits/chosen": -1.6958894729614258,
|
| 751 |
+
"logits/rejected": -1.634203553199768,
|
| 752 |
+
"logps/chosen": -303.4564208984375,
|
| 753 |
+
"logps/rejected": -303.09417724609375,
|
| 754 |
+
"loss": 0.6552,
|
| 755 |
+
"rewards/accuracies": 0.612500011920929,
|
| 756 |
+
"rewards/chosen": -0.2998581528663635,
|
| 757 |
+
"rewards/margins": 0.08801662921905518,
|
| 758 |
+
"rewards/rejected": -0.3878747820854187,
|
| 759 |
+
"step": 480
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 0.51,
|
| 763 |
+
"learning_rate": 2.8433838945460205e-07,
|
| 764 |
+
"logits/chosen": -1.7905118465423584,
|
| 765 |
+
"logits/rejected": -1.6190725564956665,
|
| 766 |
+
"logps/chosen": -353.28582763671875,
|
| 767 |
+
"logps/rejected": -327.52947998046875,
|
| 768 |
+
"loss": 0.6305,
|
| 769 |
+
"rewards/accuracies": 0.643750011920929,
|
| 770 |
+
"rewards/chosen": -0.24126093089580536,
|
| 771 |
+
"rewards/margins": 0.1726113259792328,
|
| 772 |
+
"rewards/rejected": -0.41387230157852173,
|
| 773 |
+
"step": 490
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 0.52,
|
| 777 |
+
"learning_rate": 2.752183792800671e-07,
|
| 778 |
+
"logits/chosen": -1.780226469039917,
|
| 779 |
+
"logits/rejected": -1.6524947881698608,
|
| 780 |
+
"logps/chosen": -338.72564697265625,
|
| 781 |
+
"logps/rejected": -325.1302185058594,
|
| 782 |
+
"loss": 0.6348,
|
| 783 |
+
"rewards/accuracies": 0.65625,
|
| 784 |
+
"rewards/chosen": -0.2332509458065033,
|
| 785 |
+
"rewards/margins": 0.17690421640872955,
|
| 786 |
+
"rewards/rejected": -0.41015520691871643,
|
| 787 |
+
"step": 500
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"epoch": 0.52,
|
| 791 |
+
"eval_logits/chosen": -1.8207694292068481,
|
| 792 |
+
"eval_logits/rejected": -1.6884433031082153,
|
| 793 |
+
"eval_logps/chosen": -369.601318359375,
|
| 794 |
+
"eval_logps/rejected": -346.1808166503906,
|
| 795 |
+
"eval_loss": 0.6377259492874146,
|
| 796 |
+
"eval_rewards/accuracies": 0.6646825671195984,
|
| 797 |
+
"eval_rewards/chosen": -0.26963263750076294,
|
| 798 |
+
"eval_rewards/margins": 0.18536561727523804,
|
| 799 |
+
"eval_rewards/rejected": -0.454998254776001,
|
| 800 |
+
"eval_runtime": 336.8999,
|
| 801 |
+
"eval_samples_per_second": 5.936,
|
| 802 |
+
"eval_steps_per_second": 0.187,
|
| 803 |
+
"step": 500
|
| 804 |
+
},
|
| 805 |
+
{
|
| 806 |
+
"epoch": 0.53,
|
| 807 |
+
"learning_rate": 2.6606432547836753e-07,
|
| 808 |
+
"logits/chosen": -1.7960855960845947,
|
| 809 |
+
"logits/rejected": -1.5604727268218994,
|
| 810 |
+
"logps/chosen": -385.92523193359375,
|
| 811 |
+
"logps/rejected": -332.73236083984375,
|
| 812 |
+
"loss": 0.6292,
|
| 813 |
+
"rewards/accuracies": 0.643750011920929,
|
| 814 |
+
"rewards/chosen": -0.2777213156223297,
|
| 815 |
+
"rewards/margins": 0.21707573533058167,
|
| 816 |
+
"rewards/rejected": -0.4947970509529114,
|
| 817 |
+
"step": 510
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 0.54,
|
| 821 |
+
"learning_rate": 2.5688858559204053e-07,
|
| 822 |
+
"logits/chosen": -1.7386023998260498,
|
| 823 |
+
"logits/rejected": -1.650543212890625,
|
| 824 |
+
"logps/chosen": -324.80194091796875,
|
| 825 |
+
"logps/rejected": -326.7395935058594,
|
| 826 |
+
"loss": 0.6328,
|
| 827 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 828 |
+
"rewards/chosen": -0.22357122600078583,
|
| 829 |
+
"rewards/margins": 0.23022878170013428,
|
| 830 |
+
"rewards/rejected": -0.4538000226020813,
|
| 831 |
+
"step": 520
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"epoch": 0.55,
|
| 835 |
+
"learning_rate": 2.477035464388184e-07,
|
| 836 |
+
"logits/chosen": -1.6688206195831299,
|
| 837 |
+
"logits/rejected": -1.5695548057556152,
|
| 838 |
+
"logps/chosen": -361.03753662109375,
|
| 839 |
+
"logps/rejected": -339.63226318359375,
|
| 840 |
+
"loss": 0.6309,
|
| 841 |
+
"rewards/accuracies": 0.625,
|
| 842 |
+
"rewards/chosen": -0.25370585918426514,
|
| 843 |
+
"rewards/margins": 0.17806772887706757,
|
| 844 |
+
"rewards/rejected": -0.4317736029624939,
