Model save
Browse files- README.md +80 -0
- all_results.json +9 -0
- config.json +1 -1
- generation_config.json +9 -0
- model-00001-of-00004.safetensors +1 -1
- model-00002-of-00004.safetensors +1 -1
- model-00003-of-00004.safetensors +1 -1
- model-00004-of-00004.safetensors +1 -1
- train_results.json +9 -0
- trainer_state.json +1626 -0
- training_args.bin +1 -1
README.md
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
tags:
|
| 4 |
+
- trl
|
| 5 |
+
- dpo
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
model-index:
|
| 8 |
+
- name: zephyr-8b-dpo-full
|
| 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 |
+
# zephyr-8b-dpo-full
|
| 16 |
+
|
| 17 |
+
This model was trained from scratch on an unknown dataset.
|
| 18 |
+
It achieves the following results on the evaluation set:
|
| 19 |
+
- Loss: 0.5131
|
| 20 |
+
- Rewards/chosen: -0.6905
|
| 21 |
+
- Rewards/rejected: -1.5128
|
| 22 |
+
- Rewards/accuracies: 0.7659
|
| 23 |
+
- Rewards/margins: 0.8223
|
| 24 |
+
- Logps/rejected: -432.5991
|
| 25 |
+
- Logps/chosen: -374.1649
|
| 26 |
+
- Logits/rejected: 0.2624
|
| 27 |
+
- Logits/chosen: -0.0136
|
| 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: 4
|
| 48 |
+
- eval_batch_size: 4
|
| 49 |
+
- seed: 42
|
| 50 |
+
- distributed_type: multi-GPU
|
| 51 |
+
- num_devices: 8
|
| 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_ratio: 0.1
|
| 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 |
+
| 0.6766 | 0.1047 | 100 | 0.6765 | 0.0370 | 0.0004 | 0.7103 | 0.0366 | -281.2796 | -301.4119 | -0.6479 | -0.7839 |
|
| 65 |
+
| 0.6043 | 0.2093 | 200 | 0.5948 | -0.4434 | -0.7461 | 0.7262 | 0.3027 | -355.9276 | -349.4549 | -0.2261 | -0.3967 |
|
| 66 |
+
| 0.56 | 0.3140 | 300 | 0.5452 | -0.7062 | -1.2683 | 0.7222 | 0.5621 | -408.1486 | -375.7269 | -0.0802 | -0.2757 |
|
| 67 |
+
| 0.5508 | 0.4186 | 400 | 0.5345 | -0.5684 | -1.2522 | 0.7421 | 0.6837 | -406.5327 | -361.9530 | -0.0946 | -0.2906 |
|
| 68 |
+
| 0.5182 | 0.5233 | 500 | 0.5286 | -0.9677 | -1.6478 | 0.7262 | 0.6801 | -446.0949 | -401.8783 | 0.2622 | 0.0104 |
|
| 69 |
+
| 0.4936 | 0.6279 | 600 | 0.5255 | -0.5650 | -1.3792 | 0.7778 | 0.8143 | -419.2403 | -361.6075 | 0.1520 | -0.1091 |
|
| 70 |
+
| 0.4967 | 0.7326 | 700 | 0.5153 | -0.5682 | -1.3802 | 0.7698 | 0.8121 | -419.3397 | -361.9271 | 0.1898 | -0.0745 |
|
| 71 |
+
| 0.5013 | 0.8373 | 800 | 0.5137 | -0.6747 | -1.4796 | 0.7659 | 0.8049 | -429.2723 | -372.5779 | 0.2384 | -0.0307 |
|
| 72 |
+
| 0.4983 | 0.9419 | 900 | 0.5131 | -0.6905 | -1.5128 | 0.7659 | 0.8223 | -432.5991 | -374.1649 | 0.2624 | -0.0136 |
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
### Framework versions
|
| 76 |
+
|
| 77 |
+
- Transformers 4.45.2
|
| 78 |
+
- Pytorch 2.2.2+rocm5.7
|
| 79 |
+
- Datasets 3.2.0
|
| 80 |
+
- Tokenizers 0.20.3
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9994767137624281,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5504645167845081,
|
| 5 |
+
"train_runtime": 19044.9678,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 3.21,
|
| 8 |
+
"train_steps_per_second": 0.05
|
| 9 |
+
}
|
config.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"_name_or_path": "data/zephyr-8b-sft-full/checkpoint-
|
| 3 |
"architectures": [
|
| 4 |
"LlamaForCausalLM"
|
| 5 |
],
|
|
|
|
| 1 |
{
|
| 2 |
+
"_name_or_path": "data/zephyr-8b-sft-full/checkpoint-1000",
|
| 3 |
"architectures": [
|
| 4 |
"LlamaForCausalLM"
|
| 5 |
],
|
generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 128000,
|
| 4 |
+
"do_sample": true,
|
| 5 |
+
"eos_token_id": 128001,
|
| 6 |
+
"temperature": 0.6,
|
| 7 |
+
"top_p": 0.9,
|
| 8 |
+
"transformers_version": "4.45.2"
|
| 9 |
+
}
|
model-00001-of-00004.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4976698672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:39b832847c8f68222e58779a9d20be821bd29622ecf0e3f1c8c64372d7ed98ef
|
| 3 |
size 4976698672
|
model-00002-of-00004.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4999802720
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f67c4355da2348a95f9ad045da82a74f34466a28ee0fea26baa8ced84234fd66
|
| 3 |
size 4999802720
|
model-00003-of-00004.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4915916176
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7f45dde4c599b65fba248826e93795e0f68c5ee102ffb90d909a2c47f52ed518
|
| 3 |
size 4915916176
|
model-00004-of-00004.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1168138808
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0afba220958a77d77843c0178cd2789a2d9318a78356e2bb886fb867352160b1
|
| 3 |
size 1168138808
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9994767137624281,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.5504645167845081,
|
| 5 |
+
"train_runtime": 19044.9678,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 3.21,
|
| 8 |
+
"train_steps_per_second": 0.05
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1626 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 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": 0.0010465724751439038,
|
| 13 |
+
"grad_norm": 5.451860785513885,
|
| 14 |
+
"learning_rate": 5.208333333333333e-09,
|
| 15 |
+
"logits/chosen": -1.0643162727355957,
|
| 16 |
+
"logits/rejected": -0.861487865447998,
|
| 17 |
+
"logps/chosen": -457.7091064453125,
|
| 18 |
+
"logps/rejected": -371.2662658691406,
|
| 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": 0.010465724751439037,
|
| 28 |
+
"grad_norm": 4.7663255162271385,
|
| 29 |
+
"learning_rate": 5.208333333333333e-08,
|
| 30 |
+
"logits/chosen": -0.7333142161369324,
|
| 31 |
+
"logits/rejected": -0.7134076356887817,
|
| 32 |
+
"logps/chosen": -301.405517578125,
|
| 33 |
+
"logps/rejected": -279.814697265625,
|
| 34 |
+
"loss": 0.6936,
|
| 35 |
+
"rewards/accuracies": 0.4166666567325592,
|
| 36 |
+
"rewards/chosen": -0.0006206975085660815,
|
| 37 |
+
"rewards/margins": -0.0014212429523468018,
|
| 38 |
+
"rewards/rejected": 0.0008005454437807202,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.020931449502878074,
|
| 43 |
+
"grad_norm": 5.112414364557664,
|
| 44 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 45 |
+
"logits/chosen": -0.7887303233146667,
|
| 46 |
+
"logits/rejected": -0.6987150311470032,
|
| 47 |
+
"logps/chosen": -320.1136779785156,
|
| 48 |
+
"logps/rejected": -308.3643798828125,
|
| 49 |
+
"loss": 0.6929,
|
| 50 |
+
"rewards/accuracies": 0.550000011920929,
|
| 51 |
+
"rewards/chosen": 0.0004412093257997185,
|
| 52 |
+
"rewards/margins": 0.001130643649958074,
|
| 53 |
+
"rewards/rejected": -0.000689434411469847,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.03139717425431711,
|
| 58 |
+
"grad_norm": 4.99364758669278,
|
| 59 |
+
"learning_rate": 1.5624999999999999e-07,
|
| 60 |
+
"logits/chosen": -0.8625659942626953,
|
| 61 |
+
"logits/rejected": -0.7310890555381775,
|
| 62 |
+
"logps/chosen": -304.8927307128906,
|
| 63 |
+
"logps/rejected": -282.91070556640625,
|
| 64 |
+
"loss": 0.6931,
|
| 65 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 66 |
+
"rewards/chosen": 0.00022493649157695472,
|
| 67 |
+
"rewards/margins": 0.0005593249225057662,
|
| 68 |
+
"rewards/rejected": -0.00033438848913647234,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.04186289900575615,
|
| 73 |
+
"grad_norm": 5.223938734516328,
|
| 74 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 75 |
+
"logits/chosen": -0.8336359858512878,
|
| 76 |
+
"logits/rejected": -0.7138947248458862,
|
| 77 |
+
"logps/chosen": -346.45550537109375,
|
| 78 |
+
"logps/rejected": -302.9964904785156,
|
| 79 |
+
"loss": 0.6926,
|
| 80 |
+
"rewards/accuracies": 0.5,
|
| 81 |
+
"rewards/chosen": 0.0026509915478527546,
|
| 82 |
+
"rewards/margins": 0.001462576794438064,
|
| 83 |
+
"rewards/rejected": 0.0011884148698300123,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.052328623757195186,
|
| 88 |
+
"grad_norm": 4.722375406184354,
|
| 89 |
+
"learning_rate": 2.604166666666667e-07,
|
| 90 |
+
"logits/chosen": -0.7795863151550293,
|
| 91 |
+
"logits/rejected": -0.6525732278823853,
|
| 92 |
+
"logps/chosen": -257.31060791015625,
|
| 93 |
+
"logps/rejected": -221.70748901367188,
|
| 94 |
+
"loss": 0.6922,
|
| 95 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 96 |
+
"rewards/chosen": 0.0048319087363779545,
|
| 97 |
+
"rewards/margins": 0.0019413300324231386,
|
| 98 |
+
"rewards/rejected": 0.002890578703954816,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.06279434850863422,
|
| 103 |
+
"grad_norm": 5.075388680101497,
|
| 104 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 105 |
+
"logits/chosen": -0.8222079277038574,
|
| 106 |
+
"logits/rejected": -0.7359899282455444,
|
| 107 |
+
"logps/chosen": -345.0970153808594,
|
| 108 |
+
"logps/rejected": -300.33770751953125,
|
| 109 |
+
"loss": 0.6911,
|
| 110 |
+
"rewards/accuracies": 0.612500011920929,
|
| 111 |
+
"rewards/chosen": 0.011030884459614754,
|
| 112 |
+
"rewards/margins": 0.004350438714027405,
|
| 113 |
+
"rewards/rejected": 0.006680445279926062,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.07326007326007326,
