Model save
Browse files- README.md +16 -14
- adapter_model.safetensors +1 -1
- all_results.json +18 -18
- eval_results.json +13 -13
- runs/Mar01_07-41-58_gpu4-119-4/events.out.tfevents.1709239476.gpu4-119-4.1724210.0 +2 -2
- runs/Mar01_07-41-58_gpu4-119-4/events.out.tfevents.1709246562.gpu4-119-4.1724210.1 +3 -0
- train_results.json +6 -6
- trainer_state.json +704 -108
README.md
CHANGED
|
@@ -2,13 +2,9 @@
|
|
| 2 |
license: mit
|
| 3 |
library_name: peft
|
| 4 |
tags:
|
| 5 |
-
- alignment-handbook
|
| 6 |
-
- generated_from_trainer
|
| 7 |
- trl
|
| 8 |
- dpo
|
| 9 |
- generated_from_trainer
|
| 10 |
-
datasets:
|
| 11 |
-
- HuggingFaceH4/ultrafeedback_binarized
|
| 12 |
base_model: microsoft/phi-2
|
| 13 |
model-index:
|
| 14 |
- name: phi-2-dpo-ultrachat-lora
|
|
@@ -20,17 +16,17 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 20 |
|
| 21 |
# phi-2-dpo-ultrachat-lora
|
| 22 |
|
| 23 |
-
This model is a fine-tuned version of [
|
| 24 |
It achieves the following results on the evaluation set:
|
| 25 |
-
- Loss: 0.
|
| 26 |
-
- Rewards/chosen: -0.
|
| 27 |
-
- Rewards/rejected: -0.
|
| 28 |
-
- Rewards/accuracies: 0.
|
| 29 |
-
- Rewards/margins:
|
| 30 |
-
- Logps/rejected: -
|
| 31 |
-
- Logps/chosen: -
|
| 32 |
-
- Logits/rejected: 0.
|
| 33 |
-
- Logits/chosen: 0.
|
| 34 |
|
| 35 |
## Model description
|
| 36 |
|
|
@@ -65,6 +61,12 @@ The following hyperparameters were used during training:
|
|
| 65 |
|
| 66 |
### Training results
|
| 67 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 68 |
|
| 69 |
|
| 70 |
### Framework versions
|
|
|
|
| 2 |
license: mit
|
| 3 |
library_name: peft
|
| 4 |
tags:
|
|
|
|
|
|
|
| 5 |
- trl
|
| 6 |
- dpo
|
| 7 |
- generated_from_trainer
|
|
|
|
|
|
|
| 8 |
base_model: microsoft/phi-2
|
| 9 |
model-index:
|
| 10 |
- name: phi-2-dpo-ultrachat-lora
|
|
|
|
| 16 |
|
| 17 |
# phi-2-dpo-ultrachat-lora
|
| 18 |
|
| 19 |
+
This model is a fine-tuned version of [microsoft/phi-2](https://huggingface.co/microsoft/phi-2) on the None dataset.
|
| 20 |
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.6912
|
| 22 |
+
- Rewards/chosen: -0.0072
|
| 23 |
+
- Rewards/rejected: -0.0111
|
| 24 |
+
- Rewards/accuracies: 0.3180
|
| 25 |
+
- Rewards/margins: 0.0040
|
| 26 |
+
- Logps/rejected: -95.3090
|
| 27 |
+
- Logps/chosen: -92.4438
|
| 28 |
+
- Logits/rejected: 0.8021
|
| 29 |
+
- Logits/chosen: 0.7828
|
| 30 |
|
| 31 |
## Model description
|
| 32 |
|
|
|
|
| 61 |
|
| 62 |
### Training results
|
| 63 |
|
| 64 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 65 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 66 |
+
| 0.693 | 0.21 | 100 | 0.6931 | -0.0005 | -0.0008 | 0.2680 | 0.0004 | -94.2804 | -91.7748 | 0.8176 | 0.7998 |
|
| 67 |
+
| 0.6922 | 0.42 | 200 | 0.6924 | -0.0018 | -0.0032 | 0.3020 | 0.0014 | -94.5141 | -91.9068 | 0.8121 | 0.7941 |
|
| 68 |
+
| 0.6917 | 0.63 | 300 | 0.6917 | -0.0049 | -0.0077 | 0.3100 | 0.0028 | -94.9659 | -92.2189 | 0.8057 | 0.7870 |
|
| 69 |
+
| 0.6905 | 0.84 | 400 | 0.6913 | -0.0070 | -0.0105 | 0.3280 | 0.0036 | -95.2509 | -92.4247 | 0.8012 | 0.7827 |
|
| 70 |
|
| 71 |
|
| 72 |
### Framework versions
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 41977616
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:305ec4283375fd6264f9e6b7561a8941ac32582c9088af742777fac3c094eeba
|
| 3 |
size 41977616
|
all_results.json
CHANGED
|
@@ -1,21 +1,21 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch": 0
|
| 3 |
-
"eval_logits/chosen": 0.
