Model save
Browse files- README.md +21 -6
- all_results.json +5 -6
- generation_config.json +1 -1
- model-00001-of-00003.safetensors +1 -1
- model-00002-of-00003.safetensors +1 -1
- model-00003-of-00003.safetensors +1 -1
- train_results.json +5 -6
- trainer_state.json +680 -108
- training_args.bin +1 -1
README.md
CHANGED
|
@@ -13,10 +13,19 @@ model-index:
|
|
| 13 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 14 |
should probably proofread and complete it, then remove this comment. -->
|
| 15 |
|
| 16 |
-
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/sanqiang/wdpo/runs/1co2gbm0)
|
| 17 |
# zephyr-7b-dpo-full
|
| 18 |
|
| 19 |
This model is a fine-tuned version of [HuggingFaceH4/mistral-7b-sft-beta](https://huggingface.co/HuggingFaceH4/mistral-7b-sft-beta) on the None dataset.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
|
| 21 |
## Model description
|
| 22 |
|
|
@@ -38,11 +47,11 @@ The following hyperparameters were used during training:
|
|
| 38 |
- learning_rate: 5e-07
|
| 39 |
- train_batch_size: 8
|
| 40 |
- eval_batch_size: 8
|
| 41 |
-
- seed:
|
| 42 |
- distributed_type: multi-GPU
|
| 43 |
- num_devices: 8
|
| 44 |
-
- gradient_accumulation_steps:
|
| 45 |
-
- total_train_batch_size:
|
| 46 |
- total_eval_batch_size: 64
|
| 47 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 48 |
- lr_scheduler_type: cosine
|
|
@@ -51,11 +60,17 @@ The following hyperparameters were used during training:
|
|
| 51 |
|
| 52 |
### Training results
|
| 53 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 54 |
|
| 55 |
|
| 56 |
### Framework versions
|
| 57 |
|
| 58 |
-
- Transformers 4.
|
| 59 |
- Pytorch 2.1.2+cu121
|
| 60 |
- Datasets 2.14.6
|
| 61 |
-
- Tokenizers 0.
|
|
|
|
| 13 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 14 |
should probably proofread and complete it, then remove this comment. -->
|
| 15 |
|
|
|
|
| 16 |
# zephyr-7b-dpo-full
|
| 17 |
|
| 18 |
This model is a fine-tuned version of [HuggingFaceH4/mistral-7b-sft-beta](https://huggingface.co/HuggingFaceH4/mistral-7b-sft-beta) on the None dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.0544
|
| 21 |
+
- Rewards/chosen: -2.1973
|
| 22 |
+
- Rewards/rejected: -2.9243
|
| 23 |
+
- Rewards/accuracies: 0.7070
|
| 24 |
+
- Rewards/margins: 0.7270
|
| 25 |
+
- Logps/rejected: -549.7877
|
| 26 |
+
- Logps/chosen: -476.7722
|
| 27 |
+
- Logits/rejected: -1.9407
|
| 28 |
+
- Logits/chosen: -1.9849
|
| 29 |
|
| 30 |
## Model description
|
| 31 |
|
|
|
|
| 47 |
- learning_rate: 5e-07
|
| 48 |
- train_batch_size: 8
|
| 49 |
- eval_batch_size: 8
|
| 50 |
+
- seed: 3
|
| 51 |
- distributed_type: multi-GPU
|
| 52 |
- num_devices: 8
|
| 53 |
+
- gradient_accumulation_steps: 4
|
| 54 |
+
- total_train_batch_size: 256
|
| 55 |
- total_eval_batch_size: 64
|
| 56 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 57 |
- lr_scheduler_type: cosine
|
|
|
|
| 60 |
|
| 61 |
### Training results
|
| 62 |
|
| 63 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 64 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 65 |
+
| 0.0846 | 0.23 | 100 | 0.0846 | -1.4642 | -1.8940 | 0.6484 | 0.4298 | -446.7535 | -403.4620 | -2.3302 | -2.3522 |
|
| 66 |
+
| 0.0477 | 0.45 | 200 | 0.0672 | -1.7958 | -2.4017 | 0.7148 | 0.6059 | -497.5217 | -436.6205 | -2.1284 | -2.1617 |
|
| 67 |
+
| 0.046 | 0.68 | 300 | 0.0552 | -2.1484 | -2.8722 | 0.7148 | 0.7238 | -544.5698 | -471.8781 | -1.9484 | -1.9914 |
|
| 68 |
+
| 0.0439 | 0.91 | 400 | 0.0544 | -2.1973 | -2.9243 | 0.7070 | 0.7270 | -549.7877 | -476.7722 | -1.9407 | -1.9849 |
|
| 69 |
|
| 70 |
|
| 71 |
### Framework versions
|
| 72 |
|
| 73 |
+
- Transformers 4.35.2
|
| 74 |
- Pytorch 2.1.2+cu121
|
| 75 |
- Datasets 2.14.6
|
| 76 |
+
- Tokenizers 0.14.1
|
all_results.json
CHANGED
|
@@ -1,9 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"
|
| 4 |
-
"
|
| 5 |
-
"
|
| 6 |
-
"
|
| 7 |
-
"
|
| 8 |
-
"train_steps_per_second": 0.138
|
| 9 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.07998780592802972,
|
| 4 |
+
"train_runtime": 7378.9712,
|
| 5 |
+
"train_samples": 113028,
|
| 6 |
+
"train_samples_per_second": 15.318,
|
| 7 |
+
"train_steps_per_second": 0.06
|
|
|
|
| 8 |
}
|
generation_config.json
CHANGED
|
@@ -2,5 +2,5 @@
|
|
| 2 |
"_from_model_config": true,
|
| 3 |
"bos_token_id": 1,
|
| 4 |
"eos_token_id": 2,
|
| 5 |
-
"transformers_version": "4.
