Model save
Browse files- README.md +18 -3
- adapter_model.safetensors +1 -1
- all_results.json +20 -5
- eval_results.json +14 -14
- runs/Mar25_22-32-18_uclaml04.cs.ucla.edu/events.out.tfevents.1711431194.uclaml04.cs.ucla.edu.282670.0 +2 -2
- runs/Mar25_22-32-18_uclaml04.cs.ucla.edu/events.out.tfevents.1711456098.uclaml04.cs.ucla.edu.282670.1 +3 -0
- train_results.json +5 -5
- trainer_state.json +789 -15
README.md
CHANGED
|
@@ -14,6 +14,18 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 14 |
# zephyr-ds
|
| 15 |
|
| 16 |
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
|
| 18 |
## Model description
|
| 19 |
|
|
@@ -32,14 +44,14 @@ More information needed
|
|
| 32 |
### Training hyperparameters
|
| 33 |
|
| 34 |
The following hyperparameters were used during training:
|
| 35 |
-
- learning_rate:
|
| 36 |
- train_batch_size: 4
|
| 37 |
- eval_batch_size: 4
|
| 38 |
- seed: 42
|
| 39 |
- distributed_type: multi-GPU
|
| 40 |
- num_devices: 4
|
| 41 |
-
- gradient_accumulation_steps:
|
| 42 |
-
- total_train_batch_size:
|
| 43 |
- total_eval_batch_size: 16
|
| 44 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 45 |
- lr_scheduler_type: linear
|
|
@@ -48,6 +60,9 @@ The following hyperparameters were used during training:
|
|
| 48 |
|
| 49 |
### Training results
|
| 50 |
|
|
|
|
|
|
|
|
|
|
| 51 |
|
| 52 |
|
| 53 |
### Framework versions
|
|
|
|
| 14 |
# zephyr-ds
|
| 15 |
|
| 16 |
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
|
| 17 |
+
It achieves the following results on the evaluation set:
|
| 18 |
+
- Loss: 0.5199
|
| 19 |
+
- Rewards/chosen: -0.1238
|
| 20 |
+
- Rewards/rejected: -1.1258
|
| 21 |
+
- Rewards/accuracies: 0.7300
|
| 22 |
+
- Rewards/margins: 1.0020
|
| 23 |
+
- Logps/rejected: -270.5574
|
| 24 |
+
- Logps/chosen: -285.4951
|
| 25 |
+
- Logits/rejected: -2.8178
|
| 26 |
+
- Logits/chosen: -2.8221
|
| 27 |
+
- Use Label: 0.0
|
| 28 |
+
- Pred Label: 0.0
|
| 29 |
|
| 30 |
## Model description
|
| 31 |
|
|
|
|
| 44 |
### Training hyperparameters
|
| 45 |
|
| 46 |
The following hyperparameters were used during training:
|
| 47 |
+
- learning_rate: 5e-05
|
| 48 |
- train_batch_size: 4
|
| 49 |
- eval_batch_size: 4
|
| 50 |
- seed: 42
|
| 51 |
- distributed_type: multi-GPU
|
| 52 |
- num_devices: 4
|
| 53 |
+
- gradient_accumulation_steps: 8
|
| 54 |
+
- total_train_batch_size: 128
|
| 55 |
- total_eval_batch_size: 16
|
| 56 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 57 |
- lr_scheduler_type: linear
|
|
|
|
| 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 | Use Label | Pred Label |
|
| 64 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:---------:|:----------:|
|
| 65 |
+
| 0.4951 | 1.0 | 477 | 0.5199 | -0.1238 | -1.1258 | 0.7300 | 1.0020 | -270.5574 | -285.4951 | -2.8178 | -2.8221 | 0.0 | 0.0 |
|
| 66 |
|
| 67 |
|
| 68 |
### Framework versions
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 218138576
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:113769ca75081b006e5afbb71ddc0c5d371f9107f8efec1c029f3129a997de8f
|
| 3 |
size 218138576
|
all_results.json
CHANGED
|
@@ -1,8 +1,23 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"
|
| 4 |
-
"
|
| 5 |
-
"
|
| 6 |
-
"
|
| 7 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"eval_logits/chosen": -2.822110891342163,
|
| 4 |
+
"eval_logits/rejected": -2.8178136348724365,
|
| 5 |
+
"eval_logps/chosen": -285.4951171875,
|
| 6 |
+
"eval_logps/rejected": -270.55743408203125,
|
| 7 |
+
"eval_loss": 0.5198934078216553,
|
| 8 |
+
"eval_pred_label": 0.0,
|
| 9 |
+
"eval_rewards/accuracies": 0.7300000190734863,
|
| 10 |
+
"eval_rewards/chosen": -0.12377375364303589,
|
| 11 |
+
"eval_rewards/margins": 1.001997470855713,
|
| 12 |
+
"eval_rewards/rejected": -1.1257712841033936,
|
| 13 |
+
"eval_runtime": 453.8128,
|
| 14 |
+
"eval_samples": 2000,
|
| 15 |
+
"eval_samples_per_second": 4.407,
|
| 16 |
+
"eval_steps_per_second": 0.275,
|
| 17 |
+
"eval_use_label": 0.0,
|
| 18 |
+
"train_loss": 0.5321606655040877,
|
| 19 |
+
"train_runtime": 24451.2028,
|
| 20 |
+
"train_samples": 61135,
|
| 21 |
+
"train_samples_per_second": 2.5,
|
| 22 |
+
"train_steps_per_second": 0.02
|
| 23 |
}
|
eval_results.json
CHANGED
|
@@ -1,18 +1,18 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"eval_logits/chosen": -2.
|
| 4 |
-
"eval_logits/rejected": -2.
|
| 5 |
-
"eval_logps/chosen": -
|
| 6 |
-
"eval_logps/rejected": -
|
| 7 |
-
"eval_loss": 0.
|
| 8 |
-
"eval_pred_label":
|
| 9 |
-
"eval_rewards/accuracies": 0.
|
| 10 |
-
"eval_rewards/chosen": -
|
| 11 |
-
"eval_rewards/margins":
|
| 12 |
-
"eval_rewards/rejected": -
|
| 13 |
-
"eval_runtime":
|
| 14 |
"eval_samples": 2000,
|
| 15 |
-
"eval_samples_per_second":
|
| 16 |
-
"eval_steps_per_second": 0.