|
| 4 |
-
"eval_logits/rejected": 0.
|
| 5 |
-
"eval_logps/chosen": -
|
| 6 |
-
"eval_logps/rejected": -
|
| 7 |
-
"eval_loss": 0.
|
| 8 |
-
"eval_rewards/accuracies": 0.
|
| 9 |
-
"eval_rewards/chosen": -0.
|
| 10 |
-
"eval_rewards/margins":
|
| 11 |
-
"eval_rewards/rejected": -0.
|
| 12 |
-
"eval_runtime": 273.
|
| 13 |
"eval_samples": 2000,
|
| 14 |
-
"eval_samples_per_second": 7.
|
| 15 |
-
"eval_steps_per_second": 0.
|
| 16 |
-
"train_loss": 0.
|
| 17 |
-
"train_runtime":
|
| 18 |
-
"train_samples":
|
| 19 |
-
"train_samples_per_second":
|
| 20 |
-
"train_steps_per_second": 0.
|
| 21 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"eval_logits/chosen": 0.7827913761138916,
|
| 4 |
+
"eval_logits/rejected": 0.8020623326301575,
|
| 5 |
+
"eval_logps/chosen": -92.44380187988281,
|
| 6 |
+
"eval_logps/rejected": -95.30902099609375,
|
| 7 |
+
"eval_loss": 0.6912217736244202,
|
| 8 |
+
"eval_rewards/accuracies": 0.3179999887943268,
|
| 9 |
+
"eval_rewards/chosen": -0.00716440798714757,
|
| 10 |
+
"eval_rewards/margins": 0.00395576748996973,
|
| 11 |
+
"eval_rewards/rejected": -0.011120175942778587,
|
| 12 |
+
"eval_runtime": 273.2522,
|
| 13 |
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 7.319,
|
| 15 |
+
"eval_steps_per_second": 0.457,
|
| 16 |
+
"train_loss": 0.6920521804121805,
|
| 17 |
+
"train_runtime": 6813.2628,
|
| 18 |
+
"train_samples": 30567,
|
| 19 |
+
"train_samples_per_second": 4.486,
|
| 20 |
+
"train_steps_per_second": 0.07
|
| 21 |
}
|
eval_results.json
CHANGED
|
@@ -1,16 +1,16 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch": 0
|
| 3 |
-
"eval_logits/chosen": 0.
|
| 4 |
-
"eval_logits/rejected": 0.
|
| 5 |
-
"eval_logps/chosen": -
|
| 6 |
-
"eval_logps/rejected": -
|
| 7 |
-
"eval_loss": 0.
|
| 8 |
-
"eval_rewards/accuracies": 0.
|
| 9 |
-
"eval_rewards/chosen": -0.
|
| 10 |
-
"eval_rewards/margins":
|
| 11 |
-
"eval_rewards/rejected": -0.
|
| 12 |
-
"eval_runtime": 273.
|
| 13 |
"eval_samples": 2000,
|
| 14 |
-
"eval_samples_per_second": 7.
|
| 15 |
-
"eval_steps_per_second": 0.
|
| 16 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"eval_logits/chosen": 0.7827913761138916,
|
| 4 |
+
"eval_logits/rejected": 0.8020623326301575,
|
| 5 |
+
"eval_logps/chosen": -92.44380187988281,
|
| 6 |
+
"eval_logps/rejected": -95.30902099609375,
|
| 7 |
+
"eval_loss": 0.6912217736244202,
|
| 8 |
+
"eval_rewards/accuracies": 0.3179999887943268,
|
| 9 |
+
"eval_rewards/chosen": -0.00716440798714757,
|
| 10 |
+
"eval_rewards/margins": 0.00395576748996973,
|
| 11 |
+
"eval_rewards/rejected": -0.011120175942778587,
|
| 12 |
+
"eval_runtime": 273.2522,
|
| 13 |
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 7.319,
|
| 15 |
+
"eval_steps_per_second": 0.457
|
| 16 |
}
|
runs/Mar01_07-41-58_gpu4-119-4/events.out.tfevents.1709239476.gpu4-119-4.1724210.0
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:e0914a4c2c529be24a3baad7b988b701bedae6634d92d5f0f86147dd98279695
|
| 3 |
+
size 38420
|
runs/Mar01_07-41-58_gpu4-119-4/events.out.tfevents.1709246562.gpu4-119-4.1724210.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c2ce15b9319daaf97e9395bf9eabce614a4c92f3b9d1d8ac2b369d584227fa6f
|
| 3 |
+
size 828
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch": 0
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
-
"train_samples":
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.6920521804121805,
|
| 4 |
+
"train_runtime": 6813.2628,
|
| 5 |
+
"train_samples": 30567,
|
| 6 |
+
"train_samples_per_second": 4.486,
|
| 7 |
+
"train_steps_per_second": 0.07
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -1,20 +1,20 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch": 0.