|
| 6 |
}
|
|
|
|
| 2 |
"_from_model_config": true,
|
| 3 |
"bos_token_id": 1,
|
| 4 |
"eos_token_id": 2,
|
| 5 |
+
"transformers_version": "4.35.2"
|
| 6 |
}
|
model-00001-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4943162336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:811536fd507edc48842921f5fe58009c4d68ba8670577fea3ab063a8b487349b
|
| 3 |
size 4943162336
|
model-00002-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4999819336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fd5a311a089f3ba60ceb62659015c6d5a3aad05b534064201ca1668bddfcc05d
|
| 3 |
size 4999819336
|
model-00003-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4540516344
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:629550c668df231f1f9b855ba38797fad39fafa90ac09f07dda826c88b931c9a
|
| 3 |
size 4540516344
|
train_results.json
CHANGED
|
@@ -1,9 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"
|
| 4 |
-
"
|
| 5 |
-
"
|
| 6 |
-
"
|
| 7 |
-
"
|
| 8 |
-
"train_steps_per_second": 0.138
|
| 9 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.07998780592802972,
|
| 4 |
+
"train_runtime": 7378.9712,
|
| 5 |
+
"train_samples": 113028,
|
| 6 |
+
"train_samples_per_second": 15.318,
|
| 7 |
+
"train_steps_per_second": 0.06
|
|
|
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -1,25 +1,21 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch":
|
| 5 |
"eval_steps": 100,
|
| 6 |
-
"global_step":
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
| 10 |
"log_history": [
|
| 11 |
{
|
| 12 |
-
"
|
| 13 |
-
"
|
| 14 |
-
"
|
| 15 |
-
"
|
| 16 |
-
"
|
| 17 |
-
"
|
| 18 |
-
"
|
| 19 |
-
"logits/rejected": -2.8797199726104736,
|
| 20 |
-
"logps/chosen": -320.43853759765625,
|
| 21 |
-
"logps/rejected": -340.07073974609375,
|
| 22 |
-
"loss": 0.2116,
|
| 23 |
"rewards/accuracies": 0.0,
|
| 24 |
"rewards/chosen": 0.0,
|
| 25 |
"rewards/margins": 0.0,
|
|
@@ -27,124 +23,700 @@
|
|
| 27 |
"step": 1
|
| 28 |
},
|
| 29 |
{
|
| 30 |
-
"
|
| 31 |
-
"
|
| 32 |
-
"
|
| 33 |
-
"
|
| 34 |
-
"
|
| 35 |
-
"
|
| 36 |
-
"
|
| 37 |
-
"
|
| 38 |
-
"
|
| 39 |
-
"
|
| 40 |
-
"
|
| 41 |
-
"rewards/accuracies": 0.5416666865348816,
|
| 42 |
-
"rewards/chosen": 0.0007928035338409245,
|
| 43 |
-
"rewards/margins": 0.005046091042459011,
|
| 44 |
-
"rewards/rejected": -0.004253287799656391,
|
| 45 |
"step": 10
|
| 46 |
},
|
| 47 |
{
|
| 48 |
-
"
|
| 49 |
-
"
|
| 50 |
-
"
|
| 51 |
-
"
|
| 52 |
-
"
|
| 53 |
-
"
|
| 54 |
-
"
|
| 55 |
-
"
|
| 56 |
-
"
|
| 57 |
-
"
|
| 58 |
-
"
|
| 59 |
-
"rewards/accuracies": 0.6312500238418579,
|
| 60 |
-
"rewards/chosen": 0.0035451077856123447,
|
| 61 |
-
"rewards/margins": 0.04073050990700722,
|
| 62 |
-
"rewards/rejected": -0.03718540072441101,
|
| 63 |
"step": 20
|
| 64 |
},
|
| 65 |
{
|
| 66 |
-
"
|
| 67 |
-
"
|
| 68 |
-
"
|
| 69 |
-
"
|
| 70 |
-
"
|
| 71 |
-
"
|
| 72 |
-
"
|
| 73 |
-
"
|
| 74 |
-
"
|
| 75 |
-
"
|
| 76 |
-
"
|
| 77 |
-
"rewards/accuracies": 0.7437499761581421,
|
| 78 |
-
"rewards/chosen": -0.013330144807696342,
|
| 79 |
-
"rewards/margins": 0.11589495837688446,
|
| 80 |
-
"rewards/rejected": -0.12922510504722595,
|
| 81 |
"step": 30
|
| 82 |
},
|
| 83 |
{
|
| 84 |
-
"
|
| 85 |
-
"
|
| 86 |
-
"
|
| 87 |
-
"
|
| 88 |
-
"
|
| 89 |
-
"
|
| 90 |
-
"
|
| 91 |
-
"
|
| 92 |
-
"
|
| 93 |
-
"
|
| 94 |
-
"
|
| 95 |
-
"rewards/accuracies": 0.731249988079071,
|
| 96 |
-
"rewards/chosen": -0.010476941242814064,
|
| 97 |
-
"rewards/margins": 0.30531758069992065,
|
| 98 |
-
"rewards/rejected": -0.31579452753067017,
|
| 99 |
"step": 40
|
| 100 |
},
|
| 101 |
{
|
| 102 |
-
"
|
| 103 |
-
"
|
| 104 |
-
"
|
| 105 |
-
"
|
| 106 |
-
"
|
| 107 |
-
"
|
| 108 |
-
"
|
| 109 |
-
"
|
| 110 |
-
"
|
| 111 |
-
"
|
| 112 |
-
"
|
| 113 |
-
"rewards/accuracies": 0.668749988079071,
|
| 114 |
-
"rewards/chosen": -0.026972616091370583,
|
| 115 |
-
"rewards/margins": 0.15525773167610168,
|
| 116 |
-
"rewards/rejected": -0.18223035335540771,
|
| 117 |
"step": 50
|
| 118 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 119 |
{
|
| 120 |
"epoch": 1.0,
|
| 121 |
-
"step":
|
| 122 |
"total_flos": 0.0,
|
| 123 |
-
"train_loss": 0.
|
| 124 |
-
"train_runtime":
|
| 125 |
-
"train_samples_per_second":
|
| 126 |
-
"train_steps_per_second": 0.