|
| 17 |
-
"eval_use_label":
|
| 18 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"eval_logits/chosen": -2.822110891342163,
|
| 4 |
+
"eval_logits/rejected": -2.8178136348724365,
|
| 5 |
+
"eval_logps/chosen": -285.4951171875,
|
| 6 |
+
"eval_logps/rejected": -270.55743408203125,
|
| 7 |
+
"eval_loss": 0.5198934078216553,
|
| 8 |
+
"eval_pred_label": 0.0,
|
| 9 |
+
"eval_rewards/accuracies": 0.7300000190734863,
|
| 10 |
+
"eval_rewards/chosen": -0.12377375364303589,
|
| 11 |
+
"eval_rewards/margins": 1.001997470855713,
|
| 12 |
+
"eval_rewards/rejected": -1.1257712841033936,
|
| 13 |
+
"eval_runtime": 453.8128,
|
| 14 |
"eval_samples": 2000,
|
| 15 |
+
"eval_samples_per_second": 4.407,
|
| 16 |
+
"eval_steps_per_second": 0.275,
|
| 17 |
+
"eval_use_label": 0.0
|
| 18 |
}
|
runs/Mar25_22-32-18_uclaml04.cs.ucla.edu/events.out.tfevents.1711431194.uclaml04.cs.ucla.edu.282670.0
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8bd6d8d978f8970690763999b8c0c13245b54fa582a289d7997c052e40bbf6a4
|
| 3 |
+
size 41127
|
runs/Mar25_22-32-18_uclaml04.cs.ucla.edu/events.out.tfevents.1711456098.uclaml04.cs.ucla.edu.282670.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9902476a403d5b482d88dcbfb9fac9a6a9bda2c52b4c63ce321cf0b3e5e6b684
|
| 3 |
+
size 935
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
-
"train_samples":
|
| 6 |
-
"train_samples_per_second": 2.
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.5321606655040877,
|
| 4 |
+
"train_runtime": 24451.2028,
|
| 5 |
+
"train_samples": 61135,
|
| 6 |
+
"train_samples_per_second": 2.5,
|
| 7 |
+
"train_steps_per_second": 0.02
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -1,39 +1,813 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch":
|
| 5 |
-
"eval_steps":
|
| 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 |
-
"epoch":
|
| 13 |
-
"learning_rate":
|
| 14 |
-
"
|
| 15 |
-
"
|
|
|
|
|
|
|
| 16 |
"loss": 0.6931,
|
|
|
|
| 17 |
"rewards/accuracies": 0.0,
|
| 18 |
"rewards/chosen": 0.0,
|
| 19 |
"rewards/margins": 0.0,
|
| 20 |
"rewards/rejected": 0.0,
|
| 21 |
-
"step": 1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
},
|
| 23 |
{
|
| 24 |
"epoch": 1.0,
|
| 25 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
"total_flos": 0.0,
|
| 27 |
-
"train_loss": 0.
|
| 28 |
-
"train_runtime":
|
| 29 |
-
"train_samples_per_second": 2.
|
| 30 |
-
"train_steps_per_second": 0.
|
| 31 |
}
|
| 32 |
],
|
| 33 |
"logging_steps": 10,
|
| 34 |
-
"max_steps":
|
| 35 |
"num_train_epochs": 1,
|
| 36 |
-
"save_steps":
|
| 37 |
"total_flos": 0.0,
|
| 38 |
"trial_name": null,
|
| 39 |
"trial_params": null
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.998691442030882,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 477,
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
| 10 |
"log_history": [
|
| 11 |
{
|
| 12 |
+
"epoch": 0.0,
|
| 13 |
+
"learning_rate": 1.0416666666666667e-06,
|
| 14 |
+
"logits/chosen": -2.9089105129241943,
|
| 15 |
+
"logits/rejected": -2.8982176780700684,
|
| 16 |
+
"logps/chosen": -328.48699951171875,
|
| 17 |
+
"logps/rejected": -294.0901794433594,
|
| 18 |
"loss": 0.6931,
|
| 19 |
+
"pred_label": 0.0,
|
| 20 |
"rewards/accuracies": 0.0,
|
| 21 |
"rewards/chosen": 0.0,
|
| 22 |
"rewards/margins": 0.0,
|
| 23 |
"rewards/rejected": 0.0,
|
| 24 |
+
"step": 1,
|
| 25 |
+
"use_label": 0.0
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 0.02,
|
| 29 |
+
"learning_rate": 1.0416666666666668e-05,
|
| 30 |
+
"logits/chosen": -2.802924633026123,
|
| 31 |
+
"logits/rejected": -2.845370292663574,
|
| 32 |
+
"logps/chosen": -274.2305908203125,
|
| 33 |
+
"logps/rejected": -257.4792785644531,
|
| 34 |
+
"loss": 0.6894,
|
| 35 |
+
"pred_label": 0.0,
|
| 36 |
+
"rewards/accuracies": 0.4965277910232544,
|
| 37 |
+
"rewards/chosen": 0.0021643945947289467,
|
| 38 |
+
"rewards/margins": 0.006878577638417482,
|
| 39 |
+
"rewards/rejected": -0.004714183043688536,
|
| 40 |
+
"step": 10,
|
| 41 |
+
"use_label": 0.0
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"epoch": 0.04,
|
| 45 |
+
"learning_rate": 2.0833333333333336e-05,
|
| 46 |
+
"logits/chosen": -2.8522541522979736,
|
| 47 |
+
"logits/rejected": -2.842360019683838,
|
| 48 |
+
"logps/chosen": -279.8498840332031,
|
| 49 |
+
"logps/rejected": -264.8490905761719,
|
| 50 |
+
"loss": 0.6557,
|
| 51 |
+
"pred_label": 0.0,
|
| 52 |
+
"rewards/accuracies": 0.65625,
|
| 53 |
+
"rewards/chosen": 0.02565639279782772,
|
| 54 |
+
"rewards/margins": 0.10996748507022858,
|
| 55 |
+
"rewards/rejected": -0.0843110978603363,
|
| 56 |
+
"step": 20,
|
| 57 |
+
"use_label": 0.0
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"epoch": 0.06,
|
| 61 |
+
"learning_rate": 3.125e-05,
|
| 62 |
+
"logits/chosen": -2.8415942192077637,
|
| 63 |
+
"logits/rejected": -2.829390525817871,
|
| 64 |
+
"logps/chosen": -283.43377685546875,
|
| 65 |
+
"logps/rejected": -275.16021728515625,