|
| 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 |
-
"epoch": 0.
|
| 13 |
-
"learning_rate":
|
| 14 |
-
"logits/chosen": 0.
|
| 15 |
-
"logits/rejected": 0.
|
| 16 |
-
"logps/chosen": -
|
| 17 |
-
"logps/rejected": -
|
| 18 |
"loss": 0.6931,
|
| 19 |
"rewards/accuracies": 0.0,
|
| 20 |
"rewards/chosen": 0.0,
|
|
@@ -23,143 +23,739 @@
|
|
| 23 |
"step": 1
|
| 24 |
},
|
| 25 |
{
|
| 26 |
-
"epoch": 0.
|
| 27 |
-
"learning_rate":
|
| 28 |
-
"logits/chosen": 0.
|
| 29 |
-
"logits/rejected": 0.
|
| 30 |
-
"logps/chosen": -
|
| 31 |
-
"logps/rejected": -
|
| 32 |
-
"loss": 0.
|
| 33 |
"rewards/accuracies": 0.2361111044883728,
|
| 34 |
-
"rewards/chosen":
|
| 35 |
-
"rewards/margins": -
|
| 36 |
-
"rewards/rejected":
|
| 37 |
"step": 10
|
| 38 |
},
|
| 39 |
{
|
| 40 |
-
"epoch": 0.
|
| 41 |
-
"learning_rate":
|
| 42 |
-
"logits/chosen": 0.
|
| 43 |
-
"logits/rejected": 0.
|
| 44 |
-
"logps/chosen": -
|
| 45 |
-
"logps/rejected": -
|
| 46 |
"loss": 0.6932,
|
| 47 |
-
"rewards/accuracies": 0.
|
| 48 |
-
"rewards/chosen": 0.
|
| 49 |
-
"rewards/margins": 0.
|
| 50 |
-
"rewards/rejected":
|
| 51 |
"step": 20
|
| 52 |
},
|
| 53 |
{
|
| 54 |
-
"epoch": 0.
|
| 55 |
-
"learning_rate":
|
| 56 |
-
"logits/chosen": 0.
|
| 57 |
-
"logits/rejected": 0.
|
| 58 |
-
"logps/chosen": -
|
| 59 |
-
"logps/rejected": -
|
| 60 |
-
"loss": 0.
|
| 61 |
-
"rewards/accuracies": 0.
|
| 62 |
-
"rewards/chosen": 0.
|
| 63 |
-
"rewards/margins": 0.
|
| 64 |
-
"rewards/rejected":
|
| 65 |
"step": 30
|
| 66 |
},
|
| 67 |
{
|
| 68 |
-
"epoch": 0.
|
| 69 |
-
"learning_rate":
|
| 70 |
-
"logits/chosen": 0.
|
| 71 |
-
"logits/rejected": 0.
|
| 72 |
-
"logps/chosen": -
|
| 73 |
-
"logps/rejected": -
|
| 74 |
"loss": 0.6931,
|
| 75 |
-
"rewards/accuracies": 0.
|
| 76 |
-
"rewards/chosen":
|
| 77 |
-
"rewards/margins":
|
| 78 |
-
"rewards/rejected": -
|
| 79 |
"step": 40
|
| 80 |
},
|
| 81 |
{
|
| 82 |
-
"epoch": 0.
|
| 83 |
-
"learning_rate":
|
| 84 |
-
"logits/chosen": 0.
|
| 85 |
-
"logits/rejected": 0.
|
| 86 |
-
"logps/chosen": -
|
| 87 |
-
"logps/rejected": -
|
| 88 |
"loss": 0.6932,
|
| 89 |
-
"rewards/accuracies": 0.
|
| 90 |
-
"rewards/chosen": -0.
|
| 91 |
-
"rewards/margins": -0.
|
| 92 |
-
"rewards/rejected":
|
| 93 |
"step": 50
|
| 94 |
},
|
| 95 |
{
|
| 96 |
-
"epoch": 0.
|
| 97 |
-
"learning_rate":
|
| 98 |
-
"logits/chosen": 0.
|
| 99 |
-
"logits/rejected": 0.
|
| 100 |
-
"logps/chosen": -
|
| 101 |
-
"logps/rejected": -
|
| 102 |
"loss": 0.6931,
|
| 103 |
-
"rewards/accuracies": 0.
|
| 104 |
-
"rewards/chosen":
|
| 105 |
-
"rewards/margins": 0.
|
| 106 |
-
"rewards/rejected":
|
| 107 |
"step": 60
|
| 108 |
},
|
| 109 |
{
|
| 110 |
-
"epoch": 0.
|
| 111 |
-
"learning_rate":
|
| 112 |
-
"logits/chosen": 0.
|
| 113 |
-
"logits/rejected": 0.
|
| 114 |
-
"logps/chosen": -
|
| 115 |
-
"logps/rejected": -
|
| 116 |
-
"loss": 0.
|
| 117 |
-
"rewards/accuracies": 0.