|
| 845 |
+
"step": 530
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 0.57,
|
| 849 |
+
"learning_rate": 2.3852160739000706e-07,
|
| 850 |
+
"logits/chosen": -1.827457070350647,
|
| 851 |
+
"logits/rejected": -1.7111104726791382,
|
| 852 |
+
"logps/chosen": -378.85089111328125,
|
| 853 |
+
"logps/rejected": -386.2629699707031,
|
| 854 |
+
"loss": 0.6217,
|
| 855 |
+
"rewards/accuracies": 0.65625,
|
| 856 |
+
"rewards/chosen": -0.24295127391815186,
|
| 857 |
+
"rewards/margins": 0.199497789144516,
|
| 858 |
+
"rewards/rejected": -0.44244909286499023,
|
| 859 |
+
"step": 540
|
| 860 |
+
},
|
| 861 |
+
{
|
| 862 |
+
"epoch": 0.58,
|
| 863 |
+
"learning_rate": 2.2935516363191693e-07,
|
| 864 |
+
"logits/chosen": -1.7694709300994873,
|
| 865 |
+
"logits/rejected": -1.6132080554962158,
|
| 866 |
+
"logps/chosen": -350.4745788574219,
|
| 867 |
+
"logps/rejected": -334.8966369628906,
|
| 868 |
+
"loss": 0.6415,
|
| 869 |
+
"rewards/accuracies": 0.6875,
|
| 870 |
+
"rewards/chosen": -0.30643701553344727,
|
| 871 |
+
"rewards/margins": 0.2043788731098175,
|
| 872 |
+
"rewards/rejected": -0.5108158588409424,
|
| 873 |
+
"step": 550
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 0.59,
|
| 877 |
+
"learning_rate": 2.2021658943294407e-07,
|
| 878 |
+
"logits/chosen": -1.7891242504119873,
|
| 879 |
+
"logits/rejected": -1.6618837118148804,
|
| 880 |
+
"logps/chosen": -348.5799865722656,
|
| 881 |
+
"logps/rejected": -320.85675048828125,
|
| 882 |
+
"loss": 0.6252,
|
| 883 |
+
"rewards/accuracies": 0.675000011920929,
|
| 884 |
+
"rewards/chosen": -0.30380532145500183,
|
| 885 |
+
"rewards/margins": 0.2296813428401947,
|
| 886 |
+
"rewards/rejected": -0.5334866642951965,
|
| 887 |
+
"step": 560
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 0.6,
|
| 891 |
+
"learning_rate": 2.1111822143888928e-07,
|
| 892 |
+
"logits/chosen": -1.8157535791397095,
|
| 893 |
+
"logits/rejected": -1.6647037267684937,
|
| 894 |
+
"logps/chosen": -365.7635192871094,
|
| 895 |
+
"logps/rejected": -348.455810546875,
|
| 896 |
+
"loss": 0.6396,
|
| 897 |
+
"rewards/accuracies": 0.675000011920929,
|
| 898 |
+
"rewards/chosen": -0.3315108120441437,
|
| 899 |
+
"rewards/margins": 0.16533556580543518,
|
| 900 |
+
"rewards/rejected": -0.49684637784957886,
|
| 901 |
+
"step": 570
|
| 902 |
+
},
|
| 903 |
+
{
|
| 904 |
+
"epoch": 0.61,
|
| 905 |
+
"learning_rate": 2.0207234201906545e-07,
|
| 906 |
+
"logits/chosen": -1.6991872787475586,
|
| 907 |
+
"logits/rejected": -1.5607801675796509,
|
| 908 |
+
"logps/chosen": -355.08758544921875,
|
| 909 |
+
"logps/rejected": -318.2015075683594,
|
| 910 |
+
"loss": 0.6263,
|
| 911 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 912 |
+
"rewards/chosen": -0.32228100299835205,
|
| 913 |
+
"rewards/margins": 0.16686378419399261,
|
| 914 |
+
"rewards/rejected": -0.48914486169815063,
|
| 915 |
+
"step": 580
|
| 916 |
+
},
|
| 917 |
+
{
|
| 918 |
+
"epoch": 0.62,
|
| 919 |
+
"learning_rate": 1.9309116268567671e-07,
|
| 920 |
+
"logits/chosen": -1.8247711658477783,
|
| 921 |
+
"logits/rejected": -1.7237234115600586,
|
| 922 |
+
"logps/chosen": -346.73529052734375,
|
| 923 |
+
"logps/rejected": -332.7265625,
|
| 924 |
+
"loss": 0.631,
|
| 925 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 926 |
+
"rewards/chosen": -0.2523882985115051,
|
| 927 |
+
"rewards/margins": 0.11806211620569229,
|
| 928 |
+
"rewards/rejected": -0.37045034766197205,
|
| 929 |
+
"step": 590
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 0.63,
|
| 933 |
+
"learning_rate": 1.8418680760885024e-07,
|
| 934 |
+
"logits/chosen": -1.803122878074646,
|
| 935 |
+
"logits/rejected": -1.6771843433380127,
|
| 936 |
+
"logps/chosen": -412.44671630859375,
|
| 937 |
+
"logps/rejected": -402.94281005859375,
|
| 938 |
+
"loss": 0.6308,
|
| 939 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 940 |
+
"rewards/chosen": -0.22780685126781464,
|
| 941 |
+
"rewards/margins": 0.22766557335853577,
|
| 942 |
+
"rewards/rejected": -0.455472469329834,