|
| 66 |
+
"loss": 0.6152,
|
| 67 |
+
"pred_label": 0.0,
|
| 68 |
+
"rewards/accuracies": 0.668749988079071,
|
| 69 |
+
"rewards/chosen": -0.017193807289004326,
|
| 70 |
+
"rewards/margins": 0.2977373003959656,
|
| 71 |
+
"rewards/rejected": -0.31493109464645386,
|
| 72 |
+
"step": 30,
|
| 73 |
+
"use_label": 0.0
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"epoch": 0.08,
|
| 77 |
+
"learning_rate": 4.166666666666667e-05,
|
| 78 |
+
"logits/chosen": -2.808344841003418,
|
| 79 |
+
"logits/rejected": -2.8009369373321533,
|
| 80 |
+
"logps/chosen": -299.697998046875,
|
| 81 |
+
"logps/rejected": -281.28240966796875,
|
| 82 |
+
"loss": 0.5509,
|
| 83 |
+
"pred_label": 0.0,
|
| 84 |
+
"rewards/accuracies": 0.734375,
|
| 85 |
+
"rewards/chosen": 0.011143045499920845,
|
| 86 |
+
"rewards/margins": 0.5404817461967468,
|
| 87 |
+
"rewards/rejected": -0.5293387174606323,
|
| 88 |
+
"step": 40,
|
| 89 |
+
"use_label": 0.0
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"epoch": 0.1,
|
| 93 |
+
"learning_rate": 4.976689976689977e-05,
|
| 94 |
+
"logits/chosen": -2.735532283782959,
|
| 95 |
+
"logits/rejected": -2.762310743331909,
|
| 96 |
+
"logps/chosen": -284.2992248535156,
|
| 97 |
+
"logps/rejected": -285.3406982421875,
|
| 98 |
+
"loss": 0.5757,
|
| 99 |
+
"pred_label": 0.0,
|
| 100 |
+
"rewards/accuracies": 0.731249988079071,
|
| 101 |
+
"rewards/chosen": 0.06582433730363846,
|
| 102 |
+
"rewards/margins": 0.6093484163284302,
|
| 103 |
+
"rewards/rejected": -0.5435240864753723,
|
| 104 |
+
"step": 50,
|
| 105 |
+
"use_label": 0.0
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"epoch": 0.13,
|
| 109 |
+
"learning_rate": 4.86013986013986e-05,
|
| 110 |
+
"logits/chosen": -2.7604079246520996,
|
| 111 |
+
"logits/rejected": -2.749760389328003,
|
| 112 |
+
"logps/chosen": -271.2710876464844,
|
| 113 |
+
"logps/rejected": -253.1558074951172,
|
| 114 |
+
"loss": 0.5778,
|
| 115 |
+
"pred_label": 0.0,
|
| 116 |
+
"rewards/accuracies": 0.690625011920929,
|
| 117 |
+
"rewards/chosen": 0.08226754516363144,
|
| 118 |
+
"rewards/margins": 0.5549899935722351,
|
| 119 |
+
"rewards/rejected": -0.47272244095802307,
|
| 120 |
+
"step": 60,
|
| 121 |
+
"use_label": 0.0
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.15,
|
| 125 |
+
"learning_rate": 4.7435897435897435e-05,
|
| 126 |
+
"logits/chosen": -2.825266122817993,
|
| 127 |
+
"logits/rejected": -2.8063371181488037,
|
| 128 |
+
"logps/chosen": -282.8874816894531,
|
| 129 |
+
"logps/rejected": -260.37786865234375,
|
| 130 |
+
"loss": 0.5591,
|
| 131 |
+
"pred_label": 0.0,
|
| 132 |
+
"rewards/accuracies": 0.6968749761581421,
|
| 133 |
+
"rewards/chosen": -0.020037399604916573,
|
| 134 |
+
"rewards/margins": 0.611237108707428,
|
| 135 |
+
"rewards/rejected": -0.6312744617462158,
|
| 136 |
+
"step": 70,
|
| 137 |
+
"use_label": 0.0
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 0.17,
|
| 141 |
+
"learning_rate": 4.6270396270396274e-05,
|
| 142 |
+
"logits/chosen": -2.777280807495117,
|
| 143 |
+
"logits/rejected": -2.760341167449951,
|
| 144 |
+
"logps/chosen": -278.3568420410156,
|
| 145 |
+
"logps/rejected": -281.4912109375,
|
| 146 |
+
"loss": 0.5484,
|
| 147 |
+
"pred_label": 0.0,
|
| 148 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 149 |
+
"rewards/chosen": 0.02771860361099243,
|
| 150 |
+
"rewards/margins": 0.6373428702354431,
|
| 151 |
+
"rewards/rejected": -0.6096242666244507,
|
| 152 |
+
"step": 80,
|
| 153 |
+
"use_label": 0.0
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"epoch": 0.19,
|
| 157 |
+
"learning_rate": 4.5104895104895105e-05,
|
| 158 |
+
"logits/chosen": -2.7895703315734863,
|
| 159 |
+
"logits/rejected": -2.7826099395751953,
|
| 160 |
+
"logps/chosen": -274.6077575683594,
|
| 161 |
+
"logps/rejected": -265.6002197265625,
|
| 162 |
+
"loss": 0.5476,
|
| 163 |
+
"pred_label": 0.0,
|
| 164 |
+
"rewards/accuracies": 0.71875,
|
| 165 |
+
"rewards/chosen": 0.15948975086212158,
|
| 166 |
+
"rewards/margins": 0.703137218952179,
|
| 167 |
+
"rewards/rejected": -0.5436475276947021,
|
| 168 |
+
"step": 90,
|
| 169 |
+
"use_label": 0.0
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"epoch": 0.21,
|
| 173 |
+
"learning_rate": 4.3939393939393944e-05,
|
| 174 |
+
"logits/chosen": -2.842555046081543,
|
| 175 |
+
"logits/rejected": -2.8219830989837646,
|
| 176 |
+
"logps/chosen": -282.922607421875,
|
| 177 |
+
"logps/rejected": -272.0762939453125,
|
| 178 |
+
"loss": 0.5218,
|
| 179 |
+
"pred_label": 0.0,
|
| 180 |
+
"rewards/accuracies": 0.753125011920929,
|
| 181 |
+
"rewards/chosen": 0.1369757205247879,
|
| 182 |
+
"rewards/margins": 0.8804744482040405,
|
| 183 |
+
"rewards/rejected": -0.7434987425804138,
|
| 184 |
+
"step": 100,
|
| 185 |
+
"use_label": 0.0
|
| 186 |
+
},
|
| 187 |
+
{
|
| 188 |
+
"epoch": 0.23,
|
| 189 |
+
"learning_rate": 4.2773892773892776e-05,
|
| 190 |
+
"logits/chosen": -2.7973036766052246,