|
| 118 |
-
"rewards/chosen": -0.
|
| 119 |
-
"rewards/margins": -
|
| 120 |
-
"rewards/rejected": -
|
| 121 |
"step": 70
|
| 122 |
},
|
| 123 |
{
|
| 124 |
-
"epoch": 0.
|
| 125 |
-
"learning_rate":
|
| 126 |
-
"logits/chosen": 0.
|
| 127 |
-
"logits/rejected": 0.
|
| 128 |
-
"logps/chosen": -
|
| 129 |
-
"logps/rejected": -
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 130 |
"loss": 0.6931,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 131 |
"rewards/accuracies": 0.2750000059604645,
|
| 132 |
-
"rewards/chosen": 0.
|
| 133 |
-
"rewards/margins": 0.
|
| 134 |
-
"rewards/rejected": -
|
| 135 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 136 |
},
|
| 137 |
{
|
| 138 |
"epoch": 0.94,
|
| 139 |
-
"learning_rate": 4.
|
| 140 |
-
"logits/chosen": 0.
|
| 141 |
-
"logits/rejected": 0.
|
| 142 |
-
"logps/chosen": -
|
| 143 |
-
"logps/rejected": -
|
| 144 |
-
"loss": 0.
|
| 145 |
-
"rewards/accuracies": 0.
|
| 146 |
-
"rewards/chosen": 0.
|
| 147 |
-
"rewards/margins": 0.
|
| 148 |
-
"rewards/rejected": -0.
|
| 149 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 150 |
},
|
| 151 |
{
|
| 152 |
-
"epoch": 0
|
| 153 |
-
"step":
|
| 154 |
"total_flos": 0.0,
|
| 155 |
-
"train_loss": 0.
|
| 156 |
-
"train_runtime":
|
| 157 |
-
"train_samples_per_second":
|
| 158 |
-
"train_steps_per_second": 0.
|
| 159 |
}
|
| 160 |
],
|
| 161 |
"logging_steps": 10,
|
| 162 |
-
"max_steps":
|
| 163 |
"num_input_tokens_seen": 0,
|
| 164 |
"num_train_epochs": 1,
|
| 165 |
"save_steps": 100,
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9984301412872841,
|
| 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-07,
|
| 14 |
+
"logits/chosen": 0.8531318306922913,
|
| 15 |
+
"logits/rejected": 0.7327959537506104,
|
| 16 |
+
"logps/chosen": -133.83494567871094,
|
| 17 |
+
"logps/rejected": -129.6299285888672,
|
| 18 |
"loss": 0.6931,
|
| 19 |
"rewards/accuracies": 0.0,
|
| 20 |
"rewards/chosen": 0.0,
|
|
|
|
| 23 |
"step": 1
|
| 24 |
},
|
| 25 |
{
|
| 26 |
+
"epoch": 0.02,
|
| 27 |
+
"learning_rate": 1.0416666666666667e-06,
|
| 28 |
+
"logits/chosen": 0.8151362538337708,
|
| 29 |
+
"logits/rejected": 0.7788704037666321,
|
| 30 |
+
"logps/chosen": -90.5947036743164,
|
| 31 |
+
"logps/rejected": -92.77328491210938,
|
| 32 |
+
"loss": 0.6931,
|
| 33 |
"rewards/accuracies": 0.2361111044883728,
|
| 34 |
+
"rewards/chosen": 0.00018473717500455678,
|
| 35 |
+
"rewards/margins": 6.259369547478855e-05,
|
| 36 |
+
"rewards/rejected": 0.00012214347952976823,
|
| 37 |
"step": 10
|
| 38 |
},
|
| 39 |
{
|
| 40 |
+
"epoch": 0.04,
|
| 41 |
+
"learning_rate": 2.0833333333333334e-06,
|
| 42 |
+
"logits/chosen": 0.8342723846435547,
|
| 43 |
+
"logits/rejected": 0.8797974586486816,
|
| 44 |
+
"logps/chosen": -81.16204071044922,
|
| 45 |
+
"logps/rejected": -81.1548080444336,
|
| 46 |
"loss": 0.6932,
|
| 47 |
+
"rewards/accuracies": 0.17499999701976776,
|
| 48 |
+
"rewards/chosen": -0.00041853776201605797,
|
| 49 |
+
"rewards/margins": -0.0003428882628213614,
|
| 50 |
+
"rewards/rejected": -7.564939733128995e-05,
|
| 51 |
"step": 20
|
| 52 |
},
|
| 53 |
{
|
| 54 |
+
"epoch": 0.06,
|
| 55 |
+
"learning_rate": 3.125e-06,
|
| 56 |
+
"logits/chosen": 0.8050435185432434,
|
| 57 |
+
"logits/rejected": 0.8031445741653442,
|
| 58 |
+
"logps/chosen": -101.43102264404297,
|
| 59 |
+
"logps/rejected": -98.3962631225586,
|
| 60 |
+
"loss": 0.6932,
|
| 61 |
+
"rewards/accuracies": 0.23749999701976776,
|
| 62 |
+
"rewards/chosen": 0.00019087354303337634,
|
| 63 |
+
"rewards/margins": -0.00028236288926564157,
|
| 64 |
+