|
| 943 |
+
"step": 600
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 0.63,
|
| 947 |
+
"eval_logits/chosen": -1.8269299268722534,
|
| 948 |
+
"eval_logits/rejected": -1.6965351104736328,
|
| 949 |
+
"eval_logps/chosen": -370.46728515625,
|
| 950 |
+
"eval_logps/rejected": -348.8290710449219,
|
| 951 |
+
"eval_loss": 0.6333169937133789,
|
| 952 |
+
"eval_rewards/accuracies": 0.6726190447807312,
|
| 953 |
+
"eval_rewards/chosen": -0.27829232811927795,
|
| 954 |
+
"eval_rewards/margins": 0.2031887322664261,
|
| 955 |
+
"eval_rewards/rejected": -0.48148107528686523,
|
| 956 |
+
"eval_runtime": 333.0266,
|
| 957 |
+
"eval_samples_per_second": 6.006,
|
| 958 |
+
"eval_steps_per_second": 0.189,
|
| 959 |
+
"step": 600
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 0.64,
|
| 963 |
+
"learning_rate": 1.753712972495764e-07,
|
| 964 |
+
"logits/chosen": -1.858438491821289,
|
| 965 |
+
"logits/rejected": -1.6571956872940063,
|
| 966 |
+
"logps/chosen": -356.0987854003906,
|
| 967 |
+
"logps/rejected": -304.0067138671875,
|
| 968 |
+
"loss": 0.6257,
|
| 969 |
+
"rewards/accuracies": 0.699999988079071,
|
| 970 |
+
"rewards/chosen": -0.263310968875885,
|
| 971 |
+
"rewards/margins": 0.19186006486415863,
|
| 972 |
+
"rewards/rejected": -0.45517101883888245,
|
| 973 |
+
"step": 610
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 0.65,
|
| 977 |
+
"learning_rate": 1.666565321326512e-07,
|
| 978 |
+
"logits/chosen": -1.8659627437591553,
|
| 979 |
+
"logits/rejected": -1.709458589553833,
|
| 980 |
+
"logps/chosen": -364.3775634765625,
|
| 981 |
+
"logps/rejected": -322.25396728515625,
|
| 982 |
+
"loss": 0.6239,
|
| 983 |
+
"rewards/accuracies": 0.675000011920929,
|
| 984 |
+
"rewards/chosen": -0.3321610391139984,
|
| 985 |
+
"rewards/margins": 0.1703757643699646,
|
| 986 |
+
"rewards/rejected": -0.5025367736816406,
|
| 987 |
+
"step": 620
|
| 988 |
+
},
|
| 989 |
+
{
|
| 990 |
+
"epoch": 0.66,
|
| 991 |
+
"learning_rate": 1.5805427678152674e-07,
|
| 992 |
+
"logits/chosen": -1.641649603843689,
|
| 993 |
+
"logits/rejected": -1.5464293956756592,
|
| 994 |
+
"logps/chosen": -341.80950927734375,
|
| 995 |
+
"logps/rejected": -339.9558410644531,
|
| 996 |
+
"loss": 0.6251,
|
| 997 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 998 |
+
"rewards/chosen": -0.2863956093788147,
|
| 999 |
+
"rewards/margins": 0.20320038497447968,
|
| 1000 |
+
"rewards/rejected": -0.4895959794521332,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 0.67,
|
| 1005 |
+
"learning_rate": 1.4957614383675767e-07,
|
| 1006 |
+
"logits/chosen": -1.7454828023910522,
|
| 1007 |
+
"logits/rejected": -1.5985407829284668,
|
| 1008 |
+
"logps/chosen": -347.81488037109375,
|
| 1009 |
+
"logps/rejected": -337.49676513671875,
|
| 1010 |
+
"loss": 0.6222,
|
| 1011 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1012 |
+
"rewards/chosen": -0.264575332403183,
|
| 1013 |
+
"rewards/margins": 0.2227667272090912,
|
| 1014 |
+
"rewards/rejected": -0.48734205961227417,
|
| 1015 |
+
"step": 640
|
| 1016 |
+
},
|
| 1017 |
+
{
|
| 1018 |
+
"epoch": 0.68,
|
| 1019 |
+
"learning_rate": 1.4123357837948176e-07,
|
| 1020 |
+
"logits/chosen": -1.9199473857879639,
|
| 1021 |
+
"logits/rejected": -1.715175986289978,
|
| 1022 |
+
"logps/chosen": -401.645263671875,
|
| 1023 |
+
"logps/rejected": -365.5584411621094,
|
| 1024 |
+
"loss": 0.6101,
|
| 1025 |
+
"rewards/accuracies": 0.65625,
|
| 1026 |
+
"rewards/chosen": -0.28458407521247864,
|
| 1027 |
+
"rewards/margins": 0.24697256088256836,
|
| 1028 |
+
"rewards/rejected": -0.5315566062927246,
|
| 1029 |
+
"step": 650
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 0.69,
|
| 1033 |
+
"learning_rate": 1.3303784248109808e-07,
|
| 1034 |
+
"logits/chosen": -1.8878719806671143,
|
| 1035 |
+
"logits/rejected": -1.8138688802719116,
|
| 1036 |
+
"logps/chosen": -379.57672119140625,
|
| 1037 |
+
"logps/rejected": -363.7694091796875,
|
| 1038 |
+
"loss": 0.6202,
|
| 1039 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 1040 |