|
| 191 |
+
"logits/rejected": -2.8159468173980713,
|
| 192 |
+
"logps/chosen": -287.54254150390625,
|
| 193 |
+
"logps/rejected": -269.5066223144531,
|
| 194 |
+
"loss": 0.5455,
|
| 195 |
+
"pred_label": 0.0,
|
| 196 |
+
"rewards/accuracies": 0.734375,
|
| 197 |
+
"rewards/chosen": 0.057723164558410645,
|
| 198 |
+
"rewards/margins": 0.7660588622093201,
|
| 199 |
+
"rewards/rejected": -0.7083355784416199,
|
| 200 |
+
"step": 110,
|
| 201 |
+
"use_label": 0.0
|
| 202 |
+
},
|
| 203 |
+
{
|
| 204 |
+
"epoch": 0.25,
|
| 205 |
+
"learning_rate": 4.1608391608391614e-05,
|
| 206 |
+
"logits/chosen": -2.7736809253692627,
|
| 207 |
+
"logits/rejected": -2.7753469944000244,
|
| 208 |
+
"logps/chosen": -292.09246826171875,
|
| 209 |
+
"logps/rejected": -252.3135986328125,
|
| 210 |
+
"loss": 0.5353,
|
| 211 |
+
"pred_label": 0.0,
|
| 212 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 213 |
+
"rewards/chosen": -0.07172087579965591,
|
| 214 |
+
"rewards/margins": 0.8031927347183228,
|
| 215 |
+
"rewards/rejected": -0.8749135732650757,
|
| 216 |
+
"step": 120,
|
| 217 |
+
"use_label": 0.0
|
| 218 |
+
},
|
| 219 |
+
{
|
| 220 |
+
"epoch": 0.27,
|
| 221 |
+
"learning_rate": 4.0442890442890446e-05,
|
| 222 |
+
"logits/chosen": -2.7675487995147705,
|
| 223 |
+
"logits/rejected": -2.761944532394409,
|
| 224 |
+
"logps/chosen": -271.2415771484375,
|
| 225 |
+
"logps/rejected": -269.3936767578125,
|
| 226 |
+
"loss": 0.5413,
|
| 227 |
+
"pred_label": 0.0,
|
| 228 |
+
"rewards/accuracies": 0.703125,
|
| 229 |
+
"rewards/chosen": -0.16168196499347687,
|
| 230 |
+
"rewards/margins": 0.7922881841659546,
|
| 231 |
+
"rewards/rejected": -0.9539702534675598,
|
| 232 |
+
"step": 130,
|
| 233 |
+
"use_label": 0.0
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 0.29,
|
| 237 |
+
"learning_rate": 3.9277389277389285e-05,
|
| 238 |
+
"logits/chosen": -2.821808338165283,
|
| 239 |
+
"logits/rejected": -2.828434467315674,
|
| 240 |
+
"logps/chosen": -293.92376708984375,
|
| 241 |
+
"logps/rejected": -278.75469970703125,
|
| 242 |
+
"loss": 0.5249,
|
| 243 |
+
"pred_label": 0.0,
|
| 244 |
+
"rewards/accuracies": 0.7593749761581421,
|
| 245 |
+
"rewards/chosen": -0.10061223804950714,
|
| 246 |
+
"rewards/margins": 0.9257994890213013,
|
| 247 |
+
"rewards/rejected": -1.026411771774292,
|
| 248 |
+
"step": 140,
|
| 249 |
+
"use_label": 0.0
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.31,
|
| 253 |
+
"learning_rate": 3.811188811188811e-05,
|
| 254 |
+
"logits/chosen": -2.8263511657714844,
|
| 255 |
+
"logits/rejected": -2.803342819213867,
|
| 256 |
+
"logps/chosen": -280.3300476074219,
|
| 257 |
+
"logps/rejected": -244.6310577392578,
|
| 258 |
+
"loss": 0.5312,
|
| 259 |
+
"pred_label": 0.0,
|
| 260 |
+
"rewards/accuracies": 0.721875011920929,
|
| 261 |
+
"rewards/chosen": -0.2029426395893097,
|
| 262 |
+
"rewards/margins": 0.7571284174919128,
|
| 263 |
+
"rewards/rejected": -0.9600710868835449,
|
| 264 |
+
"step": 150,
|
| 265 |
+
"use_label": 0.0
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.33,
|
| 269 |
+
"learning_rate": 3.694638694638695e-05,
|
| 270 |
+
"logits/chosen": -2.7946958541870117,
|
| 271 |
+
"logits/rejected": -2.748753309249878,
|
| 272 |
+
"logps/chosen": -282.52154541015625,
|
| 273 |
+
"logps/rejected": -296.4395446777344,
|
| 274 |
+
"loss": 0.5058,
|
| 275 |
+
"pred_label": 0.0,
|
| 276 |
+
"rewards/accuracies": 0.762499988079071,
|
| 277 |
+
"rewards/chosen": -0.10657407343387604,
|
| 278 |
+
"rewards/margins": 1.0698082447052002,
|
| 279 |
+
"rewards/rejected": -1.176382303237915,
|
| 280 |
+
"step": 160,
|
| 281 |
+
"use_label": 0.0
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"epoch": 0.36,
|
| 285 |
+
"learning_rate": 3.578088578088578e-05,
|
| 286 |
+
"logits/chosen": -2.7990992069244385,
|
| 287 |
+
"logits/rejected": -2.7839698791503906,
|
| 288 |
+
"logps/chosen": -280.283447265625,
|
| 289 |
+
"logps/rejected": -245.1365966796875,
|
| 290 |
+
"loss": 0.5741,
|
| 291 |
+
"pred_label": 0.0,
|
| 292 |
+
"rewards/accuracies": 0.675000011920929,
|
| 293 |
+
"rewards/chosen": -0.0219185221940279,
|
| 294 |
+
"rewards/margins": 0.7351241707801819,
|
| 295 |
+
"rewards/rejected": -0.7570425868034363,
|
| 296 |
+
"step": 170,
|
| 297 |
+
"use_label": 0.0
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.38,
|
| 301 |
+
"learning_rate": 3.461538461538462e-05,
|
| 302 |
+
"logits/chosen": -2.8004555702209473,
|
| 303 |
+
"logits/rejected": -2.80672025680542,
|
| 304 |
+
"logps/chosen": -286.1470031738281,
|
| 305 |
+
"logps/rejected": -280.5009765625,
|
| 306 |
+
"loss": 0.5354,
|
| 307 |
+
"pred_label": 0.0,
|
| 308 |
+
"rewards/accuracies": 0.768750011920929,
|
| 309 |
+
"rewards/chosen": -0.1152573823928833,
|
| 310 |
+
"rewards/margins": 0.79168301820755,
|
| 311 |
+
"rewards/rejected": -0.9069403409957886,
|
| 312 |
+
"step": 180,
|
| 313 |
+
"use_label": 0.0
|
| 314 |