"rewards/rejected": 0.000473236374091357,
|
| 65 |
"step": 30
|
| 66 |
},
|
| 67 |
{
|
| 68 |
+
"epoch": 0.08,
|
| 69 |
+
"learning_rate": 4.166666666666667e-06,
|
| 70 |
+
"logits/chosen": 0.8352361917495728,
|
| 71 |
+
"logits/rejected": 0.8269574046134949,
|
| 72 |
+
"logps/chosen": -111.55462646484375,
|
| 73 |
+
"logps/rejected": -99.02640533447266,
|
| 74 |
"loss": 0.6931,
|
| 75 |
+
"rewards/accuracies": 0.23749999701976776,
|
| 76 |
+
"rewards/chosen": -5.0148169975727797e-05,
|
| 77 |
+
"rewards/margins": 7.251273200381547e-05,
|
| 78 |
+
"rewards/rejected": -0.00012266085832379758,
|
| 79 |
"step": 40
|
| 80 |
},
|
| 81 |
{
|
| 82 |
+
"epoch": 0.1,
|
| 83 |
+
"learning_rate": 4.999731868769027e-06,
|
| 84 |
+
"logits/chosen": 0.8326011896133423,
|
| 85 |
+
"logits/rejected": 0.8801982998847961,
|
| 86 |
+
"logps/chosen": -98.504638671875,
|
| 87 |
+
"logps/rejected": -83.65934753417969,
|
| 88 |
"loss": 0.6932,
|
| 89 |
+
"rewards/accuracies": 0.19374999403953552,
|
| 90 |
+
"rewards/chosen": -0.0006545094074681401,
|
| 91 |
+
"rewards/margins": -0.0007917654584161937,
|
| 92 |
+
"rewards/rejected": 0.00013725618191529065,
|
| 93 |
"step": 50
|
| 94 |
},
|
| 95 |
{
|
| 96 |
+
"epoch": 0.13,
|
| 97 |
+
"learning_rate": 4.9903533134293035e-06,
|
| 98 |
+
"logits/chosen": 0.8338711857795715,
|
| 99 |
+
"logits/rejected": 0.8361239433288574,
|
| 100 |
+
"logps/chosen": -105.47569274902344,
|
| 101 |
+
"logps/rejected": -103.5544662475586,
|
| 102 |
"loss": 0.6931,
|
| 103 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 104 |
+
"rewards/chosen": -6.698282231809571e-05,
|
| 105 |
+
"rewards/margins": -0.0007280272548086941,
|
| 106 |
+
"rewards/rejected": 0.0006610444979742169,
|
| 107 |
"step": 60
|
| 108 |
},
|
| 109 |
{
|
| 110 |
+
"epoch": 0.15,
|
| 111 |
+
"learning_rate": 4.967625656594782e-06,
|
| 112 |
+
"logits/chosen": 0.8040585517883301,
|
| 113 |
+
"logits/rejected": 0.8523794412612915,
|
| 114 |
+
"logps/chosen": -83.11436462402344,
|
| 115 |
+
"logps/rejected": -77.95774841308594,
|
| 116 |
+
"loss": 0.6931,
|
| 117 |
+
"rewards/accuracies": 0.26875001192092896,
|
| 118 |
+
"rewards/chosen": -0.0002090830384986475,
|
| 119 |
+
"rewards/margins": -0.0001823374768719077,
|
| 120 |
+
"rewards/rejected": -2.6745503419078887e-05,
|
| 121 |
"step": 70
|
| 122 |
},
|
| 123 |
{
|
| 124 |
+
"epoch": 0.17,
|
| 125 |
+
"learning_rate": 4.93167072587771e-06,
|
| 126 |
+
"logits/chosen": 0.8447190523147583,
|
| 127 |
+
"logits/rejected": 0.9011589288711548,
|
| 128 |
+
"logps/chosen": -101.15557098388672,
|
| 129 |
+
"logps/rejected": -84.94987487792969,
|
| 130 |
+
"loss": 0.693,
|
| 131 |
+
"rewards/accuracies": 0.3375000059604645,
|
| 132 |
+
"rewards/chosen": 0.00010431509872432798,
|
| 133 |
+
"rewards/margins": 0.000598998973146081,
|
| 134 |
+
"rewards/rejected": -0.0004946838016621768,
|
| 135 |
+
"step": 80
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.19,
|
| 139 |
+
"learning_rate": 4.882681251368549e-06,
|
| 140 |
+
"logits/chosen": 0.8417055010795593,
|
| 141 |
+
"logits/rejected": 0.8667267560958862,
|
| 142 |
+
"logps/chosen": -88.30390930175781,
|
| 143 |
+
"logps/rejected": -110.51224517822266,
|
| 144 |
"loss": 0.6931,
|
| 145 |
+
"rewards/accuracies": 0.24375000596046448,
|
| 146 |
+
"rewards/chosen": -0.0014625315088778734,
|
| 147 |
+
"rewards/margins": -0.0006530345417559147,
|
| 148 |
+
"rewards/rejected": -0.0008094970253296196,
|
| 149 |
+
"step": 90
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.21,
|
| 153 |
+