+
"rewards/chosen": -0.2839195132255554,
|
| 1041 |
+
"rewards/margins": 0.23723697662353516,
|
| 1042 |
+
"rewards/rejected": -0.5211564898490906,
|
| 1043 |
+
"step": 660
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 0.7,
|
| 1047 |
+
"learning_rate": 1.2500000000000005e-07,
|
| 1048 |
+
"logits/chosen": -1.7933881282806396,
|
| 1049 |
+
"logits/rejected": -1.708486557006836,
|
| 1050 |
+
"logps/chosen": -361.3472595214844,
|
| 1051 |
+
"logps/rejected": -356.447509765625,
|
| 1052 |
+
"loss": 0.6204,
|
| 1053 |
+
"rewards/accuracies": 0.637499988079071,
|
| 1054 |
+
"rewards/chosen": -0.2589954733848572,
|
| 1055 |
+
"rewards/margins": 0.21957698464393616,
|
| 1056 |
+
"rewards/rejected": -0.47857245802879333,
|
| 1057 |
+
"step": 670
|
| 1058 |
+
},
|
| 1059 |
+
{
|
| 1060 |
+
"epoch": 0.71,
|
| 1061 |
+
"learning_rate": 1.1713090164588606e-07,
|
| 1062 |
+
"logits/chosen": -1.675574541091919,
|
| 1063 |
+
"logits/rejected": -1.6282098293304443,
|
| 1064 |
+
"logps/chosen": -336.2772521972656,
|
| 1065 |
+
"logps/rejected": -314.0538330078125,
|
| 1066 |
+
"loss": 0.6279,
|
| 1067 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 1068 |
+
"rewards/chosen": -0.24504590034484863,
|
| 1069 |
+
"rewards/margins": 0.09308433532714844,
|
| 1070 |
+
"rewards/rejected": -0.3381302058696747,
|
| 1071 |
+
"step": 680
|
| 1072 |
+
},
|
| 1073 |
+
{
|
| 1074 |
+
"epoch": 0.72,
|
| 1075 |
+
"learning_rate": 1.094411703318115e-07,
|
| 1076 |
+
"logits/chosen": -1.7943480014801025,
|
| 1077 |
+
"logits/rejected": -1.650317907333374,
|
| 1078 |
+
"logps/chosen": -339.38836669921875,
|
| 1079 |
+
"logps/rejected": -322.80755615234375,
|
| 1080 |
+
"loss": 0.6253,
|
| 1081 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1082 |
+
"rewards/chosen": -0.22786399722099304,
|
| 1083 |
+
"rewards/margins": 0.1742563247680664,
|
| 1084 |
+
"rewards/rejected": -0.40212029218673706,
|
| 1085 |
+
"step": 690
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 0.73,
|
| 1089 |
+
"learning_rate": 1.0194118683375502e-07,
|
| 1090 |
+
"logits/chosen": -1.752105951309204,
|
| 1091 |
+
"logits/rejected": -1.5573476552963257,
|
| 1092 |
+
"logps/chosen": -324.02362060546875,
|
| 1093 |
+
"logps/rejected": -319.4602355957031,
|
| 1094 |
+
"loss": 0.62,
|
| 1095 |
+
"rewards/accuracies": 0.643750011920929,
|
| 1096 |
+
"rewards/chosen": -0.25912636518478394,
|
| 1097 |
+
"rewards/margins": 0.1973043978214264,
|
| 1098 |
+
"rewards/rejected": -0.4564308226108551,
|
| 1099 |
+
"step": 700
|
| 1100 |
+
},
|
| 1101 |
+
{
|
| 1102 |
+
"epoch": 0.73,
|
| 1103 |
+
"eval_logits/chosen": -1.8149123191833496,
|
| 1104 |
+
"eval_logits/rejected": -1.6841386556625366,
|
| 1105 |
+
"eval_logps/chosen": -365.8656005859375,
|
| 1106 |
+
"eval_logps/rejected": -345.7306213378906,
|
| 1107 |
+
"eval_loss": 0.63123619556427,
|
| 1108 |
+
"eval_rewards/accuracies": 0.6805555820465088,
|
| 1109 |
+
"eval_rewards/chosen": -0.2322753369808197,
|
| 1110 |
+
"eval_rewards/margins": 0.21822109818458557,
|
| 1111 |
+
"eval_rewards/rejected": -0.45049646496772766,
|
| 1112 |
+
"eval_runtime": 334.8941,
|
| 1113 |
+
"eval_samples_per_second": 5.972,
|
| 1114 |
+
"eval_steps_per_second": 0.188,
|
| 1115 |
+
"step": 700
|
| 1116 |
+
},
|
| 1117 |
+
{
|
| 1118 |
+
"epoch": 0.74,
|
| 1119 |
+
"learning_rate": 9.464107577705886e-08,
|
| 1120 |
+
"logits/chosen": -1.8157602548599243,
|
| 1121 |
+
"logits/rejected": -1.607846975326538,
|
| 1122 |
+
"logps/chosen": -384.46343994140625,
|
| 1123 |
+
"logps/rejected": -319.81805419921875,
|
| 1124 |
+
"loss": 0.6142,
|
| 1125 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 1126 |
+
"rewards/chosen": -0.23041348159313202,
|
| 1127 |
+
"rewards/margins": 0.2244310826063156,
|
| 1128 |
+
"rewards/rejected": -0.45484456419944763,
|
| 1129 |
+
"step": 710
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 0.75,
|
| 1133 |
+
"learning_rate": 8.755069196866013e-08,
|
| 1134 |
+