+
},
|
| 315 |
+
{
|
| 316 |
+
"epoch": 0.4,
|
| 317 |
+
"learning_rate": 3.344988344988345e-05,
|
| 318 |
+
"logits/chosen": -2.820003032684326,
|
| 319 |
+
"logits/rejected": -2.789689302444458,
|
| 320 |
+
"logps/chosen": -281.3421936035156,
|
| 321 |
+
"logps/rejected": -278.36883544921875,
|
| 322 |
+
"loss": 0.5092,
|
| 323 |
+
"pred_label": 0.0,
|
| 324 |
+
"rewards/accuracies": 0.762499988079071,
|
| 325 |
+
"rewards/chosen": -0.06256841123104095,
|
| 326 |
+
"rewards/margins": 0.9592651128768921,
|
| 327 |
+
"rewards/rejected": -1.0218336582183838,
|
| 328 |
+
"step": 190,
|
| 329 |
+
"use_label": 0.0
|
| 330 |
+
},
|
| 331 |
+
{
|
| 332 |
+
"epoch": 0.42,
|
| 333 |
+
"learning_rate": 3.228438228438229e-05,
|
| 334 |
+
"logits/chosen": -2.7884361743927,
|
| 335 |
+
"logits/rejected": -2.782519578933716,
|
| 336 |
+
"logps/chosen": -294.6650390625,
|
| 337 |
+
"logps/rejected": -261.40435791015625,
|
| 338 |
+
"loss": 0.5052,
|
| 339 |
+
"pred_label": 0.0,
|
| 340 |
+
"rewards/accuracies": 0.734375,
|
| 341 |
+
"rewards/chosen": 0.02564082108438015,
|
| 342 |
+
"rewards/margins": 0.9850121736526489,
|
| 343 |
+
"rewards/rejected": -0.9593712687492371,
|
| 344 |
+
"step": 200,
|
| 345 |
+
"use_label": 0.0
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"epoch": 0.44,
|
| 349 |
+
"learning_rate": 3.111888111888112e-05,
|
| 350 |
+
"logits/chosen": -2.797853946685791,
|
| 351 |
+
"logits/rejected": -2.7864465713500977,
|
| 352 |
+
"logps/chosen": -269.8060607910156,
|
| 353 |
+
"logps/rejected": -262.01422119140625,
|
| 354 |
+
"loss": 0.5258,
|
| 355 |
+
"pred_label": 0.0,
|
| 356 |
+
"rewards/accuracies": 0.706250011920929,
|
| 357 |
+
"rewards/chosen": -0.046703118830919266,
|
| 358 |
+
"rewards/margins": 0.7715562582015991,
|
| 359 |
+
"rewards/rejected": -0.8182594180107117,
|
| 360 |
+
"step": 210,
|
| 361 |
+
"use_label": 0.0
|
| 362 |
+
},
|
| 363 |
+
{
|
| 364 |
+
"epoch": 0.46,
|
| 365 |
+
"learning_rate": 2.9953379953379956e-05,
|
| 366 |
+
"logits/chosen": -2.8102259635925293,
|
| 367 |
+
"logits/rejected": -2.813842296600342,
|
| 368 |
+
"logps/chosen": -280.2008972167969,
|
| 369 |
+
"logps/rejected": -262.47161865234375,
|
| 370 |
+
"loss": 0.5326,
|
| 371 |
+
"pred_label": 0.0,
|
| 372 |
+
"rewards/accuracies": 0.7093750238418579,
|
| 373 |
+
"rewards/chosen": 0.038443028926849365,
|
| 374 |
+
"rewards/margins": 0.6599202156066895,
|
| 375 |
+
"rewards/rejected": -0.6214772462844849,
|
| 376 |
+
"step": 220,
|
| 377 |
+
"use_label": 0.0
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 0.48,
|
| 381 |
+
"learning_rate": 2.878787878787879e-05,
|
| 382 |
+
"logits/chosen": -2.7520573139190674,
|
| 383 |
+
"logits/rejected": -2.744856357574463,
|
| 384 |
+
"logps/chosen": -265.61956787109375,
|
| 385 |
+
"logps/rejected": -256.87115478515625,
|
| 386 |
+
"loss": 0.5328,
|
| 387 |
+
"pred_label": 0.0,
|
| 388 |
+
"rewards/accuracies": 0.690625011920929,
|
| 389 |
+
"rewards/chosen": 0.014087711460888386,
|
| 390 |
+
"rewards/margins": 0.7686548233032227,
|
| 391 |
+
"rewards/rejected": -0.7545671463012695,
|
| 392 |
+
"step": 230,
|
| 393 |
+
"use_label": 0.0
|
| 394 |
+
},
|
| 395 |
+
{
|
| 396 |
+
"epoch": 0.5,
|
| 397 |
+
"learning_rate": 2.762237762237762e-05,
|
| 398 |
+
"logits/chosen": -2.7545018196105957,
|
| 399 |
+
"logits/rejected": -2.730861186981201,
|
| 400 |
+
"logps/chosen": -256.40753173828125,
|
| 401 |
+
"logps/rejected": -260.3459167480469,
|
| 402 |
+
"loss": 0.5315,
|
| 403 |
+
"pred_label": 0.0,
|
| 404 |
+
"rewards/accuracies": 0.7093750238418579,
|
| 405 |
+
"rewards/chosen": -0.0070175291039049625,
|
| 406 |
+
"rewards/margins": 0.7594768404960632,
|
| 407 |
+
"rewards/rejected": -0.7664943933486938,
|
| 408 |
+
"step": 240,
|
| 409 |
+
"use_label": 0.0
|
| 410 |
+
},
|
| 411 |
+
{
|
| 412 |
+
"epoch": 0.52,
|
| 413 |
+
"learning_rate": 2.6456876456876455e-05,
|
| 414 |
+
"logits/chosen": -2.8026585578918457,
|
| 415 |
+
"logits/rejected": -2.778594493865967,
|
| 416 |
+
"logps/chosen": -277.6792907714844,
|
| 417 |
+
"logps/rejected": -256.2415771484375,
|
| 418 |
+
"loss": 0.5158,
|
| 419 |
+
"pred_label": 0.0,
|
| 420 |
+
"rewards/accuracies": 0.7406250238418579,
|
| 421 |
+
"rewards/chosen": 0.056344062089920044,
|
| 422 |
+
"rewards/margins": 0.8976815938949585,
|
| 423 |
+
"rewards/rejected": -0.8413375616073608,
|
| 424 |
+
"step": 250,
|
| 425 |
+
"use_label": 0.0
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"epoch": 0.54,
|
| 429 |
+
"learning_rate": 2.529137529137529e-05,
|
| 430 |
+
"logits/chosen": -2.8391366004943848,
|
| 431 |
+
"logits/rejected": -2.8148343563079834,
|
| 432 |
+
"logps/chosen": -284.32781982421875,
|
| 433 |
+
"logps/rejected": -262.647705078125,
|
| 434 |
+
"loss": 0.4916,
|
| 435 |
+
"pred_label": 0.0,
|
| 436 |
+
"rewards/accuracies": 0.7406250238418579,
|
| 437 |
+
"rewards/chosen": -0.06821730732917786,