|
| 118 |
+
"grad_norm": 5.046679615665253,
|
| 119 |
+
"learning_rate": 3.645833333333333e-07,
|
| 120 |
+
"logits/chosen": -0.6597806811332703,
|
| 121 |
+
"logits/rejected": -0.6361770629882812,
|
| 122 |
+
"logps/chosen": -260.9971618652344,
|
| 123 |
+
"logps/rejected": -286.46697998046875,
|
| 124 |
+
"loss": 0.6894,
|
| 125 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 126 |
+
"rewards/chosen": 0.013305026106536388,
|
| 127 |
+
"rewards/margins": 0.007203704211860895,
|
| 128 |
+
"rewards/rejected": 0.0061013223603367805,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.0837257980115123,
|
| 133 |
+
"grad_norm": 4.97721377364173,
|
| 134 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 135 |
+
"logits/chosen": -0.758665144443512,
|
| 136 |
+
"logits/rejected": -0.598028302192688,
|
| 137 |
+
"logps/chosen": -276.90338134765625,
|
| 138 |
+
"logps/rejected": -261.9596862792969,
|
| 139 |
+
"loss": 0.6862,
|
| 140 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 141 |
+
"rewards/chosen": 0.022657250985503197,
|
| 142 |
+
"rewards/margins": 0.012633631005883217,
|
| 143 |
+
"rewards/rejected": 0.010023623704910278,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.09419152276295134,
|
| 148 |
+
"grad_norm": 4.157558262116518,
|
| 149 |
+
"learning_rate": 4.6874999999999996e-07,
|
| 150 |
+
"logits/chosen": -0.6880342364311218,
|
| 151 |
+
"logits/rejected": -0.6242440342903137,
|
| 152 |
+
"logps/chosen": -298.48590087890625,
|
| 153 |
+
"logps/rejected": -276.1928405761719,
|
| 154 |
+
"loss": 0.684,
|
| 155 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 156 |
+
"rewards/chosen": 0.026639845222234726,
|
| 157 |
+
"rewards/margins": 0.013844798319041729,
|
| 158 |
+
"rewards/rejected": 0.012795047834515572,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.10465724751439037,
|
| 163 |
+
"grad_norm": 4.859970034492583,
|
| 164 |
+
"learning_rate": 4.999732492681437e-07,
|
| 165 |
+
"logits/chosen": -0.7400780916213989,
|
| 166 |
+
"logits/rejected": -0.5942438840866089,
|
| 167 |
+
"logps/chosen": -298.5497131347656,
|
| 168 |
+
"logps/rejected": -291.4864196777344,
|
| 169 |
+
"loss": 0.6766,
|
| 170 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 171 |
+
"rewards/chosen": 0.036215804517269135,
|
| 172 |
+
"rewards/margins": 0.034676693379879,
|
| 173 |
+
"rewards/rejected": 0.001539108925499022,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.10465724751439037,
|
| 178 |
+
"eval_logits/chosen": -0.7839179039001465,
|
| 179 |
+
"eval_logits/rejected": -0.6478830575942993,
|
| 180 |
+
"eval_logps/chosen": -301.41192626953125,
|
| 181 |
+
"eval_logps/rejected": -281.27960205078125,
|
| 182 |
+
"eval_loss": 0.676456093788147,
|
| 183 |
+
"eval_rewards/accuracies": 0.7103174328804016,
|
| 184 |
+
"eval_rewards/chosen": 0.036981064826250076,
|
| 185 |
+
"eval_rewards/margins": 0.03660787642002106,
|
| 186 |
+
"eval_rewards/rejected": 0.0003731876495294273,
|
| 187 |
+
"eval_runtime": 229.5969,
|
| 188 |
+
"eval_samples_per_second": 8.711,
|
| 189 |
+
"eval_steps_per_second": 0.274,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.1151229722658294,
|
| 194 |
+
"grad_norm": 5.0565000677565495,
|
| 195 |
+
"learning_rate": 4.996723692767926e-07,
|
| 196 |
+
"logits/chosen": -0.7310239672660828,
|
| 197 |
+
"logits/rejected": -0.6270194053649902,
|
| 198 |
+
"logps/chosen": -276.677490234375,
|
| 199 |
+
"logps/rejected": -242.0689239501953,
|
| 200 |
+
"loss": 0.6743,
|
| 201 |
+
"rewards/accuracies": 0.675000011920929,
|
| 202 |
+
"rewards/chosen": 0.04216315224766731,
|
| 203 |
+
"rewards/margins": 0.0578165240585804,
|
| 204 |
+
"rewards/rejected": -0.015653371810913086,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.12558869701726844,
|
| 209 |
+
"grad_norm": 5.236113971991537,
|
| 210 |
+
"learning_rate": 4.990375746213598e-07,
|
| 211 |
+
"logits/chosen": -0.7704837322235107,
|
| 212 |
+
"logits/rejected": -0.7006691098213196,
|
| 213 |
+
"logps/chosen": -253.9123992919922,
|
| 214 |
+
"logps/rejected": -279.79742431640625,
|
| 215 |
+
"loss": 0.6605,
|
| 216 |
+
"rewards/accuracies": 0.762499988079071,
|
| 217 |
+
"rewards/chosen": 0.021229851990938187,
|
| 218 |
+
"rewards/margins": 0.06761517375707626,
|
| 219 |
+
"rewards/rejected": -0.04638532176613808,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.1360544217687075,
|
| 224 |
+
"grad_norm": 5.848718385521063,
|
| 225 |
+
"learning_rate": 4.980697142834314e-07,
|
| 226 |
+
"logits/chosen": -0.7273018956184387,
|
| 227 |
+
"logits/rejected": -0.6581666469573975,
|
| 228 |
+
"logps/chosen": -309.36895751953125,
|
| 229 |
+
"logps/rejected": -300.0992126464844,
|
| 230 |
+
"loss": 0.6491,
|
| 231 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 232 |
+
"rewards/chosen": -0.043165743350982666,
|
| 233 |
+
"rewards/margins": 0.11292536556720734,
|
| 234 |
+
"rewards/rejected": -0.1560911238193512,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.14652014652014653,
|
| 239 |
+
"grad_norm": 7.55656728427718,
|
| 240 |
+
"learning_rate": 4.967700826904229e-07,
|
| 241 |
+
"logits/chosen": -0.5881147384643555,
|
| 242 |
+
"logits/rejected": -0.5260018110275269,
|
| 243 |
+
"logps/chosen": -323.1036071777344,
|
| 244 |
+
"logps/rejected": -325.988037109375,
|
| 245 |
+
"loss": 0.6403,
|
| 246 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 247 |
+
"rewards/chosen": -0.07382384687662125,
|
| 248 |
+
"rewards/margins": 0.09867590665817261,
|
| 249 |
+
"rewards/rejected": -0.17249977588653564,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.15698587127158556,
|
| 254 |
+
"grad_norm": 5.993400229281012,
|
| 255 |
+
"learning_rate": 4.951404179843962e-07,
|
| 256 |
+
"logits/chosen": -0.8269667625427246,
|
| 257 |
+
"logits/rejected": -0.6397506594657898,
|
| 258 |
+
"logps/chosen": -338.58087158203125,
|
| 259 |
+
"logps/rejected": -282.53472900390625,
|
| 260 |
+
"loss": 0.6354,
|
| 261 |
+
"rewards/accuracies": 0.75,
|
| 262 |
+
"rewards/chosen": -0.023351237177848816,
|
| 263 |
+
"rewards/margins": 0.14395873248577118,
|
| 264 |
+
"rewards/rejected": -0.1673099845647812,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.1674515960230246,
|
| 269 |
+
"grad_norm": 6.52774166199853,
|
| 270 |
+
"learning_rate": 4.931828996974498e-07,
|
| 271 |
+
"logits/chosen": -0.7941682934761047,
|
| 272 |
+
"logits/rejected": -0.5335707068443298,
|
| 273 |
+
"logps/chosen": -351.42047119140625,
|
| 274 |
+
"logps/rejected": -295.7675476074219,
|
| 275 |
+
"loss": 0.6148,
|
| 276 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 277 |
+
"rewards/chosen": -0.02929142490029335,
|
| 278 |
+
"rewards/margins": 0.2560710310935974,
|
| 279 |
+
"rewards/rejected": -0.28536245226860046,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.17791732077446362,
|
| 284 |
+
"grad_norm": 7.419772604550085,
|
| 285 |
+
"learning_rate": 4.909001458367866e-07,
|
| 286 |
+
"logits/chosen": -0.6290043592453003,
|
| 287 |
+
"logits/rejected": -0.50290447473526,
|
| 288 |
+
"logps/chosen": -297.24652099609375,
|
| 289 |
+
"logps/rejected": -316.2005615234375,
|
| 290 |
+
"loss": 0.6028,
|
| 291 |
+
"rewards/accuracies": 0.737500011920929,
|
| 292 |
+
"rewards/chosen": -0.14729979634284973,
|
| 293 |
+
"rewards/margins": 0.2713707983493805,
|
| 294 |
+
"rewards/rejected": -0.4186705946922302,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.18838304552590268,
|
| 299 |
+
"grad_norm": 8.219451889017629,
|
| 300 |
+
"learning_rate": 4.882952093833627e-07,
|
| 301 |
+
"logits/chosen": -0.695152759552002,
|
| 302 |
+
"logits/rejected": -0.6777328848838806,
|
| 303 |
+
"logps/chosen": -266.3822937011719,
|
| 304 |
+
"logps/rejected": -305.38055419921875,
|
| 305 |
+
"loss": 0.5993,
|
| 306 |
+
"rewards/accuracies": 0.699999988079071,
|
| 307 |
+
"rewards/chosen": -0.1685328185558319,
|
| 308 |
+
"rewards/margins": 0.23227596282958984,
|
| 309 |
+
"rewards/rejected": -0.40080875158309937,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.1988487702773417,
|
| 314 |
+
"grad_norm": 14.401160470357306,
|
| 315 |
+
"learning_rate": 4.853715742087946e-07,
|
| 316 |
+
"logits/chosen": -0.5633553266525269,
|
| 317 |
+
"logits/rejected": -0.4553346037864685,
|
| 318 |
+
"logps/chosen": -299.26141357421875,
|
| 319 |
+
"logps/rejected": -383.54583740234375,
|
| 320 |
+
"loss": 0.5981,
|
| 321 |
+
"rewards/accuracies": 0.800000011920929,
|
| 322 |
+
"rewards/chosen": -0.2618769407272339,
|
| 323 |
+
"rewards/margins": 0.3539055287837982,
|
| 324 |
+
"rewards/rejected": -0.6157824397087097,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.20931449502878074,
|
| 329 |
+
"grad_norm": 9.77874906688374,
|
| 330 |
+
"learning_rate": 4.821331504159906e-07,
|
| 331 |
+
"logits/chosen": -0.6467507481575012,
|
| 332 |
+
"logits/rejected": -0.34418243169784546,
|
| 333 |
+
"logps/chosen": -381.84051513671875,
|
| 334 |
+
"logps/rejected": -324.20941162109375,
|
| 335 |
+
"loss": 0.6043,
|
| 336 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 337 |