|
| 127 |
}
|
| 128 |
],
|
| 129 |
"logging_steps": 10,
|
| 130 |
-
"max_steps":
|
| 131 |
-
"num_input_tokens_seen": 0,
|
| 132 |
"num_train_epochs": 1,
|
| 133 |
"save_steps": 100,
|
| 134 |
-
"stateful_callbacks": {
|
| 135 |
-
"TrainerControl": {
|
| 136 |
-
"args": {
|
| 137 |
-
"should_epoch_stop": false,
|
| 138 |
-
"should_evaluate": false,
|
| 139 |
-
"should_log": false,
|
| 140 |
-
"should_save": false,
|
| 141 |
-
"should_training_stop": false
|
| 142 |
-
},
|
| 143 |
-
"attributes": {}
|
| 144 |
-
}
|
| 145 |
-
},
|
| 146 |
"total_flos": 0.0,
|
| 147 |
-
"train_batch_size": 8,
|
| 148 |
"trial_name": null,
|
| 149 |
"trial_params": null
|
| 150 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.99830220713073,
|
| 5 |
"eval_steps": 100,
|
| 6 |
+
"global_step": 441,
|
| 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": 1.111111111111111e-08,
|
| 14 |
+
"logits/chosen": -2.669281482696533,
|
| 15 |
+
"logits/rejected": -2.675659418106079,
|
| 16 |
+
"logps/chosen": -301.2757873535156,
|
| 17 |
+
"logps/rejected": -280.8008728027344,
|
| 18 |
+
"loss": 0.2803,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
"rewards/accuracies": 0.0,
|
| 20 |
"rewards/chosen": 0.0,
|
| 21 |
"rewards/margins": 0.0,
|
|
|
|
| 23 |
"step": 1
|
| 24 |
},
|
| 25 |
{
|
| 26 |
+
"epoch": 0.02,
|
| 27 |
+
"learning_rate": 1.111111111111111e-07,
|
| 28 |
+
"logits/chosen": -2.8107264041900635,
|
| 29 |
+
"logits/rejected": -2.7811262607574463,
|
| 30 |
+
"logps/chosen": -320.0479736328125,
|
| 31 |
+
"logps/rejected": -195.18087768554688,
|
| 32 |
+
"loss": 0.2792,
|
| 33 |
+
"rewards/accuracies": 0.4826388955116272,
|
| 34 |
+
"rewards/chosen": 0.0006166233215481043,
|
| 35 |
+
"rewards/margins": 0.0009485264890827239,
|
| 36 |
+
"rewards/rejected": -0.00033190299291163683,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
"step": 10
|
| 38 |
},
|
| 39 |
{
|
| 40 |
+
"epoch": 0.05,
|
| 41 |
+
"learning_rate": 2.222222222222222e-07,
|
| 42 |
+
"logits/chosen": -2.8013899326324463,
|
| 43 |
+
"logits/rejected": -2.7626068592071533,
|
| 44 |
+
"logps/chosen": -350.6124572753906,
|
| 45 |
+
"logps/rejected": -191.4945831298828,
|
| 46 |
+
"loss": 0.2774,
|
| 47 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 48 |
+
"rewards/chosen": 0.0069363838993012905,
|
| 49 |
+
"rewards/margins": 0.013156639412045479,
|
| 50 |
+
"rewards/rejected": -0.006220255978405476,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51 |
"step": 20
|
| 52 |
},
|
| 53 |
{
|
| 54 |
+
"epoch": 0.07,
|
| 55 |
+
"learning_rate": 3.333333333333333e-07,
|
| 56 |
+
"logits/chosen": -2.7069194316864014,
|
| 57 |
+
"logits/rejected": -2.691702365875244,
|
| 58 |
+
"logps/chosen": -316.24737548828125,
|
| 59 |
+
"logps/rejected": -201.17063903808594,
|
| 60 |
+
"loss": 0.2782,
|
| 61 |
+
"rewards/accuracies": 0.6656249761581421,
|
| 62 |
+
"rewards/chosen": 0.0192975215613842,
|
| 63 |
+
"rewards/margins": 0.07039008289575577,
|
| 64 |
+
"rewards/rejected": -0.05109255388379097,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 65 |
"step": 30
|
| 66 |
},
|
| 67 |
{
|
| 68 |
+
"epoch": 0.09,
|
| 69 |
+
"learning_rate": 4.444444444444444e-07,
|
| 70 |
+
"logits/chosen": -2.606764554977417,
|
| 71 |
+
"logits/rejected": -2.5910491943359375,
|
| 72 |
+
"logps/chosen": -376.6419982910156,
|
| 73 |
+
"logps/rejected": -222.21923828125,
|
| 74 |
+
"loss": 0.2589,
|
| 75 |
+
"rewards/accuracies": 0.699999988079071,
|
| 76 |
+
"rewards/chosen": 0.015041938051581383,
|
| 77 |
+
"rewards/margins": 0.2519453465938568,
|
| 78 |
+
"rewards/rejected": -0.23690339922904968,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 79 |
"step": 40
|
| 80 |
},
|
| 81 |
{
|
| 82 |
+
"epoch": 0.11,
|
| 83 |
+
"learning_rate": 4.998033461515242e-07,
|
| 84 |
+
"logits/chosen": -2.5254263877868652,
|
| 85 |
+
"logits/rejected": -2.522778034210205,
|
| 86 |
+
"logps/chosen": -350.6036682128906,
|
| 87 |
+
"logps/rejected": -230.76931762695312,
|
| 88 |
+
"loss": 0.2151,
|
| 89 |
+
"rewards/accuracies": 0.637499988079071,
|
| 90 |
+
"rewards/chosen": -0.16026124358177185,
|
| 91 |
+
"rewards/margins": 0.36047226190567017,
|
| 92 |
+
"rewards/rejected": -0.5207335352897644,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 93 |
"step": 50
|
| 94 |
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.14,
|
| 97 |
+
"learning_rate": 4.982319711683221e-07,
|
| 98 |
+
"logits/chosen": -2.509260654449463,
|
| 99 |
+
"logits/rejected": -2.4819066524505615,
|
| 100 |
+
"logps/chosen": -343.71575927734375,
|
| 101 |
+
"logps/rejected": -281.47723388671875,
|
| 102 |
+
"loss": 0.1652,
|
| 103 |
+
"rewards/accuracies": 0.703125,
|
| 104 |
+
"rewards/chosen": -0.3321402072906494,
|
| 105 |
+
"rewards/margins": 0.506214439868927,
|
| 106 |
+
"rewards/rejected": -0.8383547067642212,
|
| 107 |
+
"step": 60
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.16,
|
| 111 |
+
"learning_rate": 4.950991058546892e-07,
|
| 112 |
+
"logits/chosen": -2.4952120780944824,