"logits/chosen": -1.7827469110488892,
|
| 1135 |
+
"logits/rejected": -1.6819953918457031,
|
| 1136 |
+
"logps/chosen": -368.9613952636719,
|
| 1137 |
+
"logps/rejected": -345.7911071777344,
|
| 1138 |
+
"loss": 0.6334,
|
| 1139 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1140 |
+
"rewards/chosen": -0.24999408423900604,
|
| 1141 |
+
"rewards/margins": 0.18186303973197937,
|
| 1142 |
+
"rewards/rejected": -0.431857168674469,
|
| 1143 |
+
"step": 720
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"epoch": 0.76,
|
| 1147 |
+
"learning_rate": 8.067960709356478e-08,
|
| 1148 |
+
"logits/chosen": -1.8376483917236328,
|
| 1149 |
+
"logits/rejected": -1.677075982093811,
|
| 1150 |
+
"logps/chosen": -381.3594665527344,
|
| 1151 |
+
"logps/rejected": -341.1945495605469,
|
| 1152 |
+
"loss": 0.6275,
|
| 1153 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 1154 |
+
"rewards/chosen": -0.25991183519363403,
|
| 1155 |
+
"rewards/margins": 0.2162000834941864,
|
| 1156 |
+
"rewards/rejected": -0.47611188888549805,
|
| 1157 |
+
"step": 730
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 0.77,
|
| 1161 |
+
"learning_rate": 7.403709679352216e-08,
|
| 1162 |
+
"logits/chosen": -1.7825815677642822,
|
| 1163 |
+
"logits/rejected": -1.5635161399841309,
|
| 1164 |
+
"logps/chosen": -405.16839599609375,
|
| 1165 |
+
"logps/rejected": -342.57476806640625,
|
| 1166 |
+
"loss": 0.6172,
|
| 1167 |
+
"rewards/accuracies": 0.731249988079071,
|
| 1168 |
+
"rewards/chosen": -0.2256811410188675,
|
| 1169 |
+
"rewards/margins": 0.23789973556995392,
|
| 1170 |
+
"rewards/rejected": -0.4635809063911438,
|
| 1171 |
+
"step": 740
|
| 1172 |
+
},
|
| 1173 |
+
{
|
| 1174 |
+
"epoch": 0.78,
|
| 1175 |
+
"learning_rate": 6.763212814534483e-08,
|
| 1176 |
+
"logits/chosen": -1.8319247961044312,
|
| 1177 |
+
"logits/rejected": -1.7559360265731812,
|
| 1178 |
+
"logps/chosen": -367.5751647949219,
|
| 1179 |
+
"logps/rejected": -367.4506530761719,
|
| 1180 |
+
"loss": 0.6208,
|
| 1181 |
+
"rewards/accuracies": 0.65625,
|
| 1182 |
+
"rewards/chosen": -0.29129502177238464,
|
| 1183 |
+
"rewards/margins": 0.18480955064296722,
|
| 1184 |
+
"rewards/rejected": -0.4761045575141907,
|
| 1185 |
+
"step": 750
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 0.8,
|
| 1189 |
+
"learning_rate": 6.147334755577596e-08,
|
| 1190 |
+
"logits/chosen": -1.729148268699646,
|
| 1191 |
+
"logits/rejected": -1.6354280710220337,
|
| 1192 |
+
"logps/chosen": -353.6579895019531,
|
| 1193 |
+
"logps/rejected": -341.3516845703125,
|
| 1194 |
+
"loss": 0.6157,
|
| 1195 |
+
"rewards/accuracies": 0.65625,
|
| 1196 |
+
"rewards/chosen": -0.3305814862251282,
|
| 1197 |
+
"rewards/margins": 0.20686399936676025,
|
| 1198 |
+
"rewards/rejected": -0.5374454259872437,
|
| 1199 |
+
"step": 760
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 0.81,
|
| 1203 |
+
"learning_rate": 5.556906908924655e-08,
|
| 1204 |
+
"logits/chosen": -1.774147629737854,
|
| 1205 |
+
"logits/rejected": -1.7218784093856812,
|
| 1206 |
+
"logps/chosen": -382.23968505859375,
|
| 1207 |
+
"logps/rejected": -393.50579833984375,
|
| 1208 |
+
"loss": 0.621,
|
| 1209 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1210 |
+
"rewards/chosen": -0.30208078026771545,
|
| 1211 |
+
"rewards/margins": 0.19531475007534027,
|
| 1212 |
+
"rewards/rejected": -0.49739551544189453,
|
| 1213 |
+
"step": 770
|
| 1214 |
+
},
|
| 1215 |
+
{
|
| 1216 |
+
"epoch": 0.82,
|
| 1217 |
+
"learning_rate": 4.992726324427901e-08,
|
| 1218 |
+
"logits/chosen": -1.6901906728744507,
|
| 1219 |
+
"logits/rejected": -1.58046555519104,
|
| 1220 |
+
"logps/chosen": -335.16583251953125,
|
| 1221 |
+
"logps/rejected": -316.52978515625,
|
| 1222 |
+
"loss": 0.6264,
|
| 1223 |
+
"rewards/accuracies": 0.581250011920929,
|
| 1224 |
+
"rewards/chosen": -0.2825678884983063,
|
| 1225 |
+
"rewards/margins": 0.153324693441391,
|
| 1226 |
+
"rewards/rejected": -0.4358925223350525,
|
| 1227 |
+
"step": 780
|
| 1228 |
+
},
|
| 1229 |
+
{
|
| 1230 |
+