+
"rewards/chosen": -0.30156201124191284,
|
| 338 |
+
"rewards/margins": 0.31678324937820435,
|
| 339 |
+
"rewards/rejected": -0.6183452606201172,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.20931449502878074,
|
| 344 |
+
"eval_logits/chosen": -0.3967326581478119,
|
| 345 |
+
"eval_logits/rejected": -0.22607146203517914,
|
| 346 |
+
"eval_logps/chosen": -349.45489501953125,
|
| 347 |
+
"eval_logps/rejected": -355.9275817871094,
|
| 348 |
+
"eval_loss": 0.5947719812393188,
|
| 349 |
+
"eval_rewards/accuracies": 0.726190447807312,
|
| 350 |
+
"eval_rewards/chosen": -0.44344887137413025,
|
| 351 |
+
"eval_rewards/margins": 0.302657812833786,
|
| 352 |
+
"eval_rewards/rejected": -0.7461066246032715,
|
| 353 |
+
"eval_runtime": 229.0804,
|
| 354 |
+
"eval_samples_per_second": 8.731,
|
| 355 |
+
"eval_steps_per_second": 0.275,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.21978021978021978,
|
| 360 |
+
"grad_norm": 10.798066517209035,
|
| 361 |
+
"learning_rate": 4.785842691097342e-07,
|
| 362 |
+
"logits/chosen": -0.3766190707683563,
|
| 363 |
+
"logits/rejected": -0.259346067905426,
|
| 364 |
+
"logps/chosen": -337.4864196777344,
|
| 365 |
+
"logps/rejected": -380.99993896484375,
|
| 366 |
+
"loss": 0.5964,
|
| 367 |
+
"rewards/accuracies": 0.625,
|
| 368 |
+
"rewards/chosen": -0.39307016134262085,
|
| 369 |
+
"rewards/margins": 0.23509669303894043,
|
| 370 |
+
"rewards/rejected": -0.6281668543815613,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.2302459445316588,
|
| 375 |
+
"grad_norm": 8.703797087912555,
|
| 376 |
+
"learning_rate": 4.7472967660421603e-07,
|
| 377 |
+
"logits/chosen": -0.5316035151481628,
|
| 378 |
+
"logits/rejected": -0.3284838795661926,
|
| 379 |
+
"logps/chosen": -331.7840270996094,
|
| 380 |
+
"logps/rejected": -323.8965759277344,
|
| 381 |
+
"loss": 0.5891,
|
| 382 |
+
"rewards/accuracies": 0.699999988079071,
|
| 383 |
+
"rewards/chosen": -0.2757546603679657,
|
| 384 |
+
"rewards/margins": 0.4368220865726471,
|
| 385 |
+
"rewards/rejected": -0.7125767469406128,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.24071166928309787,
|
| 390 |
+
"grad_norm": 10.116764545045047,
|
| 391 |
+
"learning_rate": 4.705745280752585e-07,
|
| 392 |
+
"logits/chosen": -0.4472675323486328,
|
| 393 |
+
"logits/rejected": -0.36385267972946167,
|
| 394 |
+
"logps/chosen": -318.5218200683594,
|
| 395 |
+
"logps/rejected": -355.2652893066406,
|
| 396 |
+
"loss": 0.5789,
|
| 397 |
+
"rewards/accuracies": 0.675000011920929,
|
| 398 |
+
"rewards/chosen": -0.3962463438510895,
|
| 399 |
+
"rewards/margins": 0.3614889681339264,
|
| 400 |
+
"rewards/rejected": -0.7577352523803711,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.25117739403453687,
|
| 405 |
+
"grad_norm": 9.428286192810406,
|
| 406 |
+
"learning_rate": 4.6612438066572555e-07,
|
| 407 |
+
"logits/chosen": -0.3501998782157898,
|
| 408 |
+
"logits/rejected": -0.2727964520454407,
|
| 409 |
+
"logps/chosen": -312.61175537109375,
|
| 410 |
+
"logps/rejected": -358.1639099121094,
|
| 411 |
+
"loss": 0.5635,
|
| 412 |
+
"rewards/accuracies": 0.75,
|
| 413 |
+
"rewards/chosen": -0.3030812740325928,
|
| 414 |
+
"rewards/margins": 0.5171294212341309,
|
| 415 |
+
"rewards/rejected": -0.8202105760574341,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.2616431187859759,
|
| 420 |
+
"grad_norm": 11.893713587854874,
|
| 421 |
+
"learning_rate": 4.6138518605333664e-07,
|
| 422 |
+
"logits/chosen": -0.3740696310997009,
|
| 423 |
+
"logits/rejected": -0.3431631624698639,
|
| 424 |
+
"logps/chosen": -318.5814514160156,
|
| 425 |
+
"logps/rejected": -364.1845703125,
|
| 426 |
+
"loss": 0.5365,
|
| 427 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 428 |
+
"rewards/chosen": -0.29904454946517944,
|
| 429 |
+
"rewards/margins": 0.49876856803894043,
|
| 430 |
+
"rewards/rejected": -0.7978130578994751,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.272108843537415,
|
| 435 |
+
"grad_norm": 10.66317821656739,
|
| 436 |
+
"learning_rate": 4.5636328249082514e-07,
|
| 437 |
+
"logits/chosen": -0.5066936612129211,
|
| 438 |
+
"logits/rejected": -0.2627994120121002,
|
| 439 |
+
"logps/chosen": -326.15496826171875,
|
| 440 |
+
"logps/rejected": -359.80242919921875,
|
| 441 |
+
"loss": 0.5521,
|
| 442 |
+
"rewards/accuracies": 0.699999988079071,
|
| 443 |
+
"rewards/chosen": -0.2969462275505066,
|
| 444 |
+
"rewards/margins": 0.45207586884498596,
|
| 445 |
+
"rewards/rejected": -0.7490221261978149,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.282574568288854,
|
| 450 |
+
"grad_norm": 15.074175115734294,
|
| 451 |
+
"learning_rate": 4.510653863290871e-07,
|
| 452 |
+
"logits/chosen": -0.3418571949005127,
|
| 453 |
+
"logits/rejected": -0.21116261184215546,
|
| 454 |
+
"logps/chosen": -371.8050842285156,
|
| 455 |
+
"logps/rejected": -365.5683288574219,
|
| 456 |
+
"loss": 0.5354,
|
| 457 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 458 |
+
"rewards/chosen": -0.45582056045532227,
|
| 459 |
+
"rewards/margins": 0.5173171758651733,
|
| 460 |
+
"rewards/rejected": -0.9731376767158508,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.29304029304029305,
|
| 465 |
+
"grad_norm": 11.47127961963256,
|
| 466 |
+
"learning_rate": 4.4549858303465737e-07,
|
| 467 |
+
"logits/chosen": -0.45343008637428284,
|
| 468 |
+
"logits/rejected": -0.21792340278625488,
|
| 469 |
+
"logps/chosen": -384.6351013183594,
|
| 470 |
+
"logps/rejected": -404.1702880859375,
|
| 471 |
+
"loss": 0.5688,
|
| 472 |
+
"rewards/accuracies": 0.762499988079071,
|
| 473 |
+
"rewards/chosen": -0.4517960548400879,
|
| 474 |
+
"rewards/margins": 0.690081775188446,
|
| 475 |
+
"rewards/rejected": -1.1418778896331787,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.3035060177917321,
|
| 480 |
+
"grad_norm": 10.944507091993055,
|
| 481 |
+
"learning_rate": 4.396703177135261e-07,
|
| 482 |
+
"logits/chosen": -0.5740801692008972,
|
| 483 |
+
"logits/rejected": -0.3048514723777771,
|
| 484 |
+
"logps/chosen": -362.6668395996094,
|
| 485 |
+
"logps/rejected": -351.4581604003906,
|
| 486 |
+
"loss": 0.5486,
|
| 487 |
+
"rewards/accuracies": 0.75,
|
| 488 |
+
"rewards/chosen": -0.42820802330970764,
|
| 489 |
+
"rewards/margins": 0.5140999555587769,
|
| 490 |
+
"rewards/rejected": -0.9423079490661621,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.3139717425431711,
|
| 495 |
+
"grad_norm": 15.202455626347957,
|
| 496 |
+
"learning_rate": 4.335883851539693e-07,
|
| 497 |
+
"logits/chosen": -0.24740684032440186,
|
| 498 |
+
"logits/rejected": -0.16567502915859222,
|
| 499 |
+
"logps/chosen": -326.53814697265625,
|
| 500 |
+
"logps/rejected": -362.52337646484375,
|
| 501 |
+
"loss": 0.56,
|
| 502 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 503 |
+
"rewards/chosen": -0.613074779510498,
|
| 504 |
+
"rewards/margins": 0.3882750868797302,
|
| 505 |
+
"rewards/rejected": -1.0013500452041626,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.3139717425431711,
|
| 510 |
+
"eval_logits/chosen": -0.27571603655815125,
|
| 511 |
+
"eval_logits/rejected": -0.08016987144947052,
|
| 512 |
+
"eval_logps/chosen": -375.7269287109375,
|
| 513 |
+
"eval_logps/rejected": -408.14862060546875,
|
| 514 |
+
"eval_loss": 0.5452203154563904,
|
| 515 |
+
"eval_rewards/accuracies": 0.7222222089767456,
|
| 516 |
+
"eval_rewards/chosen": -0.706169068813324,
|
| 517 |
+
"eval_rewards/margins": 0.5621480941772461,
|
| 518 |
+
"eval_rewards/rejected": -1.2683171033859253,
|
| 519 |
+
"eval_runtime": 228.6878,
|
| 520 |
+
"eval_samples_per_second": 8.746,
|
| 521 |
+
"eval_steps_per_second": 0.275,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.32443746729461015,
|
| 526 |
+
"grad_norm": 13.811587080938947,
|
| 527 |
+
"learning_rate": 4.272609194017105e-07,
|
| 528 |
+
"logits/chosen": -0.4527352452278137,
|
| 529 |
+
"logits/rejected": -0.062251877039670944,
|
| 530 |
+
"logps/chosen": -355.1481628417969,
|
| 531 |
+
"logps/rejected": -345.7646484375,
|
| 532 |
+
"loss": 0.5359,
|
| 533 |
+
"rewards/accuracies": 0.737500011920929,
|
| 534 |
+
"rewards/chosen": -0.5272838473320007,
|
| 535 |
+
"rewards/margins": 0.645318329334259,
|
| 536 |
+
"rewards/rejected": -1.1726022958755493,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.3349031920460492,
|
| 541 |
+
"grad_norm": 12.089739461632615,
|
| 542 |
+
"learning_rate": 4.2069638288135547e-07,
|
| 543 |
+
"logits/chosen": -0.3022197186946869,
|
| 544 |
+
"logits/rejected": -0.22582903504371643,
|
| 545 |
+
"logps/chosen": -320.36968994140625,
|
| 546 |
+
"logps/rejected": -406.35565185546875,
|
| 547 |
+
"loss": 0.5448,
|
| 548 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 549 |
+
"rewards/chosen": -0.435504674911499,
|
| 550 |
+
"rewards/margins": 0.6115713715553284,
|
| 551 |
+
"rewards/rejected": -1.0470759868621826,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.3453689167974882,
|
| 556 |
+
"grad_norm": 16.48349473214006,
|
| 557 |
+
"learning_rate": 4.139035550786494e-07,
|
| 558 |
+
"logits/chosen": -0.10828112065792084,