|
| 113 |
+
"logits/rejected": -2.4609100818634033,
|
| 114 |
+
"logps/chosen": -381.2798156738281,
|
| 115 |
+
"logps/rejected": -299.72149658203125,
|
| 116 |
+
"loss": 0.1193,
|
| 117 |
+
"rewards/accuracies": 0.7093750238418579,
|
| 118 |
+
"rewards/chosen": -0.518448531627655,
|
| 119 |
+
"rewards/margins": 0.5417992472648621,
|
| 120 |
+
"rewards/rejected": -1.0602478981018066,
|
| 121 |
+
"step": 70
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.18,
|
| 125 |
+
"learning_rate": 4.904244573372733e-07,
|
| 126 |
+
"logits/chosen": -2.4261202812194824,
|
| 127 |
+
"logits/rejected": -2.4001305103302,
|
| 128 |
+
"logps/chosen": -423.95233154296875,
|
| 129 |
+
"logps/rejected": -343.5025939941406,
|
| 130 |
+
"loss": 0.0926,
|
| 131 |
+
"rewards/accuracies": 0.737500011920929,
|
| 132 |
+
"rewards/chosen": -0.8466840982437134,
|
| 133 |
+
"rewards/margins": 0.7109834551811218,
|
| 134 |
+
"rewards/rejected": -1.55766761302948,
|
| 135 |
+
"step": 80
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.2,
|
| 139 |
+
"learning_rate": 4.842374312499405e-07,
|
| 140 |
+
"logits/chosen": -2.4205052852630615,
|
| 141 |
+
"logits/rejected": -2.3753650188446045,
|
| 142 |
+
"logps/chosen": -425.88934326171875,
|
| 143 |
+
"logps/rejected": -375.40478515625,
|
| 144 |
+
"loss": 0.0715,
|
| 145 |
+
"rewards/accuracies": 0.731249988079071,
|
| 146 |
+
"rewards/chosen": -1.0327210426330566,
|
| 147 |
+
"rewards/margins": 0.8036619424819946,
|
| 148 |
+
"rewards/rejected": -1.8363832235336304,
|
| 149 |
+
"step": 90
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.23,
|
| 153 |
+
"learning_rate": 4.7657694675916247e-07,
|
| 154 |
+
"logits/chosen": -2.3601014614105225,
|
| 155 |
+
"logits/rejected": -2.3345751762390137,
|
| 156 |
+
"logps/chosen": -415.061279296875,
|
| 157 |
+
"logps/rejected": -360.3291931152344,
|
| 158 |
+
"loss": 0.0846,
|
| 159 |
+
"rewards/accuracies": 0.721875011920929,
|
| 160 |
+
"rewards/chosen": -0.9150521159172058,
|
| 161 |
+
"rewards/margins": 0.7828376889228821,
|
| 162 |
+
"rewards/rejected": -1.6978899240493774,
|
| 163 |
+
"step": 100
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.23,
|
| 167 |
+
"eval_logits/chosen": -2.3521652221679688,
|
| 168 |
+
"eval_logits/rejected": -2.330230236053467,
|
| 169 |
+
"eval_logps/chosen": -403.4620361328125,
|
| 170 |
+
"eval_logps/rejected": -446.7535400390625,
|
| 171 |
+
"eval_loss": 0.08460698276758194,
|
| 172 |
+
"eval_rewards/accuracies": 0.6484375,
|
| 173 |
+
"eval_rewards/chosen": -1.4642237424850464,
|
| 174 |
+
"eval_rewards/margins": 0.42977917194366455,
|
| 175 |
+
"eval_rewards/rejected": -1.894002914428711,
|
| 176 |
+
"eval_runtime": 53.5062,
|
| 177 |
+
"eval_samples_per_second": 37.379,
|
| 178 |
+
"eval_steps_per_second": 0.598,
|
| 179 |
+
"step": 100
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 0.25,
|
| 183 |
+
"learning_rate": 4.6749119174501973e-07,
|
| 184 |
+
"logits/chosen": -2.338693857192993,
|
| 185 |
+
"logits/rejected": -2.274574041366577,
|
| 186 |
+
"logps/chosen": -452.51953125,
|
| 187 |
+
"logps/rejected": -401.04058837890625,
|
| 188 |
+
"loss": 0.0679,
|
| 189 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 190 |
+
"rewards/chosen": -1.2087438106536865,
|
| 191 |
+
"rewards/margins": 0.9248006939888,
|
| 192 |
+
"rewards/rejected": -2.133544445037842,
|
| 193 |
+
"step": 110
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 0.27,
|
| 197 |
+
"learning_rate": 4.5703731967784265e-07,
|
| 198 |
+
"logits/chosen": -2.2874975204467773,
|
| 199 |
+
"logits/rejected": -2.2298505306243896,
|
| 200 |
+
"logps/chosen": -436.90283203125,
|
| 201 |
+
"logps/rejected": -372.56170654296875,
|
| 202 |
+
"loss": 0.0738,
|
| 203 |
+
"rewards/accuracies": 0.7718750238418579,
|
| 204 |
+
"rewards/chosen": -0.8776865005493164,
|
| 205 |
+
"rewards/margins": 0.9758397340774536,
|
| 206 |
+
"rewards/rejected": -1.8535263538360596,
|
| 207 |
+
"step": 120
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 0.29,
|
| 211 |
+
"learning_rate": 4.4528109009727333e-07,
|
| 212 |
+
"logits/chosen": -2.297295093536377,
|
| 213 |
+
"logits/rejected": -2.2529804706573486,
|
| 214 |
+
"logps/chosen": -418.8360290527344,
|
| 215 |
+
"logps/rejected": -369.44451904296875,
|
| 216 |
+
"loss": 0.084,
|
| 217 |
+
"rewards/accuracies": 0.753125011920929,
|
| 218 |
+
"rewards/chosen": -0.8223110437393188,
|
| 219 |
+
"rewards/margins": 0.993333637714386,
|
| 220 |
+
"rewards/rejected": -1.8156448602676392,
|
| 221 |
+
"step": 130
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.32,
|
| 225 |
+
"learning_rate": 4.3229645495529427e-07,
|
| 226 |
+
"logits/chosen": -2.30572509765625,
|
| 227 |
+
"logits/rejected": -2.214725971221924,
|
| 228 |
+
"logps/chosen": -461.4957580566406,
|
| 229 |
+
"logps/rejected": -416.990234375,
|
| 230 |
+
"loss": 0.066,
|
| 231 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 232 |
+
"rewards/chosen": -1.087278962135315,
|
| 233 |
+