"learning_rate": 4.8209198325401815e-06,
|
| 154 |
+
"logits/chosen": 0.8801371455192566,
|
| 155 |
+
"logits/rejected": 0.9127315282821655,
|
| 156 |
+
"logps/chosen": -86.3653793334961,
|
| 157 |
+
"logps/rejected": -69.10803985595703,
|
| 158 |
+
"loss": 0.693,
|
| 159 |
+
"rewards/accuracies": 0.22499999403953552,
|
| 160 |
+
"rewards/chosen": -0.0005041404510848224,
|
| 161 |
+
"rewards/margins": 0.0003193179436493665,
|
| 162 |
+
"rewards/rejected": -0.0008234584820456803,
|
| 163 |
+
"step": 100
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.21,
|
| 167 |
+
"eval_logits/chosen": 0.7997792959213257,
|
| 168 |
+
"eval_logits/rejected": 0.8176011443138123,
|
| 169 |
+
"eval_logps/chosen": -91.77481842041016,
|
| 170 |
+
"eval_logps/rejected": -94.28044891357422,
|
| 171 |
+
"eval_loss": 0.6930604577064514,
|
| 172 |
+
"eval_rewards/accuracies": 0.2680000066757202,
|
| 173 |
+
"eval_rewards/chosen": -0.00047452302533201873,
|
| 174 |
+
"eval_rewards/margins": 0.00035985803697258234,
|
| 175 |
+
"eval_rewards/rejected": -0.0008343810332007706,
|
| 176 |
+
"eval_runtime": 273.126,
|
| 177 |
+
"eval_samples_per_second": 7.323,
|
| 178 |
+
"eval_steps_per_second": 0.458,
|
| 179 |
+
"step": 100
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 0.23,
|
| 183 |
+
"learning_rate": 4.746717530629565e-06,
|
| 184 |
+
"logits/chosen": 0.8060985803604126,
|
| 185 |
+
"logits/rejected": 0.8291690945625305,
|
| 186 |
+
"logps/chosen": -110.20963287353516,
|
| 187 |
+
"logps/rejected": -96.49766540527344,
|
| 188 |
+
"loss": 0.6929,
|
| 189 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 190 |
+
"rewards/chosen": -0.0005915694055147469,
|
| 191 |
+
"rewards/margins": 0.00022233165509533137,
|
| 192 |
+
"rewards/rejected": -0.0008139010751619935,
|
| 193 |
+
"step": 110
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 0.25,
|
| 197 |
+
"learning_rate": 4.660472094042121e-06,
|
| 198 |
+
"logits/chosen": 0.8455309867858887,
|
| 199 |
+
"logits/rejected": 0.9219174385070801,
|
| 200 |
+
"logps/chosen": -84.42987060546875,
|
| 201 |
+
"logps/rejected": -80.36955261230469,
|
| 202 |
+
"loss": 0.6931,
|
| 203 |
+
"rewards/accuracies": 0.28125,
|
| 204 |
+
"rewards/chosen": -0.0005250257672742009,
|
| 205 |
+
"rewards/margins": 0.00011121686839032918,
|
| 206 |
+
"rewards/rejected": -0.0006362426793202758,
|
| 207 |
+
"step": 120
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 0.27,
|
| 211 |
+
"learning_rate": 4.5626458262912745e-06,
|
| 212 |
+
"logits/chosen": 0.7771774530410767,
|
| 213 |
+
"logits/rejected": 0.8348148465156555,
|
| 214 |
+
"logps/chosen": -91.55365753173828,
|
| 215 |
+
"logps/rejected": -80.16468811035156,
|
| 216 |
+
"loss": 0.693,
|
| 217 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 218 |
+
"rewards/chosen": -0.0003579836920835078,
|
| 219 |
+
"rewards/margins": 0.0003911118838004768,
|
| 220 |
+
"rewards/rejected": -0.0007490954594686627,
|
| 221 |
+
"step": 130
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.29,
|
| 225 |
+
"learning_rate": 4.453763107901676e-06,
|
| 226 |
+
"logits/chosen": 0.8080952763557434,
|
| 227 |
+
"logits/rejected": 0.8902841806411743,
|
| 228 |
+
"logps/chosen": -102.3683853149414,
|
| 229 |
+
"logps/rejected": -121.3236312866211,
|
| 230 |
+
"loss": 0.6929,
|
| 231 |
+
"rewards/accuracies": 0.34375,
|
| 232 |
+
"rewards/chosen": -0.0011437158100306988,
|
| 233 |
+
"rewards/margins": 0.00023913508630357683,
|
| 234 |
+
"rewards/rejected": -0.001382850925438106,
|
| 235 |
+
"step": 140
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.31,