|
| 559 |
+
"logits/rejected": 0.048714399337768555,
|
| 560 |
+
"logps/chosen": -354.105712890625,
|
| 561 |
+
"logps/rejected": -383.7474060058594,
|
| 562 |
+
"loss": 0.5535,
|
| 563 |
+
"rewards/accuracies": 0.675000011920929,
|
| 564 |
+
"rewards/chosen": -0.7812685966491699,
|
| 565 |
+
"rewards/margins": 0.4748496413230896,
|
| 566 |
+
"rewards/rejected": -1.2561182975769043,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.35583464154892724,
|
| 571 |
+
"grad_norm": 9.897253624611022,
|
| 572 |
+
"learning_rate": 4.0689152079869306e-07,
|
| 573 |
+
"logits/chosen": -0.14965051412582397,
|
| 574 |
+
"logits/rejected": 0.014736655168235302,
|
| 575 |
+
"logps/chosen": -307.13482666015625,
|
| 576 |
+
"logps/rejected": -319.986083984375,
|
| 577 |
+
"loss": 0.5461,
|
| 578 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 579 |
+
"rewards/chosen": -0.4124748110771179,
|
| 580 |
+
"rewards/margins": 0.4596773087978363,
|
| 581 |
+
"rewards/rejected": -0.8721522092819214,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.3663003663003663,
|
| 586 |
+
"grad_norm": 10.612437127371589,
|
| 587 |
+
"learning_rate": 3.99669658015821e-07,
|
| 588 |
+
"logits/chosen": -0.07019379734992981,
|
| 589 |
+
"logits/rejected": 0.12156897783279419,
|
| 590 |
+
"logps/chosen": -348.1796875,
|
| 591 |
+
"logps/rejected": -370.327880859375,
|
| 592 |
+
"loss": 0.5437,
|
| 593 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 594 |
+
"rewards/chosen": -0.7881641983985901,
|
| 595 |
+
"rewards/margins": 0.43442434072494507,
|
| 596 |
+
"rewards/rejected": -1.2225885391235352,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.37676609105180536,
|
| 601 |
+
"grad_norm": 14.930745956052746,
|
| 602 |
+
"learning_rate": 3.92247625331392e-07,
|
| 603 |
+
"logits/chosen": -0.120775006711483,
|
| 604 |
+
"logits/rejected": -0.008257830515503883,
|
| 605 |
+
"logps/chosen": -383.41534423828125,
|
| 606 |
+
"logps/rejected": -410.1376953125,
|
| 607 |
+
"loss": 0.5203,
|
| 608 |
+
"rewards/accuracies": 0.800000011920929,
|
| 609 |
+
"rewards/chosen": -0.8889461755752563,
|
| 610 |
+
"rewards/margins": 0.48123469948768616,
|
| 611 |
+
"rewards/rejected": -1.3701808452606201,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.3872318158032444,
|
| 616 |
+
"grad_norm": 14.188357046119185,
|
| 617 |
+
"learning_rate": 3.846353490562664e-07,
|
| 618 |
+
"logits/chosen": -0.21186837553977966,
|
| 619 |
+
"logits/rejected": -0.18856562674045563,
|
| 620 |
+
"logps/chosen": -341.7081604003906,
|
| 621 |
+
"logps/rejected": -380.34039306640625,
|
| 622 |
+
"loss": 0.5258,
|
| 623 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 624 |
+
"rewards/chosen": -0.6658584475517273,
|
| 625 |
+
"rewards/margins": 0.510163426399231,
|
| 626 |
+
"rewards/rejected": -1.1760218143463135,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.3976975405546834,
|
| 631 |
+
"grad_norm": 17.153113991725967,
|
| 632 |
+
"learning_rate": 3.768430099352445e-07,
|
| 633 |
+
"logits/chosen": -0.13871899247169495,
|
| 634 |
+
"logits/rejected": -0.02639787271618843,
|
| 635 |
+
"logps/chosen": -330.0691833496094,
|
| 636 |
+
"logps/rejected": -390.55267333984375,
|
| 637 |
+
"loss": 0.5309,
|
| 638 |
+
"rewards/accuracies": 0.75,
|
| 639 |
+
"rewards/chosen": -0.5334495306015015,
|
| 640 |
+
"rewards/margins": 0.5370527505874634,
|
| 641 |
+
"rewards/rejected": -1.0705024003982544,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.40816326530612246,
|
| 646 |
+
"grad_norm": 14.72003311606877,
|
| 647 |
+
"learning_rate": 3.6888102953122304e-07,
|
| 648 |
+
"logits/chosen": -0.16793277859687805,
|
| 649 |
+
"logits/rejected": -0.06103026866912842,
|
| 650 |
+
"logps/chosen": -300.54656982421875,
|
| 651 |
+
"logps/rejected": -357.5650634765625,
|
| 652 |
+
"loss": 0.5411,
|
| 653 |
+
"rewards/accuracies": 0.6875,
|
| 654 |
+
"rewards/chosen": -0.5054274201393127,
|
| 655 |
+
"rewards/margins": 0.5430657863616943,
|
| 656 |
+
"rewards/rejected": -1.0484931468963623,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.4186289900575615,
|
| 661 |
+
"grad_norm": 20.850926855130513,
|
| 662 |
+
"learning_rate": 3.607600562872785e-07,
|
| 663 |
+
"logits/chosen": -0.2158089131116867,
|
| 664 |
+
"logits/rejected": -0.2511315941810608,
|
| 665 |
+
"logps/chosen": -333.9248046875,
|
| 666 |
+
"logps/rejected": -421.63818359375,
|
| 667 |
+
"loss": 0.5508,
|
| 668 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 669 |
+
"rewards/chosen": -0.6180536150932312,
|
| 670 |
+
"rewards/margins": 0.4438760280609131,
|
| 671 |
+
"rewards/rejected": -1.0619295835494995,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.4186289900575615,
|
| 676 |
+
"eval_logits/chosen": -0.29057949781417847,
|
| 677 |
+
"eval_logits/rejected": -0.09457352012395859,
|
| 678 |
+
"eval_logps/chosen": -361.9530334472656,
|
| 679 |
+
"eval_logps/rejected": -406.53271484375,
|
| 680 |
+
"eval_loss": 0.5345055460929871,
|
| 681 |
+
"eval_rewards/accuracies": 0.7420634627342224,
|
| 682 |
+
"eval_rewards/chosen": -0.5684300065040588,
|
| 683 |
+
"eval_rewards/margins": 0.6837278604507446,
|
| 684 |
+
"eval_rewards/rejected": -1.2521578073501587,
|
| 685 |
+
"eval_runtime": 229.6469,
|
| 686 |
+
"eval_samples_per_second": 8.709,
|
| 687 |
+
"eval_steps_per_second": 0.274,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.4290947148090005,
|
| 692 |
+
"grad_norm": 10.298070998489017,
|
| 693 |
+
"learning_rate": 3.5249095128531856e-07,
|
| 694 |
+
"logits/chosen": -0.1405165195465088,
|
| 695 |
+
"logits/rejected": 0.06610298156738281,
|
| 696 |
+
"logps/chosen": -367.21270751953125,
|
| 697 |
+
"logps/rejected": -396.1885681152344,
|
| 698 |
+
"loss": 0.5457,
|
| 699 |
+
"rewards/accuracies": 0.762499988079071,
|
| 700 |
+
"rewards/chosen": -0.6680795550346375,
|
| 701 |
+
"rewards/margins": 0.6149377822875977,
|
| 702 |
+
"rewards/rejected": -1.2830173969268799,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.43956043956043955,
|
| 707 |
+
"grad_norm": 11.725780840224937,
|
| 708 |
+
"learning_rate": 3.4408477372034736e-07,
|
| 709 |
+
"logits/chosen": -0.3442925810813904,
|
| 710 |
+
"logits/rejected": -0.0029064356349408627,
|
| 711 |
+
"logps/chosen": -413.73199462890625,
|
| 712 |
+
"logps/rejected": -388.2182312011719,
|
| 713 |
+
"loss": 0.53,
|
| 714 |
+
"rewards/accuracies": 0.675000011920929,
|
| 715 |
+
"rewards/chosen": -0.6328467130661011,
|
| 716 |
+
"rewards/margins": 0.5191168189048767,
|
| 717 |
+
"rewards/rejected": -1.151963472366333,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.4500261643118786,
|
| 722 |
+
"grad_norm": 13.27922986778142,
|
| 723 |
+
"learning_rate": 3.3555276610977276e-07,
|
| 724 |
+
"logits/chosen": -0.06918157637119293,
|
| 725 |
+
"logits/rejected": 0.0864952802658081,
|
| 726 |
+
"logps/chosen": -294.5364074707031,
|
| 727 |
+
"logps/rejected": -392.9295654296875,
|
| 728 |
+
"loss": 0.5241,
|
| 729 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 730 |
+
"rewards/chosen": -0.6482856273651123,
|
| 731 |
+
"rewards/margins": 0.7281777262687683,
|
| 732 |
+
"rewards/rejected": -1.3764632940292358,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.4604918890633176,
|
| 737 |
+
"grad_norm": 12.766468960268535,
|
| 738 |
+
"learning_rate": 3.269063392575352e-07,
|
| 739 |
+
"logits/chosen": 0.02183201164007187,
|
| 740 |
+
"logits/rejected": 0.08124883472919464,
|
| 741 |
+
"logps/chosen": -320.3900146484375,
|
| 742 |
+
"logps/rejected": -374.9771728515625,
|
| 743 |
+
"loss": 0.5365,
|
| 744 |
+
"rewards/accuracies": 0.699999988079071,
|
| 745 |
+
"rewards/chosen": -0.6313291788101196,
|
| 746 |
+
"rewards/margins": 0.6131441593170166,
|
| 747 |
+
"rewards/rejected": -1.2444733381271362,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.47095761381475665,
|
| 752 |
+
"grad_norm": 11.245463417386828,
|
| 753 |
+
"learning_rate": 3.1815705699316964e-07,
|
| 754 |
+
"logits/chosen": -0.10290797799825668,
|
| 755 |
+
"logits/rejected": 0.08914367109537125,
|
| 756 |
+
"logps/chosen": -369.95697021484375,
|
| 757 |
+
"logps/rejected": -416.80853271484375,
|
| 758 |
+
"loss": 0.5481,
|
| 759 |
+
"rewards/accuracies": 0.75,
|
| 760 |
+
"rewards/chosen": -0.7395186424255371,
|
| 761 |
+
"rewards/margins": 0.693315863609314,
|
| 762 |
+
"rewards/rejected": -1.432834506034851,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.48142333856619574,
|
| 767 |
+
"grad_norm": 12.061083908845202,
|
| 768 |
+
"learning_rate": 3.0931662070620794e-07,
|
| 769 |
+
"logits/chosen": -0.15375325083732605,
|
| 770 |
+
"logits/rejected": 0.12264730781316757,
|
| 771 |
+
"logps/chosen": -304.6129455566406,
|
| 772 |
+
"logps/rejected": -363.7560729980469,
|
| 773 |
+
"loss": 0.5406,
|
| 774 |
+
"rewards/accuracies": 0.8125,
|
| 775 |
+
"rewards/chosen": -0.6241869926452637,
|
| 776 |
+
"rewards/margins": 0.7929509878158569,
|
| 777 |
+
"rewards/rejected": -1.4171379804611206,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.49188906331763477,
|
| 782 |
+