"epoch": 0.83,
|
| 1231 |
+
"learning_rate": 4.4555546193688734e-08,
|
| 1232 |
+
"logits/chosen": -1.795194387435913,
|
| 1233 |
+
"logits/rejected": -1.722459077835083,
|
| 1234 |
+
"logps/chosen": -390.23846435546875,
|
| 1235 |
+
"logps/rejected": -379.72332763671875,
|
| 1236 |
+
"loss": 0.635,
|
| 1237 |
+
"rewards/accuracies": 0.65625,
|
| 1238 |
+
"rewards/chosen": -0.29042816162109375,
|
| 1239 |
+
"rewards/margins": 0.2163534164428711,
|
| 1240 |
+
"rewards/rejected": -0.5067815184593201,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 0.84,
|
| 1245 |
+
"learning_rate": 3.94611695031086e-08,
|
| 1246 |
+
"logits/chosen": -1.743678331375122,
|
| 1247 |
+
"logits/rejected": -1.5813570022583008,
|
| 1248 |
+
"logps/chosen": -426.080078125,
|
| 1249 |
+
"logps/rejected": -366.8641662597656,
|
| 1250 |
+
"loss": 0.6055,
|
| 1251 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1252 |
+
"rewards/chosen": -0.26492840051651,
|
| 1253 |
+
"rewards/margins": 0.23690399527549744,
|
| 1254 |
+
"rewards/rejected": -0.5018323659896851,
|
| 1255 |
+
"step": 800
|
| 1256 |
+
},
|
| 1257 |
+
{
|
| 1258 |
+
"epoch": 0.84,
|
| 1259 |
+
"eval_logits/chosen": -1.809856653213501,
|
| 1260 |
+
"eval_logits/rejected": -1.6792672872543335,
|
| 1261 |
+
"eval_logps/chosen": -371.4098815917969,
|
| 1262 |
+
"eval_logps/rejected": -352.3697204589844,
|
| 1263 |
+
"eval_loss": 0.6287124156951904,
|
| 1264 |
+
"eval_rewards/accuracies": 0.6865079402923584,
|
| 1265 |
+
"eval_rewards/chosen": -0.2877180278301239,
|
| 1266 |
+
"eval_rewards/margins": 0.2291695475578308,
|
| 1267 |
+
"eval_rewards/rejected": -0.5168876051902771,
|
| 1268 |
+
"eval_runtime": 335.4301,
|
| 1269 |
+
"eval_samples_per_second": 5.962,
|
| 1270 |
+
"eval_steps_per_second": 0.188,
|
| 1271 |
+
"step": 800
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"epoch": 0.85,
|
| 1275 |
+
"learning_rate": 3.465101034171603e-08,
|
| 1276 |
+
"logits/chosen": -1.8098558187484741,
|
| 1277 |
+
"logits/rejected": -1.5995447635650635,
|
| 1278 |
+
"logps/chosen": -368.4771423339844,
|
| 1279 |
+
"logps/rejected": -318.16156005859375,
|
| 1280 |
+
"loss": 0.6273,
|
| 1281 |
+
"rewards/accuracies": 0.612500011920929,
|
| 1282 |
+
"rewards/chosen": -0.309670627117157,
|
| 1283 |
+
"rewards/margins": 0.13772638142108917,
|
| 1284 |
+
"rewards/rejected": -0.44739705324172974,
|
| 1285 |
+
"step": 810
|
| 1286 |
+
},
|
| 1287 |
+
{
|
| 1288 |
+
"epoch": 0.86,
|
| 1289 |
+
"learning_rate": 3.013156219837776e-08,
|
| 1290 |
+
"logits/chosen": -1.638256311416626,
|
| 1291 |
+
"logits/rejected": -1.5319178104400635,
|
| 1292 |
+
"logps/chosen": -319.73828125,
|
| 1293 |
+
"logps/rejected": -316.17022705078125,
|
| 1294 |
+
"loss": 0.6382,
|
| 1295 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 1296 |
+
"rewards/chosen": -0.28010720014572144,
|
| 1297 |
+
"rewards/margins": 0.2056259661912918,
|
| 1298 |
+
"rewards/rejected": -0.48573318123817444,
|
| 1299 |
+
"step": 820
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"epoch": 0.87,
|
| 1303 |
+
"learning_rate": 2.5908926115744994e-08,
|
| 1304 |
+
"logits/chosen": -1.6816116571426392,
|
| 1305 |
+
"logits/rejected": -1.6694707870483398,
|
| 1306 |
+
"logps/chosen": -323.09210205078125,
|
| 1307 |
+
"logps/rejected": -325.642822265625,
|
| 1308 |
+
"loss": 0.6157,
|
| 1309 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1310 |
+
"rewards/chosen": -0.2693836987018585,
|
| 1311 |
+
"rewards/margins": 0.23751957714557648,
|
| 1312 |
+
"rewards/rejected": -0.5069032907485962,
|
| 1313 |
+
"step": 830
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 0.88,
|
| 1317 |
+
"learning_rate": 2.19888024541324e-08,
|
| 1318 |
+
"logits/chosen": -1.7210891246795654,
|
| 1319 |
+
"logits/rejected": -1.6208484172821045,
|
| 1320 |
+
"logps/chosen": -391.98187255859375,
|
| 1321 |
+
"logps/rejected": -361.2024841308594,
|
| 1322 |
+
"loss": 0.6279,
|
| 1323 |
+
"rewards/accuracies": 0.574999988079071,
|
| 1324 |
+