|
| 239 |
+
"learning_rate": 4.33440758555951e-06,
|
| 240 |
+
"logits/chosen": 0.8286693692207336,
|
| 241 |
+
"logits/rejected": 0.8885159492492676,
|
| 242 |
+
"logps/chosen": -98.10747528076172,
|
| 243 |
+
"logps/rejected": -75.61839294433594,
|
| 244 |
+
"loss": 0.6929,
|
| 245 |
+
"rewards/accuracies": 0.26875001192092896,
|
| 246 |
+
"rewards/chosen": -0.001014741021208465,
|
| 247 |
+
"rewards/margins": 0.00025303030270151794,
|
| 248 |
+
"rewards/rejected": -0.0012677714694291353,
|
| 249 |
+
"step": 150
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.33,
|
| 253 |
+
"learning_rate": 4.205219043576955e-06,
|
| 254 |
+
"logits/chosen": 0.7178612947463989,
|
| 255 |
+
"logits/rejected": 0.7653802633285522,
|
| 256 |
+
"logps/chosen": -80.03684997558594,
|
| 257 |
+
"logps/rejected": -73.78407287597656,
|
| 258 |
+
"loss": 0.6928,
|
| 259 |
+
"rewards/accuracies": 0.2562499940395355,
|
| 260 |
+
"rewards/chosen": -0.001032956875860691,
|
| 261 |
+
"rewards/margins": 0.0007252781069837511,
|
| 262 |
+
"rewards/rejected": -0.0017582349246367812,
|
| 263 |
+
"step": 160
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"epoch": 0.36,
|
| 267 |
+
"learning_rate": 4.066889974440757e-06,
|
| 268 |
+
"logits/chosen": 0.8685981631278992,
|
| 269 |
+
"logits/rejected": 0.9337660670280457,
|
| 270 |
+
"logps/chosen": -72.55715942382812,
|
| 271 |
+
"logps/rejected": -81.37494659423828,
|
| 272 |
+
"loss": 0.6929,
|
| 273 |
+
"rewards/accuracies": 0.26875001192092896,
|
| 274 |
+
"rewards/chosen": -0.0013817875878885388,
|
| 275 |
+
"rewards/margins": 0.0006504356861114502,
|
| 276 |
+
"rewards/rejected": -0.002032223390415311,
|
| 277 |
+
"step": 170
|
| 278 |
+
},
|
| 279 |
+
{
|
| 280 |
+
"epoch": 0.38,
|
| 281 |
+
"learning_rate": 3.92016186682789e-06,
|
| 282 |
+
"logits/chosen": 0.7872442007064819,
|
| 283 |
+
"logits/rejected": 0.8030352592468262,
|
| 284 |
+
"logps/chosen": -98.71012878417969,
|
| 285 |
+
"logps/rejected": -94.06107330322266,
|
| 286 |
+
"loss": 0.6928,
|
| 287 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 288 |
+
"rewards/chosen": -0.0005101398564875126,
|
| 289 |
+
"rewards/margins": 0.0011010088492184877,
|
| 290 |
+
"rewards/rejected": -0.0016111487057060003,
|
| 291 |
+
"step": 180
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"epoch": 0.4,
|
| 295 |
+
"learning_rate": 3.7658212309857576e-06,
|
| 296 |
+
"logits/chosen": 0.7398073077201843,
|
| 297 |
+
"logits/rejected": 0.8337036967277527,
|
| 298 |
+
"logps/chosen": -97.44853973388672,
|
| 299 |
+
"logps/rejected": -107.1574478149414,
|
| 300 |
+
"loss": 0.6926,
|
| 301 |
+
"rewards/accuracies": 0.3187499940395355,
|
| 302 |
+
"rewards/chosen": -0.0017274795100092888,
|
| 303 |
+
"rewards/margins": 0.0012662711087614298,
|
| 304 |
+
"rewards/rejected": -0.0029937506187707186,
|
| 305 |
+
"step": 190
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 0.42,
|
| 309 |
+
"learning_rate": 3.604695382782159e-06,
|
| 310 |
+
"logits/chosen": 0.7769767642021179,
|
| 311 |
+
"logits/rejected": 0.8381963968276978,
|
| 312 |
+
"logps/chosen": -77.53779602050781,
|
| 313 |
+
"logps/rejected": -86.9189224243164,
|
| 314 |
+
"loss": 0.6922,
|
| 315 |
+
"rewards/accuracies": 0.3125,
|
| 316 |
+
"rewards/chosen": -0.0016526943072676659,
|
| 317 |
+
"rewards/margins": 0.0023483093827962875,
|
| 318 |
+
"rewards/rejected": -0.004001003224402666,
|
| 319 |
+
"step": 200
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"epoch": 0.42,
|
| 323 |
+
"eval_logits/chosen": 0.7941381931304932,