"grad_norm": 14.419339908169533,
|
| 783 |
+
"learning_rate": 3.003968536966078e-07,
|
| 784 |
+
"logits/chosen": -0.03872384503483772,
|
| 785 |
+
"logits/rejected": 0.09880894422531128,
|
| 786 |
+
"logps/chosen": -359.7276916503906,
|
| 787 |
+
"logps/rejected": -385.1877746582031,
|
| 788 |
+
"loss": 0.5015,
|
| 789 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 790 |
+
"rewards/chosen": -0.6624265909194946,
|
| 791 |
+
"rewards/margins": 0.4815330505371094,
|
| 792 |
+
"rewards/rejected": -1.1439597606658936,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.5023547880690737,
|
| 797 |
+
"grad_norm": 17.404394779820596,
|
| 798 |
+
"learning_rate": 2.9140968536213693e-07,
|
| 799 |
+
"logits/chosen": -0.11226484924554825,
|
| 800 |
+
"logits/rejected": 0.18643446266651154,
|
| 801 |
+
"logps/chosen": -360.540771484375,
|
| 802 |
+
"logps/rejected": -442.9476013183594,
|
| 803 |
+
"loss": 0.5182,
|
| 804 |
+
"rewards/accuracies": 0.8125,
|
| 805 |
+
"rewards/chosen": -0.6307897567749023,
|
| 806 |
+
"rewards/margins": 0.9138120412826538,
|
| 807 |
+
"rewards/rejected": -1.5446019172668457,
|
| 808 |
+
"step": 480
|
| 809 |
+
},
|
| 810 |
+
{
|
| 811 |
+
"epoch": 0.5128205128205128,
|
| 812 |
+
"grad_norm": 12.356577288104344,
|
| 813 |
+
"learning_rate": 2.823671352438608e-07,
|
| 814 |
+
"logits/chosen": -0.19844231009483337,
|
| 815 |
+
"logits/rejected": 0.178545281291008,
|
| 816 |
+
"logps/chosen": -383.7408752441406,
|
| 817 |
+
"logps/rejected": -408.270263671875,
|
| 818 |
+
"loss": 0.5192,
|
| 819 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 820 |
+
"rewards/chosen": -0.5568854212760925,
|
| 821 |
+
"rewards/margins": 0.7981684803962708,
|
| 822 |
+
"rewards/rejected": -1.3550540208816528,
|
| 823 |
+
"step": 490
|
| 824 |
+
},
|
| 825 |
+
{
|
| 826 |
+
"epoch": 0.5232862375719518,
|
| 827 |
+
"grad_norm": 11.064578503088146,
|
| 828 |
+
"learning_rate": 2.73281296951072e-07,
|
| 829 |
+
"logits/chosen": -0.09013144671916962,
|
| 830 |
+
"logits/rejected": 0.18840348720550537,
|
| 831 |
+
"logps/chosen": -322.2913513183594,
|
| 832 |
+
"logps/rejected": -380.41986083984375,
|
| 833 |
+
"loss": 0.5182,
|
| 834 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 835 |
+
"rewards/chosen": -0.6678560972213745,
|
| 836 |
+
"rewards/margins": 0.7828424572944641,
|
| 837 |
+
"rewards/rejected": -1.4506986141204834,
|
| 838 |
+
"step": 500
|
| 839 |
+
},
|
| 840 |
+
{
|
| 841 |
+
"epoch": 0.5232862375719518,
|
| 842 |
+
"eval_logits/chosen": 0.010395529679954052,
|
| 843 |
+
"eval_logits/rejected": 0.26224514842033386,
|
| 844 |
+
"eval_logps/chosen": -401.8782653808594,
|
| 845 |
+
"eval_logps/rejected": -446.0949401855469,
|
| 846 |
+
"eval_loss": 0.5285578370094299,
|
| 847 |
+
"eval_rewards/accuracies": 0.726190447807312,
|
| 848 |
+
"eval_rewards/chosen": -0.967681884765625,
|
| 849 |
+
"eval_rewards/margins": 0.6800985932350159,
|
| 850 |
+
"eval_rewards/rejected": -1.6477805376052856,
|
| 851 |
+
"eval_runtime": 228.783,
|
| 852 |
+
"eval_samples_per_second": 8.742,
|
| 853 |
+
"eval_steps_per_second": 0.275,
|
| 854 |
+
"step": 500
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"epoch": 0.533751962323391,
|
| 858 |
+
"grad_norm": 15.367822240197881,
|
| 859 |
+
"learning_rate": 2.641643219871597e-07,
|
| 860 |
+
"logits/chosen": 0.03556225821375847,
|
| 861 |
+
"logits/rejected": 0.202899768948555,
|
| 862 |
+
"logps/chosen": -343.4957580566406,
|
| 863 |
+
"logps/rejected": -412.44451904296875,
|
| 864 |
+
"loss": 0.5094,
|
| 865 |
+
"rewards/accuracies": 0.762499988079071,
|
| 866 |
+
"rewards/chosen": -0.7577770948410034,
|
| 867 |
+
"rewards/margins": 0.8080625534057617,
|
| 868 |
+
"rewards/rejected": -1.5658397674560547,
|
| 869 |
+
"step": 510
|
| 870 |
+
},
|
| 871 |
+
{
|
| 872 |
+
"epoch": 0.54421768707483,
|
| 873 |
+
"grad_norm": 13.43430915472883,
|
| 874 |
+
"learning_rate": 2.550284034980507e-07,
|
| 875 |
+
"logits/chosen": 0.07642324268817902,
|
| 876 |
+
"logits/rejected": 0.16387151181697845,
|
| 877 |
+
"logps/chosen": -363.08551025390625,
|
| 878 |
+
"logps/rejected": -401.5746154785156,
|
| 879 |
+
"loss": 0.5326,
|
| 880 |
+
"rewards/accuracies": 0.675000011920929,
|
| 881 |
+
"rewards/chosen": -0.887750506401062,
|
| 882 |
+
"rewards/margins": 0.5795798301696777,
|
| 883 |
+
"rewards/rejected": -1.4673302173614502,
|
| 884 |
+
"step": 520
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"epoch": 0.554683411826269,
|
| 888 |
+
"grad_norm": 16.90441101569984,
|
| 889 |
+
"learning_rate": 2.4588575996495794e-07,
|
| 890 |
+
"logits/chosen": -0.1090664491057396,
|
| 891 |
+
"logits/rejected": 0.1268138587474823,
|
| 892 |
+
"logps/chosen": -387.29510498046875,
|
| 893 |
+
"logps/rejected": -459.40264892578125,
|
| 894 |
+
"loss": 0.5107,
|
| 895 |
+
"rewards/accuracies": 0.8125,
|
| 896 |
+
"rewards/chosen": -0.6459606885910034,
|
| 897 |
+
"rewards/margins": 0.941811203956604,
|
| 898 |
+
"rewards/rejected": -1.5877716541290283,
|
| 899 |
+
"step": 530
|
| 900 |
+
},
|
| 901 |
+
{
|
| 902 |
+
"epoch": 0.565149136577708,
|
| 903 |
+
"grad_norm": 12.507368084637326,
|
| 904 |
+
"learning_rate": 2.367486188632446e-07,
|
| 905 |
+
"logits/chosen": -0.039569877088069916,
|
| 906 |
+
"logits/rejected": 0.09773418307304382,
|
| 907 |
+
"logps/chosen": -366.95806884765625,
|
| 908 |
+
"logps/rejected": -394.683349609375,
|
| 909 |
+
"loss": 0.5247,
|
| 910 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 911 |
+
"rewards/chosen": -0.6594400405883789,
|
| 912 |
+
"rewards/margins": 0.6299028992652893,
|
| 913 |
+
"rewards/rejected": -1.289342999458313,
|
| 914 |
+
"step": 540
|
| 915 |
+
},
|
| 916 |
+
{
|
| 917 |
+
"epoch": 0.5756148613291471,
|
| 918 |
+
"grad_norm": 13.366833820050662,
|
| 919 |
+
"learning_rate": 2.276292003092593e-07,
|
| 920 |
+
"logits/chosen": -0.1256514936685562,
|
| 921 |
+
"logits/rejected": 0.14441490173339844,
|
| 922 |
+
"logps/chosen": -369.62310791015625,
|
| 923 |
+
"logps/rejected": -368.75726318359375,
|
| 924 |
+
"loss": 0.5278,
|
| 925 |
+
"rewards/accuracies": 0.699999988079071,
|
| 926 |
+
"rewards/chosen": -0.6649243235588074,
|
| 927 |
+
"rewards/margins": 0.5402246117591858,
|
| 928 |
+
"rewards/rejected": -1.2051489353179932,
|
| 929 |
+
"step": 550
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 0.5860805860805861,
|
| 933 |
+
"grad_norm": 15.734482090071854,
|
| 934 |
+
"learning_rate": 2.185397007170141e-07,
|
| 935 |
+
"logits/chosen": 0.1283525973558426,
|
| 936 |
+
"logits/rejected": 0.24960267543792725,
|
| 937 |
+
"logps/chosen": -285.6956787109375,
|
| 938 |
+
"logps/rejected": -360.90509033203125,
|
| 939 |
+
"loss": 0.532,
|
| 940 |
+
"rewards/accuracies": 0.737500011920929,
|
| 941 |
+
"rewards/chosen": -0.4988901615142822,
|
| 942 |
+
"rewards/margins": 0.7280929684638977,
|
| 943 |
+
"rewards/rejected": -1.2269830703735352,
|
| 944 |
+
"step": 560
|
| 945 |
+
},
|
| 946 |
+
{
|
| 947 |
+
"epoch": 0.5965463108320251,
|
| 948 |
+
"grad_norm": 12.793432900556661,
|
| 949 |
+
"learning_rate": 2.094922764865619e-07,
|
| 950 |
+
"logits/chosen": 0.18585723638534546,
|
| 951 |
+
"logits/rejected": 0.2675357460975647,
|
| 952 |
+
"logps/chosen": -364.28521728515625,
|
| 953 |
+
"logps/rejected": -457.3734436035156,
|
| 954 |
+
"loss": 0.5118,
|
| 955 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 956 |
+
"rewards/chosen": -1.0480437278747559,
|
| 957 |
+
"rewards/margins": 0.5522451996803284,
|
| 958 |
+
"rewards/rejected": -1.600288987159729,
|
| 959 |
+
"step": 570
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 0.6070120355834642,
|
| 963 |
+
"grad_norm": 14.54170281598928,
|
| 964 |
+
"learning_rate": 2.0049902774588797e-07,
|
| 965 |
+
"logits/chosen": 0.08888915926218033,
|
| 966 |
+
"logits/rejected": 0.3060843050479889,
|
| 967 |
+
"logps/chosen": -333.07373046875,
|
| 968 |
+
"logps/rejected": -380.82818603515625,
|
| 969 |
+
"loss": 0.5291,
|
| 970 |
+
"rewards/accuracies": 0.6875,
|
| 971 |
+
"rewards/chosen": -0.7626341581344604,
|
| 972 |
+
"rewards/margins": 0.7598274946212769,
|
| 973 |
+
"rewards/rejected": -1.5224617719650269,
|
| 974 |
+
"step": 580
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 0.6174777603349032,
|
| 978 |
+
"grad_norm": 14.79212886128343,
|
| 979 |
+
"learning_rate": 1.9157198216806238e-07,
|
| 980 |
+
"logits/chosen": -0.1875981092453003,
|
| 981 |
+
"logits/rejected": 0.05693904310464859,
|
| 982 |
+
"logps/chosen": -382.46868896484375,
|
| 983 |
+
"logps/rejected": -439.0375061035156,
|
| 984 |
+
"loss": 0.4968,
|
| 985 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 986 |
+
"rewards/chosen": -0.6531257629394531,
|
| 987 |
+
"rewards/margins": 0.8373476266860962,
|
| 988 |
+
"rewards/rejected": -1.4904735088348389,
|
| 989 |
+
"step": 590
|
| 990 |
+
},
|
| 991 |
+
{
|
| 992 |
+
"epoch": 0.6279434850863422,
|
| 993 |
+
"grad_norm": 13.087506840381133,
|
| 994 |
+
"learning_rate": 1.8272307888529274e-07,
|
| 995 |
+
"logits/chosen": 0.026107722893357277,
|
| 996 |
+