"rewards/chosen": -0.33281436562538147,
|
| 1325 |
+
"rewards/margins": 0.1134597659111023,
|
| 1326 |
+
"rewards/rejected": -0.4462741017341614,
|
| 1327 |
+
"step": 840
|
| 1328 |
+
},
|
| 1329 |
+
{
|
| 1330 |
+
"epoch": 0.89,
|
| 1331 |
+
"learning_rate": 1.8376483196299558e-08,
|
| 1332 |
+
"logits/chosen": -1.8490521907806396,
|
| 1333 |
+
"logits/rejected": -1.7024778127670288,
|
| 1334 |
+
"logps/chosen": -389.59765625,
|
| 1335 |
+
"logps/rejected": -336.8246154785156,
|
| 1336 |
+
"loss": 0.6253,
|
| 1337 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1338 |
+
"rewards/chosen": -0.25476616621017456,
|
| 1339 |
+
"rewards/margins": 0.24750569462776184,
|
| 1340 |
+
"rewards/rejected": -0.502271831035614,
|
| 1341 |
+
"step": 850
|
| 1342 |
+
},
|
| 1343 |
+
{
|
| 1344 |
+
"epoch": 0.9,
|
| 1345 |
+
"learning_rate": 1.507684480352292e-08,
|
| 1346 |
+
"logits/chosen": -1.7873809337615967,
|
| 1347 |
+
"logits/rejected": -1.711810827255249,
|
| 1348 |
+
"logps/chosen": -362.58319091796875,
|
| 1349 |
+
"logps/rejected": -348.2912292480469,
|
| 1350 |
+
"loss": 0.6287,
|
| 1351 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 1352 |
+
"rewards/chosen": -0.2954009175300598,
|
| 1353 |
+
"rewards/margins": 0.25710874795913696,
|
| 1354 |
+
"rewards/rejected": -0.5525097846984863,
|
| 1355 |
+
"step": 860
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 0.91,
|
| 1359 |
+
"learning_rate": 1.2094341632602062e-08,
|
| 1360 |
+
"logits/chosen": -1.7503105401992798,
|
| 1361 |
+
"logits/rejected": -1.698735237121582,
|
| 1362 |
+
"logps/chosen": -318.9074401855469,
|
| 1363 |
+
"logps/rejected": -333.00909423828125,
|
| 1364 |
+
"loss": 0.6184,
|
| 1365 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 1366 |
+
"rewards/chosen": -0.25144487619400024,
|
| 1367 |
+
"rewards/margins": 0.24724820256233215,
|
| 1368 |
+
"rewards/rejected": -0.49869304895401,
|
| 1369 |
+
"step": 870
|
| 1370 |
+
},
|
| 1371 |
+
{
|
| 1372 |
+
"epoch": 0.92,
|
| 1373 |
+
"learning_rate": 9.432999922687396e-09,
|
| 1374 |
+
"logits/chosen": -1.721398115158081,
|
| 1375 |
+
"logits/rejected": -1.566706895828247,
|
| 1376 |
+
"logps/chosen": -368.14556884765625,
|
| 1377 |
+
"logps/rejected": -340.03277587890625,
|
| 1378 |
+
"loss": 0.6259,
|
| 1379 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 1380 |
+
"rewards/chosen": -0.3243555426597595,
|
| 1381 |
+
"rewards/margins": 0.1919844150543213,
|
| 1382 |
+
"rewards/rejected": -0.516339898109436,
|
| 1383 |
+
"step": 880
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 0.93,
|
| 1387 |
+
"learning_rate": 7.096412360046544e-09,
|
| 1388 |
+
"logits/chosen": -1.7436186075210571,
|
| 1389 |
+
"logits/rejected": -1.6783809661865234,
|
| 1390 |
+
"logps/chosen": -356.3089294433594,
|
| 1391 |
+
"logps/rejected": -356.8370666503906,
|
| 1392 |
+
"loss": 0.6375,
|
| 1393 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 1394 |
+
"rewards/chosen": -0.3216833472251892,
|
| 1395 |
+
"rewards/margins": 0.15957646071910858,
|
| 1396 |
+
"rewards/rejected": -0.4812597632408142,
|
| 1397 |
+
"step": 890
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 0.94,
|
| 1401 |
+
"learning_rate": 5.087733228106517e-09,
|
| 1402 |
+
"logits/chosen": -1.7303647994995117,
|
| 1403 |
+
"logits/rejected": -1.5875599384307861,
|
| 1404 |
+
"logps/chosen": -385.02294921875,
|
| 1405 |
+
"logps/rejected": -370.8409729003906,
|
| 1406 |
+
"loss": 0.6357,
|
| 1407 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1408 |
+
"rewards/chosen": -0.29238641262054443,
|
| 1409 |
+
"rewards/margins": 0.15909823775291443,
|
| 1410 |
+
"rewards/rejected": -0.451484739780426,
|
| 1411 |
+
"step": 900
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 0.94,
|
| 1415 |
+
"eval_logits/chosen": -1.8041099309921265,
|
| 1416 |
+
"eval_logits/rejected": -1.6730901002883911,
|
| 1417 |
+
"eval_logps/chosen": -373.1398620605469,
|
| 1418 |
+
"eval_logps/rejected": -354.2070617675781,
|
| 1419 |
+