"rewards/margins": 1.041991949081421,
|
| 234 |
+
"rewards/rejected": -2.1292710304260254,
|
| 235 |
+
"step": 140
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.34,
|
| 239 |
+
"learning_rate": 4.1816509342531317e-07,
|
| 240 |
+
"logits/chosen": -2.2726972103118896,
|
| 241 |
+
"logits/rejected": -2.196261405944824,
|
| 242 |
+
"logps/chosen": -419.60479736328125,
|
| 243 |
+
"logps/rejected": -361.95831298828125,
|
| 244 |
+
"loss": 0.0853,
|
| 245 |
+
"rewards/accuracies": 0.78125,
|
| 246 |
+
"rewards/chosen": -0.6499409675598145,
|
| 247 |
+
"rewards/margins": 1.046847939491272,
|
| 248 |
+
"rewards/rejected": -1.696789026260376,
|
| 249 |
+
"step": 150
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.36,
|
| 253 |
+
"learning_rate": 4.0297589810356166e-07,
|
| 254 |
+
"logits/chosen": -2.188967227935791,
|
| 255 |
+
"logits/rejected": -2.119161605834961,
|
| 256 |
+
"logps/chosen": -450.3418884277344,
|
| 257 |
+
"logps/rejected": -422.260986328125,
|
| 258 |
+
"loss": 0.0621,
|
| 259 |
+
"rewards/accuracies": 0.734375,
|
| 260 |
+
"rewards/chosen": -1.2301867008209229,
|
| 261 |
+
"rewards/margins": 1.049591064453125,
|
| 262 |
+
"rewards/rejected": -2.279777765274048,
|
| 263 |
+
"step": 160
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"epoch": 0.38,
|
| 267 |
+
"learning_rate": 3.868244158348331e-07,
|
| 268 |
+
"logits/chosen": -2.135490894317627,
|
| 269 |
+
"logits/rejected": -2.0569894313812256,
|
| 270 |
+
"logps/chosen": -508.18511962890625,
|
| 271 |
+
"logps/rejected": -465.81689453125,
|
| 272 |
+
"loss": 0.0461,
|
| 273 |
+
"rewards/accuracies": 0.746874988079071,
|
| 274 |
+
"rewards/chosen": -1.621382474899292,
|
| 275 |
+
"rewards/margins": 1.0623977184295654,
|
| 276 |
+
"rewards/rejected": -2.6837801933288574,
|
| 277 |
+
"step": 170
|
| 278 |
+
},
|
| 279 |
+
{
|
| 280 |
+
"epoch": 0.41,
|
| 281 |
+
"learning_rate": 3.698122466800142e-07,
|
| 282 |
+
"logits/chosen": -2.1748709678649902,
|
| 283 |
+
"logits/rejected": -2.0735726356506348,
|
| 284 |
+
"logps/chosen": -480.3050842285156,
|
| 285 |
+
"logps/rejected": -428.5189514160156,
|
| 286 |
+
"loss": 0.0508,
|
| 287 |
+
"rewards/accuracies": 0.7093750238418579,
|
| 288 |
+
"rewards/chosen": -1.518842101097107,
|
| 289 |
+
"rewards/margins": 0.8754765391349792,
|
| 290 |
+
"rewards/rejected": -2.3943190574645996,
|
| 291 |
+
"step": 180
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"epoch": 0.43,
|
| 295 |
+
"learning_rate": 3.5204640480617574e-07,
|
| 296 |
+
"logits/chosen": -2.104471206665039,
|
| 297 |
+
"logits/rejected": -2.028428077697754,
|
| 298 |
+
"logps/chosen": -493.2584533691406,
|
| 299 |
+
"logps/rejected": -451.6133728027344,
|
| 300 |
+
"loss": 0.0534,
|
| 301 |
+
"rewards/accuracies": 0.6968749761581421,
|
| 302 |
+
"rewards/chosen": -1.6701205968856812,
|
| 303 |
+
"rewards/margins": 0.9617946743965149,
|
| 304 |
+
"rewards/rejected": -2.63191556930542,
|
| 305 |
+
"step": 190
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 0.45,
|
| 309 |
+
"learning_rate": 3.336386453195088e-07,
|
| 310 |
+
"logits/chosen": -2.155747890472412,
|
| 311 |
+
"logits/rejected": -2.0854601860046387,
|
| 312 |
+
"logps/chosen": -486.59954833984375,
|
| 313 |
+
"logps/rejected": -467.2601623535156,
|
| 314 |
+
"loss": 0.0477,
|
| 315 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 316 |
+
"rewards/chosen": -1.5419719219207764,
|
| 317 |
+
"rewards/margins": 1.163585901260376,
|
| 318 |
+
"rewards/rejected": -2.7055578231811523,
|
| 319 |
+
"step": 200
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"epoch": 0.45,
|
| 323 |
+
"eval_logits/chosen": -2.1616616249084473,
|
| 324 |
+
"eval_logits/rejected": -2.128385305404663,
|
| 325 |
+
"eval_logps/chosen": -436.6204833984375,
|
| 326 |
+
"eval_logps/rejected": -497.52166748046875,
|
| 327 |
+
"eval_loss": 0.06721889227628708,
|
| 328 |
+
"eval_rewards/accuracies": 0.71484375,
|
| 329 |
+
"eval_rewards/chosen": -1.7958087921142578,
|
| 330 |
+
"eval_rewards/margins": 0.6058750152587891,
|
| 331 |
+
"eval_rewards/rejected": -2.401683807373047,
|
| 332 |
+
"eval_runtime": 53.3708,
|
| 333 |
+
"eval_samples_per_second": 37.474,
|
| 334 |
+
"eval_steps_per_second": 0.6,
|
| 335 |
+
"step": 200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 0.48,
|
| 339 |
+
"learning_rate": 3.147047612756302e-07,
|
| 340 |
+
"logits/chosen": -2.1386637687683105,
|
| 341 |
+
"logits/rejected": -2.041881561279297,
|
| 342 |
+
"logps/chosen": -493.37530517578125,
|
| 343 |
+
"logps/rejected": -443.93377685546875,
|
| 344 |
+
"loss": 0.0514,
|
| 345 |
+
"rewards/accuracies": 0.765625,
|
| 346 |
+
"rewards/chosen": -1.43257737159729,
|
| 347 |
+
"rewards/margins": 1.1513398885726929,
|
| 348 |
+
"rewards/rejected": -2.5839171409606934,
|
| 349 |
+
"step": 210
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 0.5,
|
| 353 |
+
"learning_rate": 2.9536385528937565e-07,
|
| 354 |
+
"logits/chosen": -2.1365461349487305,
|
| 355 |
+
"logits/rejected": -2.0607683658599854,