|
| 438 |
+
"rewards/margins": 1.0360552072525024,
|
| 439 |
+
"rewards/rejected": -1.104272484779358,
|
| 440 |
+
"step": 260,
|
| 441 |
+
"use_label": 0.0
|
| 442 |
+
},
|
| 443 |
+
{
|
| 444 |
+
"epoch": 0.57,
|
| 445 |
+
"learning_rate": 2.4125874125874125e-05,
|
| 446 |
+
"logits/chosen": -2.841184139251709,
|
| 447 |
+
"logits/rejected": -2.813793659210205,
|
| 448 |
+
"logps/chosen": -295.7508544921875,
|
| 449 |
+
"logps/rejected": -291.97467041015625,
|
| 450 |
+
"loss": 0.5071,
|
| 451 |
+
"pred_label": 0.0,
|
| 452 |
+
"rewards/accuracies": 0.746874988079071,
|
| 453 |
+
"rewards/chosen": -0.13246159255504608,
|
| 454 |
+
"rewards/margins": 1.0678224563598633,
|
| 455 |
+
"rewards/rejected": -1.2002841234207153,
|
| 456 |
+
"step": 270,
|
| 457 |
+
"use_label": 0.0
|
| 458 |
+
},
|
| 459 |
+
{
|
| 460 |
+
"epoch": 0.59,
|
| 461 |
+
"learning_rate": 2.296037296037296e-05,
|
| 462 |
+
"logits/chosen": -2.8141586780548096,
|
| 463 |
+
"logits/rejected": -2.803541898727417,
|
| 464 |
+
"logps/chosen": -283.2509765625,
|
| 465 |
+
"logps/rejected": -250.5514373779297,
|
| 466 |
+
"loss": 0.5358,
|
| 467 |
+
"pred_label": 0.0,
|
| 468 |
+
"rewards/accuracies": 0.715624988079071,
|
| 469 |
+
"rewards/chosen": -0.14472146332263947,
|
| 470 |
+
"rewards/margins": 0.9390050768852234,
|
| 471 |
+
"rewards/rejected": -1.0837266445159912,
|
| 472 |
+
"step": 280,
|
| 473 |
+
"use_label": 0.0
|
| 474 |
+
},
|
| 475 |
+
{
|
| 476 |
+
"epoch": 0.61,
|
| 477 |
+
"learning_rate": 2.1794871794871795e-05,
|
| 478 |
+
"logits/chosen": -2.821655035018921,
|
| 479 |
+
"logits/rejected": -2.8060765266418457,
|
| 480 |
+
"logps/chosen": -274.44903564453125,
|
| 481 |
+
"logps/rejected": -272.57745361328125,
|
| 482 |
+
"loss": 0.5206,
|
| 483 |
+
"pred_label": 0.0,
|
| 484 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 485 |
+
"rewards/chosen": -0.06504921615123749,
|
| 486 |
+
"rewards/margins": 0.8676969408988953,
|
| 487 |
+
"rewards/rejected": -0.9327462315559387,
|
| 488 |
+
"step": 290,
|
| 489 |
+
"use_label": 0.0
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"epoch": 0.63,
|
| 493 |
+
"learning_rate": 2.062937062937063e-05,
|
| 494 |
+
"logits/chosen": -2.8019015789031982,
|
| 495 |
+
"logits/rejected": -2.801147222518921,
|
| 496 |
+
"logps/chosen": -292.081787109375,
|
| 497 |
+
"logps/rejected": -301.0537414550781,
|
| 498 |
+
"loss": 0.5177,
|
| 499 |
+
"pred_label": 0.0,
|
| 500 |
+
"rewards/accuracies": 0.734375,
|
| 501 |
+
"rewards/chosen": 0.03986026719212532,
|
| 502 |
+
"rewards/margins": 0.968209445476532,
|
| 503 |
+
"rewards/rejected": -0.9283491969108582,
|
| 504 |
+
"step": 300,
|
| 505 |
+
"use_label": 0.0
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 0.65,
|
| 509 |
+
"learning_rate": 1.9463869463869462e-05,
|
| 510 |
+
"logits/chosen": -2.809706449508667,
|
| 511 |
+
"logits/rejected": -2.781670570373535,
|
| 512 |
+
"logps/chosen": -293.81451416015625,
|
| 513 |
+
"logps/rejected": -256.2749938964844,
|
| 514 |
+
"loss": 0.5066,
|
| 515 |
+
"pred_label": 0.0,
|
| 516 |
+
"rewards/accuracies": 0.7281249761581421,
|
| 517 |
+
"rewards/chosen": -0.08690959960222244,
|
| 518 |
+
"rewards/margins": 0.949812114238739,
|
| 519 |
+
"rewards/rejected": -1.0367217063903809,
|
| 520 |
+
"step": 310,
|
| 521 |
+
"use_label": 0.0
|
| 522 |
+
},
|
| 523 |
+
{
|
| 524 |
+
"epoch": 0.67,
|
| 525 |
+
"learning_rate": 1.8298368298368298e-05,
|
| 526 |
+
"logits/chosen": -2.797341823577881,
|
| 527 |
+
"logits/rejected": -2.787775754928589,
|
| 528 |
+
"logps/chosen": -269.12139892578125,
|
| 529 |
+
"logps/rejected": -266.01080322265625,
|
| 530 |
+
"loss": 0.4996,
|
| 531 |
+
"pred_label": 0.0,
|
| 532 |
+
"rewards/accuracies": 0.768750011920929,
|
| 533 |
+
"rewards/chosen": 0.0034091435372829437,
|
| 534 |
+
"rewards/margins": 1.1171700954437256,
|
| 535 |
+
"rewards/rejected": -1.1137609481811523,
|
| 536 |
+
"step": 320,
|
| 537 |
+
"use_label": 0.0
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.69,
|
| 541 |
+
"learning_rate": 1.7132867132867133e-05,
|
| 542 |
+
"logits/chosen": -2.8247230052948,
|
| 543 |
+
"logits/rejected": -2.80410099029541,
|
| 544 |
+
"logps/chosen": -299.09442138671875,
|
| 545 |
+
"logps/rejected": -273.0837707519531,
|
| 546 |
+
"loss": 0.5236,
|
| 547 |
+
"pred_label": 0.0,
|
| 548 |
+
"rewards/accuracies": 0.731249988079071,
|
| 549 |
+
"rewards/chosen": -0.04382320120930672,
|
| 550 |
+
"rewards/margins": 0.871612548828125,
|
| 551 |
+
"rewards/rejected": -0.915435791015625,
|
| 552 |
+
"step": 330,
|
| 553 |
+
"use_label": 0.0
|
| 554 |
+
},
|
| 555 |
+
{
|
| 556 |
+
"epoch": 0.71,
|
| 557 |
+
"learning_rate": 1.5967365967365968e-05,
|
| 558 |
+
"logits/chosen": -2.806962251663208,
|
| 559 |
+
"logits/rejected": -2.79986572265625,
|
| 560 |
+
"logps/chosen": -284.2406005859375,
|
| 561 |
+
"logps/rejected": -258.4219665527344,
|
| 562 |
+