"logits/rejected": 0.29632943868637085,
|
| 997 |
+
"logps/chosen": -438.74761962890625,
|
| 998 |
+
"logps/rejected": -426.4642639160156,
|
| 999 |
+
"loss": 0.4936,
|
| 1000 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1001 |
+
"rewards/chosen": -0.7559309601783752,
|
| 1002 |
+
"rewards/margins": 0.568900465965271,
|
| 1003 |
+
"rewards/rejected": -1.324831247329712,
|
| 1004 |
+
"step": 600
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"epoch": 0.6279434850863422,
|
| 1008 |
+
"eval_logits/chosen": -0.10911048203706741,
|
| 1009 |
+
"eval_logits/rejected": 0.15195031464099884,
|
| 1010 |
+
"eval_logps/chosen": -361.6075439453125,
|
| 1011 |
+
"eval_logps/rejected": -419.2403259277344,
|
| 1012 |
+
"eval_loss": 0.5254898071289062,
|
| 1013 |
+
"eval_rewards/accuracies": 0.7777777910232544,
|
| 1014 |
+
"eval_rewards/chosen": -0.5649750232696533,
|
| 1015 |
+
"eval_rewards/margins": 0.8142591118812561,
|
| 1016 |
+
"eval_rewards/rejected": -1.3792341947555542,
|
| 1017 |
+
"eval_runtime": 228.5101,
|
| 1018 |
+
"eval_samples_per_second": 8.752,
|
| 1019 |
+
"eval_steps_per_second": 0.276,
|
| 1020 |
+
"step": 600
|
| 1021 |
+
},
|
| 1022 |
+
{
|
| 1023 |
+
"epoch": 0.6384092098377813,
|
| 1024 |
+
"grad_norm": 16.705575529622894,
|
| 1025 |
+
"learning_rate": 1.7396415252139288e-07,
|
| 1026 |
+
"logits/chosen": -0.07063977420330048,
|
| 1027 |
+
"logits/rejected": 0.1474047303199768,
|
| 1028 |
+
"logps/chosen": -378.7034912109375,
|
| 1029 |
+
"logps/rejected": -424.65838623046875,
|
| 1030 |
+
"loss": 0.5334,
|
| 1031 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1032 |
+
"rewards/chosen": -0.5486348271369934,
|
| 1033 |
+
"rewards/margins": 0.8419520258903503,
|
| 1034 |
+
"rewards/rejected": -1.3905869722366333,
|
| 1035 |
+
"step": 610
|
| 1036 |
+
},
|
| 1037 |
+
{
|
| 1038 |
+
"epoch": 0.6488749345892203,
|
| 1039 |
+
"grad_norm": 16.680744128054243,
|
| 1040 |
+
"learning_rate": 1.6530691736402316e-07,
|
| 1041 |
+
"logits/chosen": -0.06789754331111908,
|
| 1042 |
+
"logits/rejected": 0.1724042445421219,
|
| 1043 |
+
"logps/chosen": -380.02203369140625,
|
| 1044 |
+
"logps/rejected": -394.5411071777344,
|
| 1045 |
+
"loss": 0.4879,
|
| 1046 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1047 |
+
"rewards/chosen": -0.6499656438827515,
|
| 1048 |
+
"rewards/margins": 0.8682801127433777,
|
| 1049 |
+
"rewards/rejected": -1.5182459354400635,
|
| 1050 |
+
"step": 620
|
| 1051 |
+
},
|
| 1052 |
+
{
|
| 1053 |
+
"epoch": 0.6593406593406593,
|
| 1054 |
+
"grad_norm": 12.347293383994431,
|
| 1055 |
+
"learning_rate": 1.5676295169786864e-07,
|
| 1056 |
+
"logits/chosen": -0.008408618159592152,
|
| 1057 |
+
"logits/rejected": 0.24951598048210144,
|
| 1058 |
+
"logps/chosen": -339.1716003417969,
|
| 1059 |
+
"logps/rejected": -378.724609375,
|
| 1060 |
+
"loss": 0.4863,
|
| 1061 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1062 |
+
"rewards/chosen": -0.5663776397705078,
|
| 1063 |
+
"rewards/margins": 0.7724849581718445,
|
| 1064 |
+
"rewards/rejected": -1.338862657546997,
|
| 1065 |
+
"step": 630
|
| 1066 |
+
},
|
| 1067 |
+
{
|
| 1068 |
+
"epoch": 0.6698063840920984,
|
| 1069 |
+
"grad_norm": 16.072115808650597,
|
| 1070 |
+
"learning_rate": 1.483436823197092e-07,
|
| 1071 |
+
"logits/chosen": 0.09198600053787231,
|
| 1072 |
+
"logits/rejected": 0.36397668719291687,
|
| 1073 |
+
"logps/chosen": -319.0302734375,
|
| 1074 |
+
"logps/rejected": -378.3144836425781,
|
| 1075 |
+
"loss": 0.508,
|
| 1076 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1077 |
+
"rewards/chosen": -0.670558750629425,
|
| 1078 |
+
"rewards/margins": 0.8327127695083618,
|
| 1079 |
+
"rewards/rejected": -1.5032716989517212,
|
| 1080 |
+
"step": 640
|
| 1081 |
+
},
|
| 1082 |
+
{
|
| 1083 |
+
"epoch": 0.6802721088435374,
|
| 1084 |
+
"grad_norm": 13.081700015736937,
|
| 1085 |
+
"learning_rate": 1.4006036925609243e-07,
|
| 1086 |
+
"logits/chosen": 0.07456495612859726,
|
| 1087 |
+
"logits/rejected": 0.13473407924175262,
|
| 1088 |
+
"logps/chosen": -359.57244873046875,
|
| 1089 |
+
"logps/rejected": -442.2086486816406,
|
| 1090 |
+
"loss": 0.5214,
|
| 1091 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1092 |
+
"rewards/chosen": -0.6933525800704956,
|
| 1093 |
+
"rewards/margins": 0.744998574256897,
|
| 1094 |
+
"rewards/rejected": -1.4383512735366821,
|
| 1095 |
+
"step": 650
|
| 1096 |
+
},
|
| 1097 |
+
{
|
| 1098 |
+
"epoch": 0.6907378335949764,
|
| 1099 |
+
"grad_norm": 12.02667717260737,
|
| 1100 |
+
"learning_rate": 1.319240907040458e-07,
|
| 1101 |
+
"logits/chosen": -0.20533649623394012,
|
| 1102 |
+
"logits/rejected": -0.07553025335073471,
|
| 1103 |
+
"logps/chosen": -349.34393310546875,
|
| 1104 |
+
"logps/rejected": -402.1936950683594,
|
| 1105 |
+
"loss": 0.5014,
|
| 1106 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1107 |
+
"rewards/chosen": -0.6376605033874512,
|
| 1108 |
+
"rewards/margins": 0.7424057126045227,
|
| 1109 |
+
"rewards/rejected": -1.3800660371780396,
|
| 1110 |
+
"step": 660
|
| 1111 |
+
},
|
| 1112 |
+
{
|
| 1113 |
+
"epoch": 0.7012035583464155,
|
| 1114 |
+
"grad_norm": 15.027313729601833,
|
| 1115 |
+
"learning_rate": 1.239457282149695e-07,
|
| 1116 |
+
"logits/chosen": -0.05018684267997742,
|
| 1117 |
+
"logits/rejected": 0.18970593810081482,
|
| 1118 |
+
"logps/chosen": -376.6749267578125,
|
| 1119 |
+
"logps/rejected": -431.7818298339844,
|
| 1120 |
+
"loss": 0.504,
|
| 1121 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1122 |
+
"rewards/chosen": -0.8920845985412598,
|
| 1123 |
+
"rewards/margins": 0.7974964380264282,
|
| 1124 |
+
"rewards/rejected": -1.6895811557769775,
|
| 1125 |
+
"step": 670
|
| 1126 |
+
},
|
| 1127 |
+
{
|
| 1128 |
+
"epoch": 0.7116692830978545,
|
| 1129 |
+
"grad_norm": 14.878904971459766,
|
| 1130 |
+
"learning_rate": 1.1613595214152711e-07,
|
| 1131 |
+
"logits/chosen": 0.09591469913721085,
|
| 1132 |
+
"logits/rejected": 0.25994688272476196,
|
| 1133 |
+
"logps/chosen": -417.44451904296875,
|
| 1134 |
+
"logps/rejected": -451.24029541015625,
|
| 1135 |
+
"loss": 0.532,
|
| 1136 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1137 |
+
"rewards/chosen": -0.8660572171211243,
|
| 1138 |
+
"rewards/margins": 0.6547313928604126,
|
| 1139 |
+
"rewards/rejected": -1.520788550376892,
|
| 1140 |
+
"step": 680
|
| 1141 |
+
},
|
| 1142 |
+
{
|
| 1143 |
+
"epoch": 0.7221350078492935,
|
| 1144 |
+
"grad_norm": 14.634724746319717,
|
| 1145 |
+
"learning_rate": 1.0850520736699362e-07,
|
| 1146 |
+
"logits/chosen": -0.1941656470298767,
|
| 1147 |
+
"logits/rejected": 0.12699861824512482,
|
| 1148 |
+
"logps/chosen": -326.013671875,
|
| 1149 |
+
"logps/rejected": -370.91943359375,
|
| 1150 |
+
"loss": 0.4932,
|
| 1151 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1152 |
+
"rewards/chosen": -0.4801979064941406,
|
| 1153 |
+
"rewards/margins": 0.7524018883705139,
|
| 1154 |
+
"rewards/rejected": -1.2325998544692993,
|
| 1155 |
+
"step": 690
|
| 1156 |
+
},
|
| 1157 |
+
{
|
| 1158 |
+
"epoch": 0.7326007326007326,
|
| 1159 |
+
"grad_norm": 14.113815152461443,
|
| 1160 |
+
"learning_rate": 1.0106369933615042e-07,
|
| 1161 |
+
"logits/chosen": 0.024108683690428734,
|
| 1162 |
+
"logits/rejected": 0.23957733809947968,
|
| 1163 |
+
"logps/chosen": -328.45379638671875,
|
| 1164 |
+
"logps/rejected": -388.8372802734375,
|
| 1165 |
+
"loss": 0.4967,
|
| 1166 |
+
"rewards/accuracies": 0.8125,
|
| 1167 |
+
"rewards/chosen": -0.5869916677474976,
|
| 1168 |
+
"rewards/margins": 0.8734583854675293,
|
| 1169 |
+
"rewards/rejected": -1.4604499340057373,
|
| 1170 |
+
"step": 700
|
| 1171 |
+
},
|
| 1172 |
+
{
|
| 1173 |
+
"epoch": 0.7326007326007326,
|
| 1174 |
+
"eval_logits/chosen": -0.07450958341360092,
|
| 1175 |
+
"eval_logits/rejected": 0.1898072510957718,
|
| 1176 |
+
"eval_logps/chosen": -361.9271240234375,
|
| 1177 |
+
"eval_logps/rejected": -419.3396911621094,
|
| 1178 |
+
"eval_loss": 0.5152686834335327,
|
| 1179 |
+
"eval_rewards/accuracies": 0.7698412537574768,
|
| 1180 |
+
"eval_rewards/chosen": -0.5681709051132202,
|
| 1181 |
+
"eval_rewards/margins": 0.8120573163032532,
|
| 1182 |
+
"eval_rewards/rejected": -1.3802281618118286,
|
| 1183 |
+
"eval_runtime": 227.5622,
|
| 1184 |
+
"eval_samples_per_second": 8.789,
|
| 1185 |
+
"eval_steps_per_second": 0.277,
|
| 1186 |
+
"step": 700
|
| 1187 |
+
},
|
| 1188 |
+
{
|
| 1189 |
+
"epoch": 0.7430664573521716,
|
| 1190 |
+
"grad_norm": 14.271632027776764,
|
| 1191 |
+
"learning_rate": 9.382138040640714e-08,
|
| 1192 |
+
"logits/chosen": 0.04855802655220032,
|
| 1193 |
+
"logits/rejected": 0.28160515427589417,
|
| 1194 |
+
"logps/chosen": -356.8811340332031,
|
| 1195 |
+
"logps/rejected": -407.11016845703125,
|
| 1196 |
+
"loss": 0.54,
|
| 1197 |
+
"rewards/accuracies": 0.75,
|
| 1198 |
+
"rewards/chosen": -0.633429229259491,
|
| 1199 |
+
"rewards/margins": 0.8234784007072449,
|
| 1200 |
+
"rewards/rejected": -1.4569077491760254,
|
| 1201 |
+
"step": 710
|
| 1202 |
+
},
|
| 1203 |
+
{
|
| 1204 |
+
"epoch": 0.7535321821036107,