"eval_loss": 0.6285176873207092,
|
| 1420 |
+
"eval_rewards/accuracies": 0.6805555820465088,
|
| 1421 |
+
"eval_rewards/chosen": -0.30501797795295715,
|
| 1422 |
+
"eval_rewards/margins": 0.23024281859397888,
|
| 1423 |
+
"eval_rewards/rejected": -0.535260796546936,
|
| 1424 |
+
"eval_runtime": 333.5177,
|
| 1425 |
+
"eval_samples_per_second": 5.997,
|
| 1426 |
+
"eval_steps_per_second": 0.189,
|
| 1427 |
+
"step": 900
|
| 1428 |
+
},
|
| 1429 |
+
{
|
| 1430 |
+
"epoch": 0.95,
|
| 1431 |
+
"learning_rate": 3.4096741493194193e-09,
|
| 1432 |
+
"logits/chosen": -1.7266037464141846,
|
| 1433 |
+
"logits/rejected": -1.596879482269287,
|
| 1434 |
+
"logps/chosen": -330.7945556640625,
|
| 1435 |
+
"logps/rejected": -328.31414794921875,
|
| 1436 |
+
"loss": 0.6192,
|
| 1437 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1438 |
+
"rewards/chosen": -0.3427024781703949,
|
| 1439 |
+
"rewards/margins": 0.17462757229804993,
|
| 1440 |
+
"rewards/rejected": -0.5173300504684448,
|
| 1441 |
+
"step": 910
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 0.96,
|
| 1445 |
+
"learning_rate": 2.064500424599436e-09,
|
| 1446 |
+
"logits/chosen": -1.789232611656189,
|
| 1447 |
+
"logits/rejected": -1.6246925592422485,
|
| 1448 |
+
"logps/chosen": -370.4477233886719,
|
| 1449 |
+
"logps/rejected": -351.1963806152344,
|
| 1450 |
+
"loss": 0.6148,
|
| 1451 |
+
"rewards/accuracies": 0.731249988079071,
|
| 1452 |
+
"rewards/chosen": -0.2979881167411804,
|
| 1453 |
+
"rewards/margins": 0.2658146917819977,
|
| 1454 |
+
"rewards/rejected": -0.5638028383255005,
|
| 1455 |
+
"step": 920
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 0.97,
|
| 1459 |
+
"learning_rate": 1.0540279752731252e-09,
|
| 1460 |
+
"logits/chosen": -1.856554388999939,
|
| 1461 |
+
"logits/rejected": -1.6683080196380615,
|
| 1462 |
+
"logps/chosen": -354.5975036621094,
|
| 1463 |
+
"logps/rejected": -335.6180725097656,
|
| 1464 |
+
"loss": 0.6047,
|
| 1465 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 1466 |
+
"rewards/chosen": -0.24988925457000732,
|
| 1467 |
+
"rewards/margins": 0.27003243565559387,
|
| 1468 |
+
"rewards/rejected": -0.5199216604232788,
|
| 1469 |
+
"step": 930
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 0.98,
|
| 1473 |
+
"learning_rate": 3.7962089167095645e-10,
|
| 1474 |
+
"logits/chosen": -1.79756760597229,
|
| 1475 |
+
"logits/rejected": -1.610603928565979,
|
| 1476 |
+
"logps/chosen": -409.74462890625,
|
| 1477 |
+
"logps/rejected": -392.2386169433594,
|
| 1478 |
+
"loss": 0.6098,
|
| 1479 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 1480 |
+
"rewards/chosen": -0.29289019107818604,
|
| 1481 |
+
"rewards/margins": 0.28694167733192444,
|
| 1482 |
+
"rewards/rejected": -0.5798318386077881,
|
| 1483 |
+
"step": 940
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"epoch": 0.99,
|
| 1487 |
+
"learning_rate": 4.2189591669322674e-11,
|
| 1488 |
+
"logits/chosen": -1.8283344507217407,
|
| 1489 |
+
"logits/rejected": -1.6751607656478882,
|
| 1490 |
+
"logps/chosen": -369.1002197265625,
|
| 1491 |
+
"logps/rejected": -350.08013916015625,
|
| 1492 |
+
"loss": 0.621,
|
| 1493 |
+
"rewards/accuracies": 0.71875,
|
| 1494 |
+
"rewards/chosen": -0.29409345984458923,
|
| 1495 |
+
"rewards/margins": 0.25061991810798645,
|
| 1496 |
+
"rewards/rejected": -0.5447134375572205,
|
| 1497 |
+
"step": 950
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 1.0,
|
| 1501 |
+
"step": 955,
|
| 1502 |
+
"total_flos": 0.0,
|
| 1503 |
+
"train_loss": 0.6456358959537526,
|
| 1504 |
+
"train_runtime": 17140.5232,
|
| 1505 |
+
"train_samples_per_second": 3.567,
|
| 1506 |
+
"train_steps_per_second": 0.056
|
| 1507 |
+
}
|
| 1508 |
+
],
|
| 1509 |
+
"logging_steps": 10,
|
| 1510 |
+
"max_steps": 955,
|
| 1511 |
+
"num_input_tokens_seen": 0,
|
| 1512 |
+
"num_train_epochs": 1,
|
| 1513 |
+
"save_steps": 100,
|
| 1514 |
+
"total_flos": 0.0,
|
| 1515 |
+
"train_batch_size": 8,
|
| 1516 |
+
"trial_name": null,
|
| 1517 |
+
"trial_params": null
|
| 1518 |
+
}
|