|
| 356 |
+
"logps/chosen": -505.8017578125,
|
| 357 |
+
"logps/rejected": -459.7229919433594,
|
| 358 |
+
"loss": 0.0508,
|
| 359 |
+
"rewards/accuracies": 0.7718750238418579,
|
| 360 |
+
"rewards/chosen": -1.424095869064331,
|
| 361 |
+
"rewards/margins": 1.3351457118988037,
|
| 362 |
+
"rewards/rejected": -2.7592415809631348,
|
| 363 |
+
"step": 220
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 0.52,
|
| 367 |
+
"learning_rate": 2.7573759032598365e-07,
|
| 368 |
+
"logits/chosen": -2.12797474861145,
|
| 369 |
+
"logits/rejected": -2.06542706489563,
|
| 370 |
+
"logps/chosen": -517.1934814453125,
|
| 371 |
+
"logps/rejected": -477.3455505371094,
|
| 372 |
+
"loss": 0.049,
|
| 373 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 374 |
+
"rewards/chosen": -1.5975525379180908,
|
| 375 |
+
"rewards/margins": 1.0971721410751343,
|
| 376 |
+
"rewards/rejected": -2.6947245597839355,
|
| 377 |
+
"step": 230
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 0.54,
|
| 381 |
+
"learning_rate": 2.5594942438652685e-07,
|
| 382 |
+
"logits/chosen": -2.052783250808716,
|
| 383 |
+
"logits/rejected": -1.9777238368988037,
|
| 384 |
+
"logps/chosen": -483.03057861328125,
|
| 385 |
+
"logps/rejected": -474.64654541015625,
|
| 386 |
+
"loss": 0.0412,
|
| 387 |
+
"rewards/accuracies": 0.7406250238418579,
|
| 388 |
+
"rewards/chosen": -1.6363245248794556,
|
| 389 |
+
"rewards/margins": 1.1237612962722778,
|
| 390 |
+
"rewards/rejected": -2.7600855827331543,
|
| 391 |
+
"step": 240
|
| 392 |
+
},
|
| 393 |
+
{
|
| 394 |
+
"epoch": 0.57,
|
| 395 |
+
"learning_rate": 2.36123833901765e-07,
|
| 396 |
+
"logits/chosen": -2.0231640338897705,
|
| 397 |
+
"logits/rejected": -1.9347641468048096,
|
| 398 |
+
"logps/chosen": -503.3892517089844,
|
| 399 |
+
"logps/rejected": -482.2828674316406,
|
| 400 |
+
"loss": 0.0391,
|
| 401 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 402 |
+
"rewards/chosen": -1.7641624212265015,
|
| 403 |
+
"rewards/margins": 1.1827863454818726,
|
| 404 |
+
"rewards/rejected": -2.946949005126953,
|
| 405 |
+
"step": 250
|
| 406 |
+
},
|
| 407 |
+
{
|
| 408 |
+
"epoch": 0.59,
|
| 409 |
+
"learning_rate": 2.1638553071961704e-07,
|
| 410 |
+
"logits/chosen": -1.9997365474700928,
|
| 411 |
+
"logits/rejected": -1.919660210609436,
|
| 412 |
+
"logps/chosen": -549.3283081054688,
|
| 413 |
+
"logps/rejected": -506.68707275390625,
|
| 414 |
+
"loss": 0.0364,
|
| 415 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 416 |
+
"rewards/chosen": -1.842703104019165,
|
| 417 |
+
"rewards/margins": 1.4222638607025146,
|
| 418 |
+
"rewards/rejected": -3.2649669647216797,
|
| 419 |
+
"step": 260
|
| 420 |
+
},
|
| 421 |
+
{
|
| 422 |
+
"epoch": 0.61,
|
| 423 |
+
"learning_rate": 1.968586776117558e-07,
|
| 424 |
+
"logits/chosen": -2.023411273956299,
|
| 425 |
+
"logits/rejected": -1.9109961986541748,
|
| 426 |
+
"logps/chosen": -555.9891357421875,
|
| 427 |
+
"logps/rejected": -512.0736083984375,
|
| 428 |
+
"loss": 0.0427,
|
| 429 |
+
"rewards/accuracies": 0.765625,
|
| 430 |
+
"rewards/chosen": -1.7782520055770874,
|
| 431 |
+
"rewards/margins": 1.3554075956344604,
|
| 432 |
+
"rewards/rejected": -3.133659839630127,
|
| 433 |
+
"step": 270
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.63,
|
| 437 |
+
"learning_rate": 1.7766610723413684e-07,
|
| 438 |
+
"logits/chosen": -2.0062713623046875,
|
| 439 |
+
"logits/rejected": -1.9274126291275024,
|
| 440 |
+
"logps/chosen": -505.69110107421875,
|
| 441 |
+
"logps/rejected": -496.41864013671875,
|
| 442 |
+
"loss": 0.0421,
|
| 443 |
+
"rewards/accuracies": 0.7593749761581421,
|
| 444 |
+
"rewards/chosen": -1.7307058572769165,
|
| 445 |
+
"rewards/margins": 1.3143250942230225,
|
| 446 |
+
"rewards/rejected": -3.0450305938720703,
|
| 447 |
+
"step": 280
|
| 448 |
+
},
|
| 449 |
+
{
|
| 450 |
+
"epoch": 0.66,
|
| 451 |
+
"learning_rate": 1.589285494545514e-07,
|
| 452 |
+
"logits/chosen": -2.0006046295166016,
|
| 453 |
+
"logits/rejected": -1.9192641973495483,
|
| 454 |
+
"logps/chosen": -496.51055908203125,
|
| 455 |
+
"logps/rejected": -483.0226135253906,
|
| 456 |
+
"loss": 0.0435,
|
| 457 |
+
"rewards/accuracies": 0.762499988079071,
|
| 458 |
+
"rewards/chosen": -1.7974494695663452,
|
| 459 |
+
"rewards/margins": 1.1823097467422485,
|
| 460 |
+
"rewards/rejected": -2.979759454727173,
|
| 461 |
+
"step": 290
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.68,
|
| 465 |
+
"learning_rate": 1.4076387190766014e-07,
|
| 466 |
+
"logits/chosen": -1.9719831943511963,
|
| 467 |
+
"logits/rejected": -1.9076999425888062,
|
| 468 |
+
"logps/chosen": -481.85968017578125,
|
| 469 |
+
"logps/rejected": -481.77691650390625,
|
| 470 |
+
"loss": 0.046,
|
| 471 |
+
"rewards/accuracies": 0.753125011920929,
|
| 472 |
+
"rewards/chosen": -1.7264801263809204,
|
| 473 |
+
"rewards/margins": 1.1051851511001587,
|
| 474 |
+
"rewards/rejected": -2.831665277481079,
|
| 475 |
+
"step": 300
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 0.68,