|
| 1205 |
+
"grad_norm": 14.7255604500235,
|
| 1206 |
+
"learning_rate": 8.678793653740632e-08,
|
| 1207 |
+
"logits/chosen": -0.11811058223247528,
|
| 1208 |
+
"logits/rejected": 0.15506412088871002,
|
| 1209 |
+
"logps/chosen": -412.05828857421875,
|
| 1210 |
+
"logps/rejected": -441.4915466308594,
|
| 1211 |
+
"loss": 0.5102,
|
| 1212 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1213 |
+
"rewards/chosen": -0.5466042757034302,
|
| 1214 |
+
"rewards/margins": 0.7746764421463013,
|
| 1215 |
+
"rewards/rejected": -1.321280837059021,
|
| 1216 |
+
"step": 720
|
| 1217 |
+
},
|
| 1218 |
+
{
|
| 1219 |
+
"epoch": 0.7639979068550498,
|
| 1220 |
+
"grad_norm": 12.966643699102985,
|
| 1221 |
+
"learning_rate": 7.997277433690983e-08,
|
| 1222 |
+
"logits/chosen": 0.04536755010485649,
|
| 1223 |
+
"logits/rejected": 0.23015692830085754,
|
| 1224 |
+
"logps/chosen": -310.0247802734375,
|
| 1225 |
+
"logps/rejected": -372.56414794921875,
|
| 1226 |
+
"loss": 0.5061,
|
| 1227 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1228 |
+
"rewards/chosen": -0.5736085772514343,
|
| 1229 |
+
"rewards/margins": 0.6976566314697266,
|
| 1230 |
+
"rewards/rejected": -1.2712653875350952,
|
| 1231 |
+
"step": 730
|
| 1232 |
+
},
|
| 1233 |
+
{
|
| 1234 |
+
"epoch": 0.7744636316064888,
|
| 1235 |
+
"grad_norm": 15.46077305724092,
|
| 1236 |
+
"learning_rate": 7.338500848029602e-08,
|
| 1237 |
+
"logits/chosen": -0.012909619137644768,
|
| 1238 |
+
"logits/rejected": 0.290834903717041,
|
| 1239 |
+
"logps/chosen": -359.9085693359375,
|
| 1240 |
+
"logps/rejected": -388.9847412109375,
|
| 1241 |
+
"loss": 0.5014,
|
| 1242 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1243 |
+
"rewards/chosen": -0.6467846632003784,
|
| 1244 |
+
"rewards/margins": 0.8261871337890625,
|
| 1245 |
+
"rewards/rejected": -1.4729719161987305,
|
| 1246 |
+
"step": 740
|
| 1247 |
+
},
|
| 1248 |
+
{
|
| 1249 |
+
"epoch": 0.7849293563579278,
|
| 1250 |
+
"grad_norm": 12.699158494833881,
|
| 1251 |
+
"learning_rate": 6.70334495204884e-08,
|
| 1252 |
+
"logits/chosen": 0.2477736920118332,
|
| 1253 |
+
"logits/rejected": 0.3875717520713806,
|
| 1254 |
+
"logps/chosen": -341.1034240722656,
|
| 1255 |
+
"logps/rejected": -454.39508056640625,
|
| 1256 |
+
"loss": 0.4955,
|
| 1257 |
+
"rewards/accuracies": 0.75,
|
| 1258 |
+
"rewards/chosen": -0.8063145875930786,
|
| 1259 |
+
"rewards/margins": 1.0078232288360596,
|
| 1260 |
+
"rewards/rejected": -1.8141378164291382,
|
| 1261 |
+
"step": 750
|
| 1262 |
+
},
|
| 1263 |
+
{
|
| 1264 |
+
"epoch": 0.7953950811093669,
|
| 1265 |
+
"grad_norm": 14.871002690544678,
|
| 1266 |
+
"learning_rate": 6.092659210462231e-08,
|
| 1267 |
+
"logits/chosen": 0.16597142815589905,
|
| 1268 |
+
"logits/rejected": 0.29944753646850586,
|
| 1269 |
+
"logps/chosen": -337.59442138671875,
|
| 1270 |
+
"logps/rejected": -443.64886474609375,
|
| 1271 |
+
"loss": 0.527,
|
| 1272 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1273 |
+
"rewards/chosen": -0.8204069137573242,
|
| 1274 |
+
"rewards/margins": 0.8414122462272644,
|
| 1275 |
+
"rewards/rejected": -1.6618191003799438,
|
| 1276 |
+
"step": 760
|
| 1277 |
+
},
|
| 1278 |
+
{
|
| 1279 |
+
"epoch": 0.8058608058608059,
|
| 1280 |
+
"grad_norm": 10.845535445418564,
|
| 1281 |
+
"learning_rate": 5.507260361320737e-08,
|
| 1282 |
+
"logits/chosen": -0.017479026690125465,
|
| 1283 |
+
"logits/rejected": 0.06391821056604385,
|
| 1284 |
+
"logps/chosen": -433.59112548828125,
|
| 1285 |
+
"logps/rejected": -509.65447998046875,
|
| 1286 |
+
"loss": 0.4918,
|
| 1287 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1288 |
+
"rewards/chosen": -0.8355442881584167,
|
| 1289 |
+
"rewards/margins": 0.6438878774642944,
|
| 1290 |
+
"rewards/rejected": -1.479432225227356,
|
| 1291 |
+
"step": 770
|
| 1292 |
+
},
|
| 1293 |
+
{
|
| 1294 |
+
"epoch": 0.8163265306122449,
|
| 1295 |
+
"grad_norm": 11.773034698588397,
|
| 1296 |
+
"learning_rate": 4.947931323697982e-08,
|
| 1297 |
+
"logits/chosen": -0.04249686002731323,
|
| 1298 |
+
"logits/rejected": 0.2673026919364929,
|
| 1299 |
+
"logps/chosen": -412.539306640625,
|
| 1300 |
+
"logps/rejected": -427.43341064453125,
|
| 1301 |
+
"loss": 0.5184,
|
| 1302 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1303 |
+
"rewards/chosen": -0.6462075114250183,
|
| 1304 |
+
"rewards/margins": 0.8836655616760254,
|
| 1305 |
+
"rewards/rejected": -1.5298731327056885,
|
| 1306 |
+
"step": 780
|
| 1307 |
+
},
|
| 1308 |
+
{
|
| 1309 |
+
"epoch": 0.826792255363684,
|
| 1310 |
+
"grad_norm": 11.606763284832994,
|
| 1311 |
+
"learning_rate": 4.415420150605398e-08,
|
| 1312 |
+
"logits/chosen": 0.10231854766607285,
|
| 1313 |
+
"logits/rejected": 0.40341711044311523,
|
| 1314 |
+
"logps/chosen": -354.8089904785156,
|
| 1315 |
+
"logps/rejected": -397.8169250488281,
|
| 1316 |
+
"loss": 0.5183,
|
| 1317 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1318 |
+
"rewards/chosen": -0.7357110977172852,
|
| 1319 |
+
"rewards/margins": 0.7533496022224426,
|
| 1320 |
+
"rewards/rejected": -1.489060640335083,
|
| 1321 |
+
"step": 790
|
| 1322 |
+
},
|
| 1323 |
+
{
|
| 1324 |
+
"epoch": 0.837257980115123,
|
| 1325 |
+
"grad_norm": 12.816494511651417,
|
| 1326 |
+
"learning_rate": 3.9104390285376374e-08,
|
| 1327 |
+
"logits/chosen": 0.12218976020812988,
|
| 1328 |
+
"logits/rejected": 0.38090264797210693,
|
| 1329 |
+
"logps/chosen": -338.58837890625,
|
| 1330 |
+
"logps/rejected": -401.3483581542969,
|
| 1331 |
+
"loss": 0.5013,
|
| 1332 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1333 |
+
"rewards/chosen": -0.7186459302902222,
|
| 1334 |
+
"rewards/margins": 0.7460201978683472,
|
| 1335 |
+
"rewards/rejected": -1.4646661281585693,
|
| 1336 |
+
"step": 800
|
| 1337 |
+
},
|
| 1338 |
+
{
|
| 1339 |
+
"epoch": 0.837257980115123,
|
| 1340 |
+
"eval_logits/chosen": -0.030678851529955864,
|
| 1341 |
+
"eval_logits/rejected": 0.2384355366230011,
|
| 1342 |
+
"eval_logps/chosen": -372.57794189453125,
|
| 1343 |
+
"eval_logps/rejected": -429.2723388671875,
|
| 1344 |
+
"eval_loss": 0.51365065574646,
|
| 1345 |
+
"eval_rewards/accuracies": 0.7658730149269104,
|
| 1346 |
+
"eval_rewards/chosen": -0.6746787428855896,
|
| 1347 |
+
"eval_rewards/margins": 0.8048755526542664,
|
| 1348 |
+
"eval_rewards/rejected": -1.479554295539856,
|
| 1349 |
+
"eval_runtime": 228.9904,
|
| 1350 |
+
"eval_samples_per_second": 8.734,
|
| 1351 |
+
"eval_steps_per_second": 0.275,
|
| 1352 |
+
"step": 800
|
| 1353 |
+
},
|
| 1354 |
+
{
|
| 1355 |
+
"epoch": 0.847723704866562,
|
| 1356 |
+
"grad_norm": 11.081154469888522,
|
| 1357 |
+
"learning_rate": 3.433663324986208e-08,
|
| 1358 |
+
"logits/chosen": -0.09540579468011856,
|
| 1359 |
+
"logits/rejected": 0.3088434338569641,
|
| 1360 |
+
"logps/chosen": -419.98004150390625,
|
| 1361 |
+
"logps/rejected": -432.2977600097656,
|
| 1362 |
+
"loss": 0.5109,
|
| 1363 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1364 |
+
"rewards/chosen": -0.7216218709945679,
|
| 1365 |
+
"rewards/margins": 0.8065996170043945,
|
| 1366 |
+
"rewards/rejected": -1.5282217264175415,
|
| 1367 |
+
"step": 810
|
| 1368 |
+
},
|
| 1369 |
+
{
|
| 1370 |
+
"epoch": 0.858189429618001,
|
| 1371 |
+
"grad_norm": 13.289574828613516,
|
| 1372 |
+
"learning_rate": 2.9857306851953897e-08,
|
| 1373 |
+
"logits/chosen": -0.0814175009727478,
|
| 1374 |
+
"logits/rejected": 0.054249562323093414,
|
| 1375 |
+
"logps/chosen": -355.91827392578125,
|
| 1376 |
+
"logps/rejected": -427.5276794433594,
|
| 1377 |
+
"loss": 0.5071,
|
| 1378 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1379 |
+
"rewards/chosen": -0.6057306528091431,
|
| 1380 |
+
"rewards/margins": 0.9160734415054321,
|
| 1381 |
+
"rewards/rejected": -1.521803855895996,
|
| 1382 |
+
"step": 820
|
| 1383 |
+
},
|
| 1384 |
+
{
|
| 1385 |
+
"epoch": 0.8686551543694401,
|
| 1386 |
+
"grad_norm": 13.234567409276869,
|
| 1387 |
+
"learning_rate": 2.567240179368185e-08,
|
| 1388 |
+
"logits/chosen": 0.14148227870464325,
|
| 1389 |
+
"logits/rejected": 0.24102012813091278,
|
| 1390 |
+
"logps/chosen": -332.89105224609375,
|
| 1391 |
+
"logps/rejected": -437.22869873046875,
|
| 1392 |
+
"loss": 0.5167,
|
| 1393 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1394 |
+
"rewards/chosen": -0.7850676774978638,
|
| 1395 |
+
"rewards/margins": 0.8141494989395142,
|
| 1396 |
+
"rewards/rejected": -1.599217176437378,
|
| 1397 |
+
"step": 830
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 0.8791208791208791,
|
| 1401 |
+
"grad_norm": 12.940970521328797,
|
| 1402 |
+
"learning_rate": 2.1787515014630357e-08,
|
| 1403 |
+
"logits/chosen": 0.15607428550720215,
|
| 1404 |
+
"logits/rejected": 0.3544492721557617,
|
| 1405 |
+
"logps/chosen": -375.6202087402344,
|
| 1406 |
+
"logps/rejected": -448.9365234375,
|
| 1407 |
+
"loss": 0.4882,
|
| 1408 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1409 |
+
"rewards/chosen": -0.8064106702804565,
|
| 1410 |
+
"rewards/margins": 0.7966355681419373,
|
| 1411 |
+
"rewards/rejected": -1.603046178817749,