"loss": 0.5258,
|
| 563 |
+
"pred_label": 0.0,
|
| 564 |
+
"rewards/accuracies": 0.7093750238418579,
|
| 565 |
+
"rewards/chosen": -0.023638445883989334,
|
| 566 |
+
"rewards/margins": 0.8371860384941101,
|
| 567 |
+
"rewards/rejected": -0.8608245849609375,
|
| 568 |
+
"step": 340,
|
| 569 |
+
"use_label": 0.0
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"epoch": 0.73,
|
| 573 |
+
"learning_rate": 1.4801864801864803e-05,
|
| 574 |
+
"logits/chosen": -2.803683280944824,
|
| 575 |
+
"logits/rejected": -2.7878894805908203,
|
| 576 |
+
"logps/chosen": -263.1024169921875,
|
| 577 |
+
"logps/rejected": -239.3209686279297,
|
| 578 |
+
"loss": 0.4963,
|
| 579 |
+
"pred_label": 0.0,
|
| 580 |
+
"rewards/accuracies": 0.765625,
|
| 581 |
+
"rewards/chosen": -0.0065319957211613655,
|
| 582 |
+
"rewards/margins": 1.0474125146865845,
|
| 583 |
+
"rewards/rejected": -1.05394446849823,
|
| 584 |
+
"step": 350,
|
| 585 |
+
"use_label": 0.0
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 0.75,
|
| 589 |
+
"learning_rate": 1.3636363636363637e-05,
|
| 590 |
+
"logits/chosen": -2.8253402709960938,
|
| 591 |
+
"logits/rejected": -2.817469596862793,
|
| 592 |
+
"logps/chosen": -280.67022705078125,
|
| 593 |
+
"logps/rejected": -258.00579833984375,
|
| 594 |
+
"loss": 0.4967,
|
| 595 |
+
"pred_label": 0.0,
|
| 596 |
+
"rewards/accuracies": 0.734375,
|
| 597 |
+
"rewards/chosen": -0.015683341771364212,
|
| 598 |
+
"rewards/margins": 1.0513224601745605,
|
| 599 |
+
"rewards/rejected": -1.0670057535171509,
|
| 600 |
+
"step": 360,
|
| 601 |
+
"use_label": 0.0
|
| 602 |
+
},
|
| 603 |
+
{
|
| 604 |
+
"epoch": 0.77,
|
| 605 |
+
"learning_rate": 1.2470862470862472e-05,
|
| 606 |
+
"logits/chosen": -2.793775796890259,
|
| 607 |
+
"logits/rejected": -2.796365261077881,
|
| 608 |
+
"logps/chosen": -302.306884765625,
|
| 609 |
+
"logps/rejected": -273.70635986328125,
|
| 610 |
+
"loss": 0.5115,
|
| 611 |
+
"pred_label": 0.0,
|
| 612 |
+
"rewards/accuracies": 0.737500011920929,
|
| 613 |
+
"rewards/chosen": 0.010395345278084278,
|
| 614 |
+
"rewards/margins": 0.973157525062561,
|
| 615 |
+
"rewards/rejected": -0.9627620577812195,
|
| 616 |
+
"step": 370,
|
| 617 |
+
"use_label": 0.0
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 0.8,
|
| 621 |
+
"learning_rate": 1.1305361305361307e-05,
|
| 622 |
+
"logits/chosen": -2.8007330894470215,
|
| 623 |
+
"logits/rejected": -2.809499740600586,
|
| 624 |
+
"logps/chosen": -277.95599365234375,
|
| 625 |
+
"logps/rejected": -271.8043212890625,
|
| 626 |
+
"loss": 0.5023,
|
| 627 |
+
"pred_label": 0.0,
|
| 628 |
+
"rewards/accuracies": 0.768750011920929,
|
| 629 |
+
"rewards/chosen": 0.0019541799556463957,
|
| 630 |
+
"rewards/margins": 1.0503606796264648,
|
| 631 |
+
"rewards/rejected": -1.0484063625335693,
|
| 632 |
+
"step": 380,
|
| 633 |
+
"use_label": 0.0
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"epoch": 0.82,
|
| 637 |
+
"learning_rate": 1.013986013986014e-05,
|
| 638 |
+
"logits/chosen": -2.838869571685791,
|
| 639 |
+
"logits/rejected": -2.8424456119537354,
|
| 640 |
+
"logps/chosen": -291.6543884277344,
|
| 641 |
+
"logps/rejected": -275.7422790527344,
|
| 642 |
+
"loss": 0.5223,
|
| 643 |
+
"pred_label": 0.0,
|
| 644 |
+
"rewards/accuracies": 0.7406250238418579,
|
| 645 |
+
"rewards/chosen": 0.009039236232638359,
|
| 646 |
+
"rewards/margins": 0.990314781665802,
|
| 647 |
+
"rewards/rejected": -0.9812755584716797,
|
| 648 |
+
"step": 390,
|
| 649 |
+
"use_label": 0.0
|
| 650 |
+
},
|
| 651 |
+
{
|
| 652 |
+
"epoch": 0.84,
|
| 653 |
+
"learning_rate": 8.974358974358976e-06,
|
| 654 |
+
"logits/chosen": -2.7799577713012695,
|
| 655 |
+
"logits/rejected": -2.7700963020324707,
|
| 656 |
+
"logps/chosen": -292.00604248046875,
|
| 657 |
+
"logps/rejected": -278.8124694824219,
|
| 658 |
+
"loss": 0.5059,
|
| 659 |
+
"pred_label": 0.0,
|
| 660 |
+
"rewards/accuracies": 0.7281249761581421,
|
| 661 |
+
"rewards/chosen": -0.08654220402240753,
|
| 662 |
+
"rewards/margins": 0.9550994634628296,
|
| 663 |
+
"rewards/rejected": -1.0416417121887207,
|
| 664 |
+
"step": 400,
|
| 665 |
+
"use_label": 0.0
|
| 666 |
+
},
|
| 667 |
+
{
|
| 668 |
+
"epoch": 0.86,
|
| 669 |
+
"learning_rate": 7.808857808857809e-06,
|
| 670 |
+
"logits/chosen": -2.7790863513946533,
|
| 671 |
+
"logits/rejected": -2.7839720249176025,
|
| 672 |
+
"logps/chosen": -288.0146484375,
|
| 673 |
+
"logps/rejected": -246.18386840820312,
|
| 674 |
+
"loss": 0.5193,
|
| 675 |
+
"pred_label": 0.0,
|
| 676 |
+
"rewards/accuracies": 0.7593749761581421,
|
| 677 |
+
"rewards/chosen": -0.0843573808670044,
|
| 678 |
+
"rewards/margins": 1.01529860496521,
|
| 679 |
+
"rewards/rejected": -1.0996559858322144,
|
| 680 |
+
"step": 410,
|
| 681 |
+
"use_label": 0.0
|
| 682 |
+
},
|
| 683 |
+
{
|
| 684 |
+
"epoch": 0.88,
|
| 685 |
+
"learning_rate": 6.643356643356643e-06,
|
| 686 |
+
"logits/chosen": -2.8042919635772705,
|