|
| 479 |
+
"eval_logits/chosen": -1.991421103477478,
|
| 480 |
+
"eval_logits/rejected": -1.9484151601791382,
|
| 481 |
+
"eval_logps/chosen": -471.8780517578125,
|
| 482 |
+
"eval_logps/rejected": -544.56982421875,
|
| 483 |
+
"eval_loss": 0.055175185203552246,
|
| 484 |
+
"eval_rewards/accuracies": 0.71484375,
|
| 485 |
+
"eval_rewards/chosen": -2.1483840942382812,
|
| 486 |
+
"eval_rewards/margins": 0.7237809896469116,
|
| 487 |
+
"eval_rewards/rejected": -2.8721652030944824,
|
| 488 |
+
"eval_runtime": 53.342,
|
| 489 |
+
"eval_samples_per_second": 37.494,
|
| 490 |
+
"eval_steps_per_second": 0.6,
|
| 491 |
+
"step": 300
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.7,
|
| 495 |
+
"learning_rate": 1.232863385547543e-07,
|
| 496 |
+
"logits/chosen": -1.9471126794815063,
|
| 497 |
+
"logits/rejected": -1.8728282451629639,
|
| 498 |
+
"logps/chosen": -499.8617248535156,
|
| 499 |
+
"logps/rejected": -500.9742126464844,
|
| 500 |
+
"loss": 0.0415,
|
| 501 |
+
"rewards/accuracies": 0.768750011920929,
|
| 502 |
+
"rewards/chosen": -1.7383416891098022,
|
| 503 |
+
"rewards/margins": 1.4027913808822632,
|
| 504 |
+
"rewards/rejected": -3.1411330699920654,
|
| 505 |
+
"step": 310
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 0.72,
|
| 509 |
+
"learning_rate": 1.0660589091223854e-07,
|
| 510 |
+
"logits/chosen": -1.9511429071426392,
|
| 511 |
+
"logits/rejected": -1.8690084218978882,
|
| 512 |
+
"logps/chosen": -515.9614868164062,
|
| 513 |
+
"logps/rejected": -489.18853759765625,
|
| 514 |
+
"loss": 0.0424,
|
| 515 |
+
"rewards/accuracies": 0.765625,
|
| 516 |
+
"rewards/chosen": -1.7780258655548096,
|
| 517 |
+
"rewards/margins": 1.2824206352233887,
|
| 518 |
+
"rewards/rejected": -3.060446262359619,
|
| 519 |
+
"step": 320
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"epoch": 0.75,
|
| 523 |
+
"learning_rate": 9.082745647022797e-08,
|
| 524 |
+
"logits/chosen": -1.9991905689239502,
|
| 525 |
+
"logits/rejected": -1.9214942455291748,
|
| 526 |
+
"logps/chosen": -518.4646606445312,
|
| 527 |
+
"logps/rejected": -501.2618713378906,
|
| 528 |
+
"loss": 0.0439,
|
| 529 |
+
"rewards/accuracies": 0.768750011920929,
|
| 530 |
+
"rewards/chosen": -1.7493177652359009,
|
| 531 |
+
"rewards/margins": 1.3284794092178345,
|
| 532 |
+
"rewards/rejected": -3.0777969360351562,
|
| 533 |
+
"step": 330
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 0.77,
|
| 537 |
+
"learning_rate": 7.605028865161809e-08,
|
| 538 |
+
"logits/chosen": -1.994573950767517,
|
| 539 |
+
"logits/rejected": -1.9131485223770142,
|
| 540 |
+
"logps/chosen": -523.8150024414062,
|
| 541 |
+
"logps/rejected": -497.4427795410156,
|
| 542 |
+
"loss": 0.0414,
|
| 543 |
+
"rewards/accuracies": 0.778124988079071,
|
| 544 |
+
"rewards/chosen": -1.7702951431274414,
|
| 545 |
+
"rewards/margins": 1.2257777452468872,
|
| 546 |
+
"rewards/rejected": -2.996073007583618,
|
| 547 |
+
"step": 340
|
| 548 |
+
},
|
| 549 |
+
{
|
| 550 |
+
"epoch": 0.79,
|
| 551 |
+
"learning_rate": 6.236734246357947e-08,
|
| 552 |
+
"logits/chosen": -1.9538482427597046,
|
| 553 |
+
"logits/rejected": -1.849259614944458,
|
| 554 |
+
"logps/chosen": -493.80450439453125,
|
| 555 |
+
"logps/rejected": -493.0531311035156,
|
| 556 |
+
"loss": 0.0447,
|
| 557 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 558 |
+
"rewards/chosen": -1.6410820484161377,
|
| 559 |
+
"rewards/margins": 1.3493789434432983,
|
| 560 |
+
"rewards/rejected": -2.9904608726501465,
|
| 561 |
+
"step": 350
|
| 562 |
+
},
|
| 563 |
+
{
|
| 564 |
+
"epoch": 0.81,
|
| 565 |
+
"learning_rate": 4.986468976890992e-08,
|
| 566 |
+
"logits/chosen": -1.9808590412139893,
|
| 567 |
+
"logits/rejected": -1.8965733051300049,
|
| 568 |
+
"logps/chosen": -507.7171936035156,
|
| 569 |
+
"logps/rejected": -484.1426696777344,
|
| 570 |
+
"loss": 0.0423,
|
| 571 |
+
"rewards/accuracies": 0.768750011920929,
|
| 572 |
+
"rewards/chosen": -1.6426427364349365,
|
| 573 |
+
"rewards/margins": 1.3561906814575195,
|
| 574 |
+
"rewards/rejected": -2.998833417892456,
|
| 575 |
+
"step": 360
|
| 576 |
+
},
|
| 577 |
+
{
|
| 578 |
+
"epoch": 0.84,
|
| 579 |
+
"learning_rate": 3.8620977855448936e-08,
|
| 580 |
+
"logits/chosen": -1.9780423641204834,
|
| 581 |
+
"logits/rejected": -1.884387731552124,
|
| 582 |
+
"logps/chosen": -519.645751953125,
|
| 583 |
+
"logps/rejected": -487.0723571777344,
|
| 584 |
+
"loss": 0.0454,
|
| 585 |
+
"rewards/accuracies": 0.71875,
|
| 586 |
+
"rewards/chosen": -1.856579065322876,
|
| 587 |
+
"rewards/margins": 1.1135355234146118,
|
| 588 |
+
"rewards/rejected": -2.9701147079467773,
|
| 589 |
+
"step": 370
|
| 590 |
+
},
|
| 591 |
+
{
|
| 592 |
+
"epoch": 0.86,
|
| 593 |
+
"learning_rate": 2.8706934709395893e-08,
|
| 594 |
+
"logits/chosen": -1.99289870262146,
|
| 595 |
+
"logits/rejected": -1.9017293453216553,
|
| 596 |
+
"logps/chosen": -520.5479736328125,
|
| 597 |
+
"logps/rejected": -500.9642028808594,
|
| 598 |
+
"loss": 0.0424,
|