|
| 1412 |
+
"step": 840
|
| 1413 |
+
},
|
| 1414 |
+
{
|
| 1415 |
+
"epoch": 0.8895866038723181,
|
| 1416 |
+
"grad_norm": 10.926513179024628,
|
| 1417 |
+
"learning_rate": 1.820784220652766e-08,
|
| 1418 |
+
"logits/chosen": 0.1379309892654419,
|
| 1419 |
+
"logits/rejected": 0.2978866696357727,
|
| 1420 |
+
"logps/chosen": -351.2433166503906,
|
| 1421 |
+
"logps/rejected": -427.3173828125,
|
| 1422 |
+
"loss": 0.4724,
|
| 1423 |
+
"rewards/accuracies": 0.8125,
|
| 1424 |
+
"rewards/chosen": -0.6311872005462646,
|
| 1425 |
+
"rewards/margins": 0.944290280342102,
|
| 1426 |
+
"rewards/rejected": -1.5754774808883667,
|
| 1427 |
+
"step": 850
|
| 1428 |
+
},
|
| 1429 |
+
{
|
| 1430 |
+
"epoch": 0.9000523286237572,
|
| 1431 |
+
"grad_norm": 12.635492079106957,
|
| 1432 |
+
"learning_rate": 1.4938170864468636e-08,
|
| 1433 |
+
"logits/chosen": 0.00023215674445964396,
|
| 1434 |
+
"logits/rejected": 0.2956046462059021,
|
| 1435 |
+
"logps/chosen": -393.70452880859375,
|
| 1436 |
+
"logps/rejected": -441.1329040527344,
|
| 1437 |
+
"loss": 0.4956,
|
| 1438 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1439 |
+
"rewards/chosen": -0.7465956807136536,
|
| 1440 |
+
"rewards/margins": 0.7083271741867065,
|
| 1441 |
+
"rewards/rejected": -1.4549229145050049,
|
| 1442 |
+
"step": 860
|
| 1443 |
+
},
|
| 1444 |
+
{
|
| 1445 |
+
"epoch": 0.9105180533751962,
|
| 1446 |
+
"grad_norm": 12.131310754312283,
|
| 1447 |
+
"learning_rate": 1.1982873884064465e-08,
|
| 1448 |
+
"logits/chosen": -0.05417024344205856,
|
| 1449 |
+
"logits/rejected": 0.28210294246673584,
|
| 1450 |
+
"logps/chosen": -321.20526123046875,
|
| 1451 |
+
"logps/rejected": -356.74053955078125,
|
| 1452 |
+
"loss": 0.5037,
|
| 1453 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1454 |
+
"rewards/chosen": -0.6472475528717041,
|
| 1455 |
+
"rewards/margins": 0.7646856904029846,
|
| 1456 |
+
"rewards/rejected": -1.411933183670044,
|
| 1457 |
+
"step": 870
|
| 1458 |
+
},
|
| 1459 |
+
{
|
| 1460 |
+
"epoch": 0.9209837781266352,
|
| 1461 |
+
"grad_norm": 14.916064360686926,
|
| 1462 |
+
"learning_rate": 9.345903713082304e-09,
|
| 1463 |
+
"logits/chosen": -0.02495613694190979,
|
| 1464 |
+
"logits/rejected": 0.08404093235731125,
|
| 1465 |
+
"logps/chosen": -360.6625061035156,
|
| 1466 |
+
"logps/rejected": -386.7431640625,
|
| 1467 |
+
"loss": 0.5094,
|
| 1468 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1469 |
+
"rewards/chosen": -0.6419483423233032,
|
| 1470 |
+
"rewards/margins": 0.6380220651626587,
|
| 1471 |
+
"rewards/rejected": -1.279970407485962,
|
| 1472 |
+
"step": 880
|
| 1473 |
+
},
|
| 1474 |
+
{
|
| 1475 |
+
"epoch": 0.9314495028780743,
|
| 1476 |
+
"grad_norm": 11.521916154155225,
|
| 1477 |
+
"learning_rate": 7.030787065396865e-09,
|
| 1478 |
+
"logits/chosen": 0.1541689932346344,
|
| 1479 |
+
"logits/rejected": 0.43220075964927673,
|
| 1480 |
+
"logps/chosen": -333.72039794921875,
|
| 1481 |
+
"logps/rejected": -392.5670166015625,
|
| 1482 |
+
"loss": 0.5092,
|
| 1483 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1484 |
+
"rewards/chosen": -0.6941490173339844,
|
| 1485 |
+
"rewards/margins": 0.8684225082397461,
|
| 1486 |
+
"rewards/rejected": -1.5625712871551514,
|
| 1487 |
+
"step": 890
|
| 1488 |
+
},
|
| 1489 |
+
{
|
| 1490 |
+
"epoch": 0.9419152276295133,
|
| 1491 |
+
"grad_norm": 14.786097584627694,
|
| 1492 |
+
"learning_rate": 5.04062020432286e-09,
|
| 1493 |
+
"logits/chosen": 0.10343728214502335,
|
| 1494 |
+
"logits/rejected": 0.26344841718673706,
|
| 1495 |
+
"logps/chosen": -381.5856018066406,
|
| 1496 |
+
"logps/rejected": -468.1319274902344,
|
| 1497 |
+
"loss": 0.4983,
|
| 1498 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1499 |
+
"rewards/chosen": -0.8143836855888367,
|
| 1500 |
+
"rewards/margins": 0.7843345999717712,
|
| 1501 |
+
"rewards/rejected": -1.5987184047698975,
|
| 1502 |
+
"step": 900
|
| 1503 |
+
},
|
| 1504 |
+
{
|
| 1505 |
+
"epoch": 0.9419152276295133,
|
| 1506 |
+
"eval_logits/chosen": -0.013600568287074566,
|
| 1507 |
+
"eval_logits/rejected": 0.2624098062515259,
|
| 1508 |
+
"eval_logps/chosen": -374.1649475097656,
|
| 1509 |
+
"eval_logps/rejected": -432.5990905761719,
|
| 1510 |
+
"eval_loss": 0.5130622386932373,
|
| 1511 |
+
"eval_rewards/accuracies": 0.7658730149269104,
|
| 1512 |
+
"eval_rewards/chosen": -0.6905492544174194,
|
| 1513 |
+
"eval_rewards/margins": 0.8222722411155701,
|
| 1514 |
+
"eval_rewards/rejected": -1.5128216743469238,
|
| 1515 |
+
"eval_runtime": 231.4541,
|
| 1516 |
+
"eval_samples_per_second": 8.641,
|
| 1517 |
+
"eval_steps_per_second": 0.272,
|
| 1518 |
+
"step": 900
|
| 1519 |
+
},
|
| 1520 |
+
{
|
| 1521 |
+
"epoch": 0.9523809523809523,
|
| 1522 |
+
"grad_norm": 17.98388390329237,
|
| 1523 |
+
"learning_rate": 3.3780648016376866e-09,
|
| 1524 |
+
"logits/chosen": 0.04564341902732849,
|
| 1525 |
+
"logits/rejected": 0.3436654210090637,
|
| 1526 |
+
"logps/chosen": -323.76763916015625,
|
| 1527 |
+
"logps/rejected": -363.12225341796875,
|
| 1528 |
+
"loss": 0.5159,
|
| 1529 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1530 |
+
"rewards/chosen": -0.7020602822303772,
|
| 1531 |
+
"rewards/margins": 0.7331494092941284,
|
| 1532 |
+
"rewards/rejected": -1.4352096319198608,
|
| 1533 |
+
"step": 910
|
| 1534 |
+
},
|
| 1535 |
+
{
|
| 1536 |
+
"epoch": 0.9628466771323915,
|
| 1537 |
+
"grad_norm": 15.497923846134174,
|
| 1538 |
+
"learning_rate": 2.0453443778310766e-09,
|
| 1539 |
+
"logits/chosen": 0.021750714629888535,
|
| 1540 |
+
"logits/rejected": 0.3918168544769287,
|
| 1541 |
+
"logps/chosen": -381.59539794921875,
|
| 1542 |
+
"logps/rejected": -412.8221130371094,
|
| 1543 |
+
"loss": 0.5177,
|
| 1544 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1545 |
+
"rewards/chosen": -0.6423409581184387,
|
| 1546 |
+
"rewards/margins": 0.835128903388977,
|
| 1547 |
+
"rewards/rejected": -1.477469801902771,
|
| 1548 |
+
"step": 920
|
| 1549 |
+
},
|
| 1550 |
+
{
|
| 1551 |
+
"epoch": 0.9733124018838305,
|
| 1552 |
+
"grad_norm": 12.503315786051827,
|
| 1553 |
+
"learning_rate": 1.0442413283435758e-09,
|
| 1554 |
+
"logits/chosen": -0.060670822858810425,
|
| 1555 |
+
"logits/rejected": 0.328230619430542,
|
| 1556 |
+
"logps/chosen": -422.365234375,
|
| 1557 |
+
"logps/rejected": -407.7621765136719,
|
| 1558 |
+
"loss": 0.4789,
|
| 1559 |
+
"rewards/accuracies": 0.6875,
|
| 1560 |
+
"rewards/chosen": -0.6927906274795532,
|
| 1561 |
+
"rewards/margins": 0.6564691662788391,
|
| 1562 |
+
"rewards/rejected": -1.3492597341537476,
|
| 1563 |
+
"step": 930
|
| 1564 |
+
},
|
| 1565 |
+
{
|
| 1566 |
+
"epoch": 0.9837781266352695,
|
| 1567 |
+
"grad_norm": 12.580335513343133,
|
| 1568 |
+
"learning_rate": 3.760945397705828e-10,
|
| 1569 |
+
"logits/chosen": 0.11628331989049911,
|
| 1570 |
+
"logits/rejected": 0.15096750855445862,
|
| 1571 |
+
"logps/chosen": -334.9276428222656,
|
| 1572 |
+
"logps/rejected": -456.6903381347656,
|
| 1573 |
+
"loss": 0.5032,
|
| 1574 |
+
"rewards/accuracies": 0.6875,
|
| 1575 |
+
"rewards/chosen": -0.6851450800895691,
|
| 1576 |
+
"rewards/margins": 0.7425065040588379,
|
| 1577 |
+
"rewards/rejected": -1.4276516437530518,
|
| 1578 |
+
"step": 940
|
| 1579 |
+
},
|
| 1580 |
+
{
|
| 1581 |
+
"epoch": 0.9942438513867086,
|
| 1582 |
+
"grad_norm": 15.517565032289047,
|
| 1583 |
+
"learning_rate": 4.17975992204056e-11,
|
| 1584 |
+
"logits/chosen": -0.20424561202526093,
|
| 1585 |
+
"logits/rejected": 0.1298888772726059,
|
| 1586 |
+
"logps/chosen": -400.3074951171875,
|
| 1587 |
+
"logps/rejected": -450.6595153808594,
|
| 1588 |
+
"loss": 0.5105,
|
| 1589 |
+
"rewards/accuracies": 0.75,
|
| 1590 |
+
"rewards/chosen": -0.7332125306129456,
|
| 1591 |
+
"rewards/margins": 0.8044403195381165,
|
| 1592 |
+
"rewards/rejected": -1.537652850151062,
|
| 1593 |
+
"step": 950
|
| 1594 |
+
},
|
| 1595 |
+
{
|
| 1596 |
+
"epoch": 0.9994767137624281,
|
| 1597 |
+
"step": 955,
|
| 1598 |
+
"total_flos": 0.0,
|
| 1599 |
+
"train_loss": 0.5504645167845081,
|
| 1600 |
+
"train_runtime": 19044.9678,
|
| 1601 |
+
"train_samples_per_second": 3.21,
|
| 1602 |
+
"train_steps_per_second": 0.05
|
| 1603 |
+
}
|
| 1604 |
+
],
|
| 1605 |
+
"logging_steps": 10,
|
| 1606 |
+
"max_steps": 955,
|
| 1607 |
+
"num_input_tokens_seen": 0,
|
| 1608 |
+
"num_train_epochs": 1,
|
| 1609 |
+
"save_steps": 100,
|
| 1610 |
+
"stateful_callbacks": {
|
| 1611 |
+
"TrainerControl": {
|
| 1612 |
+
"args": {
|
| 1613 |
+
"should_epoch_stop": false,
|
| 1614 |
+
"should_evaluate": false,
|
| 1615 |
+
"should_log": false,
|
| 1616 |
+
"should_save": true,
|
| 1617 |
+
"should_training_stop": true
|
| 1618 |
+
},
|
| 1619 |
+
"attributes": {}
|
| 1620 |
+
}
|
| 1621 |
+
},
|
| 1622 |
+
"total_flos": 0.0,
|
| 1623 |
+
"train_batch_size": 4,
|
| 1624 |
+
"trial_name": null,
|
| 1625 |
+
"trial_params": null
|
| 1626 |
+
}
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7608
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fae8eb2d0bd0b9b8c23b074d41f71a59bccbb31a96063eb67527f8c6659d7db1
|
| 3 |
size 7608
|