| 687 |
+
"logits/rejected": -2.814549446105957,
|
| 688 |
+
"logps/chosen": -267.73272705078125,
|
| 689 |
+
"logps/rejected": -282.6224060058594,
|
| 690 |
+
"loss": 0.5131,
|
| 691 |
+
"pred_label": 0.0,
|
| 692 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 693 |
+
"rewards/chosen": -0.23174908757209778,
|
| 694 |
+
"rewards/margins": 0.7961767315864563,
|
| 695 |
+
"rewards/rejected": -1.027925729751587,
|
| 696 |
+
"step": 420,
|
| 697 |
+
"use_label": 0.0
|
| 698 |
+
},
|
| 699 |
+
{
|
| 700 |
+
"epoch": 0.9,
|
| 701 |
+
"learning_rate": 5.477855477855478e-06,
|
| 702 |
+
"logits/chosen": -2.8008358478546143,
|
| 703 |
+
"logits/rejected": -2.827676773071289,
|
| 704 |
+
"logps/chosen": -291.685791015625,
|
| 705 |
+
"logps/rejected": -269.49896240234375,
|
| 706 |
+
"loss": 0.5135,
|
| 707 |
+
"pred_label": 0.0,
|
| 708 |
+
"rewards/accuracies": 0.7593749761581421,
|
| 709 |
+
"rewards/chosen": -0.1730644553899765,
|
| 710 |
+
"rewards/margins": 0.9842265248298645,
|
| 711 |
+
"rewards/rejected": -1.157291054725647,
|
| 712 |
+
"step": 430,
|
| 713 |
+
"use_label": 0.0
|
| 714 |
+
},
|
| 715 |
+
{
|
| 716 |
+
"epoch": 0.92,
|
| 717 |
+
"learning_rate": 4.312354312354312e-06,
|
| 718 |
+
"logits/chosen": -2.8276031017303467,
|
| 719 |
+
"logits/rejected": -2.81488299369812,
|
| 720 |
+
"logps/chosen": -265.67242431640625,
|
| 721 |
+
"logps/rejected": -261.6026611328125,
|
| 722 |
+
"loss": 0.51,
|
| 723 |
+
"pred_label": 0.0,
|
| 724 |
+
"rewards/accuracies": 0.734375,
|
| 725 |
+
"rewards/chosen": -0.08840381354093552,
|
| 726 |
+
"rewards/margins": 0.9768926501274109,
|
| 727 |
+
"rewards/rejected": -1.0652964115142822,
|
| 728 |
+
"step": 440,
|
| 729 |
+
"use_label": 0.0
|
| 730 |
+
},
|
| 731 |
+
{
|
| 732 |
+
"epoch": 0.94,
|
| 733 |
+
"learning_rate": 3.1468531468531472e-06,
|
| 734 |
+
"logits/chosen": -2.794529676437378,
|
| 735 |
+
"logits/rejected": -2.7928192615509033,
|
| 736 |
+
"logps/chosen": -280.88995361328125,
|
| 737 |
+
"logps/rejected": -275.9464416503906,
|
| 738 |
+
"loss": 0.5128,
|
| 739 |
+
"pred_label": 0.0,
|
| 740 |
+
"rewards/accuracies": 0.765625,
|
| 741 |
+
"rewards/chosen": -0.06786171346902847,
|
| 742 |
+
"rewards/margins": 0.9994968175888062,
|
| 743 |
+
"rewards/rejected": -1.0673584938049316,
|
| 744 |
+
"step": 450,
|
| 745 |
+
"use_label": 0.0
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 0.96,
|
| 749 |
+
"learning_rate": 1.981351981351981e-06,
|
| 750 |
+
"logits/chosen": -2.7951102256774902,
|
| 751 |
+
"logits/rejected": -2.7973313331604004,
|
| 752 |
+
"logps/chosen": -297.4739990234375,
|
| 753 |
+
"logps/rejected": -272.06024169921875,
|
| 754 |
+
"loss": 0.4871,
|
| 755 |
+
"pred_label": 0.0,
|
| 756 |
+
"rewards/accuracies": 0.737500011920929,
|
| 757 |
+
"rewards/chosen": -0.08221320062875748,
|
| 758 |
+
"rewards/margins": 0.9878571629524231,
|
| 759 |
+
"rewards/rejected": -1.0700703859329224,
|
| 760 |
+
"step": 460,
|
| 761 |
+
"use_label": 0.0
|
| 762 |
+
},
|
| 763 |
+
{
|
| 764 |
+
"epoch": 0.98,
|
| 765 |
+
"learning_rate": 8.158508158508159e-07,
|
| 766 |
+
"logits/chosen": -2.8256263732910156,
|
| 767 |
+
"logits/rejected": -2.7975211143493652,
|
| 768 |
+
"logps/chosen": -278.83154296875,
|
| 769 |
+
"logps/rejected": -266.3514099121094,
|
| 770 |
+
"loss": 0.4951,
|
| 771 |
+
"pred_label": 0.0,
|
| 772 |
+
"rewards/accuracies": 0.7593749761581421,
|
| 773 |
+
"rewards/chosen": -0.10287537425756454,
|
| 774 |
+
"rewards/margins": 1.0982601642608643,
|
| 775 |
+
"rewards/rejected": -1.2011353969573975,
|
| 776 |
+
"step": 470,
|
| 777 |
+
"use_label": 0.0
|
| 778 |
},
|
| 779 |
{
|
| 780 |
"epoch": 1.0,
|
| 781 |
+
"eval_logits/chosen": -2.822110891342163,
|
| 782 |
+
"eval_logits/rejected": -2.8178136348724365,
|
| 783 |
+
"eval_logps/chosen": -285.4951171875,
|
| 784 |
+
"eval_logps/rejected": -270.55743408203125,
|
| 785 |
+
"eval_loss": 0.5198934078216553,
|
| 786 |
+
"eval_pred_label": 0.0,
|
| 787 |
+
"eval_rewards/accuracies": 0.7300000190734863,
|
| 788 |
+
"eval_rewards/chosen": -0.12377375364303589,
|
| 789 |
+
"eval_rewards/margins": 1.001997470855713,
|
| 790 |
+
"eval_rewards/rejected": -1.1257712841033936,
|
| 791 |
+
"eval_runtime": 453.8631,
|
| 792 |
+
"eval_samples_per_second": 4.407,
|
| 793 |
+
"eval_steps_per_second": 0.275,
|
| 794 |
+
"eval_use_label": 0.0,
|
| 795 |
+
"step": 477
|
| 796 |
+
},
|
| 797 |
+
{
|
| 798 |
+
"epoch": 1.0,
|
| 799 |
+
"step": 477,
|
| 800 |
"total_flos": 0.0,
|
| 801 |
+
"train_loss": 0.5321606655040877,
|
| 802 |
+
"train_runtime": 24451.2028,
|
| 803 |
+
"train_samples_per_second": 2.5,
|
| 804 |
+
"train_steps_per_second": 0.02
|
| 805 |
}
|
| 806 |
],
|
| 807 |
"logging_steps": 10,
|
| 808 |
+
"max_steps": 477,
|
| 809 |
"num_train_epochs": 1,
|
| 810 |
+
"save_steps": 50,
|
| 811 |
"total_flos": 0.0,
|
| 812 |
"trial_name": null,
|
| 813 |
"trial_params": null
|