| 599 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 600 |
+
"rewards/chosen": -1.6913812160491943,
|
| 601 |
+
"rewards/margins": 1.3811187744140625,
|
| 602 |
+
"rewards/rejected": -3.072500228881836,
|
| 603 |
+
"step": 380
|
| 604 |
+
},
|
| 605 |
+
{
|
| 606 |
+
"epoch": 0.88,
|
| 607 |
+
"learning_rate": 2.0184924104583612e-08,
|
| 608 |
+
"logits/chosen": -1.988921880722046,
|
| 609 |
+
"logits/rejected": -1.891033411026001,
|
| 610 |
+
"logps/chosen": -521.970458984375,
|
| 611 |
+
"logps/rejected": -489.20501708984375,
|
| 612 |
+
"loss": 0.0425,
|
| 613 |
+
"rewards/accuracies": 0.796875,
|
| 614 |
+
"rewards/chosen": -1.6584064960479736,
|
| 615 |
+
"rewards/margins": 1.3673207759857178,
|
| 616 |
+
"rewards/rejected": -3.0257275104522705,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 0.91,
|
| 621 |
+
"learning_rate": 1.3108553306396263e-08,
|
| 622 |
+
"logits/chosen": -2.0064077377319336,
|
| 623 |
+
"logits/rejected": -1.9138851165771484,
|
| 624 |
+
"logps/chosen": -516.9049072265625,
|
| 625 |
+
"logps/rejected": -482.49859619140625,
|
| 626 |
+
"loss": 0.0439,
|
| 627 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 628 |
+
"rewards/chosen": -1.8265987634658813,
|
| 629 |
+
"rewards/margins": 1.1020760536193848,
|
| 630 |
+
"rewards/rejected": -2.9286749362945557,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 0.91,
|
| 635 |
+
"eval_logits/chosen": -1.9848593473434448,
|
| 636 |
+
"eval_logits/rejected": -1.94070303440094,
|
| 637 |
+
"eval_logps/chosen": -476.7722473144531,
|
| 638 |
+
"eval_logps/rejected": -549.7876586914062,
|
| 639 |
+
"eval_loss": 0.054434459656476974,
|
| 640 |
+
"eval_rewards/accuracies": 0.70703125,
|
| 641 |
+
"eval_rewards/chosen": -2.1973259449005127,
|
| 642 |
+
"eval_rewards/margins": 0.7270177602767944,
|
| 643 |
+
"eval_rewards/rejected": -2.9243435859680176,
|
| 644 |
+
"eval_runtime": 53.3853,
|
| 645 |
+
"eval_samples_per_second": 37.463,
|
| 646 |
+
"eval_steps_per_second": 0.599,
|
| 647 |
+
"step": 400
|
| 648 |
+
},
|
| 649 |
+
{
|
| 650 |
+
"epoch": 0.93,
|
| 651 |
+
"learning_rate": 7.522335858048705e-09,
|
| 652 |
+
"logits/chosen": -1.963323950767517,
|
| 653 |
+
"logits/rejected": -1.8955034017562866,
|
| 654 |
+
"logps/chosen": -518.6998901367188,
|
| 655 |
+
"logps/rejected": -522.4810791015625,
|
| 656 |
+
"loss": 0.0451,
|
| 657 |
+
"rewards/accuracies": 0.784375011920929,
|
| 658 |
+
"rewards/chosen": -1.8480857610702515,
|
| 659 |
+
"rewards/margins": 1.2641350030899048,
|
| 660 |
+
"rewards/rejected": -3.112220287322998,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.95,
|
| 665 |
+
"learning_rate": 3.4614115704533766e-09,
|
| 666 |
+
"logits/chosen": -1.9797197580337524,
|
| 667 |
+
"logits/rejected": -1.9050052165985107,
|
| 668 |
+
"logps/chosen": -495.5513610839844,
|
| 669 |
+
"logps/rejected": -489.12384033203125,
|
| 670 |
+
"loss": 0.0427,
|
| 671 |
+
"rewards/accuracies": 0.7718750238418579,
|
| 672 |
+
"rewards/chosen": -1.9035917520523071,
|
| 673 |
+
"rewards/margins": 1.1141357421875,
|
| 674 |
+
"rewards/rejected": -3.0177273750305176,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 0.97,
|
| 679 |
+
"learning_rate": 9.513254770636137e-10,
|
| 680 |
+
"logits/chosen": -1.9577579498291016,
|
| 681 |
+
"logits/rejected": -1.8832132816314697,
|
| 682 |
+
"logps/chosen": -509.35406494140625,
|
| 683 |
+
"logps/rejected": -482.31640625,
|
| 684 |
+
"loss": 0.0437,
|
| 685 |
+
"rewards/accuracies": 0.737500011920929,
|
| 686 |
+
"rewards/chosen": -1.8531081676483154,
|
| 687 |
+
"rewards/margins": 1.0920084714889526,
|
| 688 |
+
"rewards/rejected": -2.9451169967651367,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 1.0,
|
| 693 |
+
"learning_rate": 7.867144166728844e-12,
|
| 694 |
+
"logits/chosen": -1.9990612268447876,
|
| 695 |
+
"logits/rejected": -1.9326860904693604,
|
| 696 |
+
"logps/chosen": -519.8091430664062,
|
| 697 |
+
"logps/rejected": -503.39385986328125,
|
| 698 |
+
"loss": 0.0409,
|
| 699 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 700 |
+
"rewards/chosen": -1.7259622812271118,
|
| 701 |
+
"rewards/margins": 1.2851760387420654,
|
| 702 |
+
"rewards/rejected": -3.0111382007598877,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
{
|
| 706 |
"epoch": 1.0,
|
| 707 |
+
"step": 441,
|
| 708 |
"total_flos": 0.0,
|
| 709 |
+
"train_loss": 0.07998780592802972,
|
| 710 |
+
"train_runtime": 7378.9712,
|
| 711 |
+
"train_samples_per_second": 15.318,
|
| 712 |
+
"train_steps_per_second": 0.06
|
| 713 |
}
|
| 714 |
],
|
| 715 |
"logging_steps": 10,
|
| 716 |
+
"max_steps": 441,
|
|
|
|
| 717 |
"num_train_epochs": 1,
|
| 718 |
"save_steps": 100,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 719 |
"total_flos": 0.0,
|
|
|
|
| 720 |
"trial_name": null,
|
| 721 |
"trial_params": null
|
| 722 |
}
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 5944
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d1bc12ab9a777f3e948ad4888c0b78fbb69873a1cca9e5f6148ba6edfc1e8386
|
| 3 |
size 5944
|