|
| 324 |
+
"eval_logits/rejected": 0.8121381998062134,
|
| 325 |
+
"eval_logps/chosen": -91.90680694580078,
|
| 326 |
+
"eval_logps/rejected": -94.51410675048828,
|
| 327 |
+
"eval_loss": 0.692441999912262,
|
| 328 |
+
"eval_rewards/accuracies": 0.3019999861717224,
|
| 329 |
+
"eval_rewards/chosen": -0.0017943703569471836,
|
| 330 |
+
"eval_rewards/margins": 0.001376640284433961,
|
| 331 |
+
"eval_rewards/rejected": -0.0031710106413811445,
|
| 332 |
+
"eval_runtime": 273.1548,
|
| 333 |
+
"eval_samples_per_second": 7.322,
|
| 334 |
+
"eval_steps_per_second": 0.458,
|
| 335 |
+
"step": 200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 0.44,
|
| 339 |
+
"learning_rate": 3.437648009023905e-06,
|
| 340 |
+
"logits/chosen": 0.837793231010437,
|
| 341 |
+
"logits/rejected": 0.834112286567688,
|
| 342 |
+
"logps/chosen": -93.53672790527344,
|
| 343 |
+
"logps/rejected": -86.74687194824219,
|
| 344 |
+
"loss": 0.6926,
|
| 345 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 346 |
+
"rewards/chosen": -0.0022572961170226336,
|
| 347 |
+
"rewards/margins": 0.0016887232195585966,
|
| 348 |
+
"rewards/rejected": -0.00394601933658123,
|
| 349 |
+
"step": 210
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 0.46,
|
| 353 |
+
"learning_rate": 3.265574537815398e-06,
|
| 354 |
+
"logits/chosen": 0.7990659475326538,
|
| 355 |
+
"logits/rejected": 0.8434357643127441,
|
| 356 |
+
"logps/chosen": -101.04875946044922,
|
| 357 |
+
"logps/rejected": -95.13258361816406,
|
| 358 |
+
"loss": 0.6927,
|
| 359 |
+
"rewards/accuracies": 0.3187499940395355,
|
| 360 |
+
"rewards/chosen": -0.002797973807901144,
|
| 361 |
+
"rewards/margins": 0.0009291036985814571,
|
| 362 |
+
"rewards/rejected": -0.003727077739313245,
|
| 363 |
+
"step": 220
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 0.48,
|
| 367 |
+
"learning_rate": 3.089397338773569e-06,
|
| 368 |
+
"logits/chosen": 0.8693816065788269,
|
| 369 |
+
"logits/rejected": 0.8417167663574219,
|
| 370 |
+
"logps/chosen": -88.75157165527344,
|
| 371 |
+
"logps/rejected": -93.32563018798828,
|
| 372 |
+
"loss": 0.6923,
|
| 373 |
+
"rewards/accuracies": 0.35624998807907104,
|
| 374 |
+
"rewards/chosen": -0.0035379533655941486,
|
| 375 |
+
"rewards/margins": 0.001493574702180922,
|
| 376 |
+
"rewards/rejected": -0.005031527951359749,
|
| 377 |
+
"step": 230
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 0.5,
|
| 381 |
+
"learning_rate": 2.9100607788275547e-06,
|
| 382 |
+
"logits/chosen": 0.7980669736862183,
|
| 383 |
+
"logits/rejected": 0.8121516108512878,
|
| 384 |
+
"logps/chosen": -115.3040771484375,
|
| 385 |
+
"logps/rejected": -101.95747375488281,
|
| 386 |
+
"loss": 0.6917,
|
| 387 |
+
"rewards/accuracies": 0.36250001192092896,
|
| 388 |
+
"rewards/chosen": -0.0025563673116266727,
|
| 389 |
+
"rewards/margins": 0.0030108934734016657,
|
| 390 |
+
"rewards/rejected": -0.005567261017858982,
|
| 391 |
+
"step": 240
|
| 392 |
+
},
|
| 393 |
+
{
|
| 394 |
+
"epoch": 0.52,
|
| 395 |
+
"learning_rate": 2.72852616010567e-06,
|
| 396 |
+
"logits/chosen": 0.7765026688575745,
|
| 397 |
+
"logits/rejected": 0.8406999707221985,
|
| 398 |
+
"logps/chosen": -116.62477111816406,
|
| 399 |
+
"logps/rejected": -107.66845703125,
|
| 400 |
+
"loss": 0.692,
|
| 401 |
+
"rewards/accuracies": 0.36250001192092896,
|
| 402 |
+
"rewards/chosen": -0.0028718970715999603,
|
| 403 |
+
"rewards/margins": 0.0028597547207027674,
|
| 404 |
+
"rewards/rejected": -0.005731652025133371,
|
| 405 |
+
"step": 250
|
| 406 |
+
},
|
| 407 |
+
{
|
| 408 |
+
"epoch": 0.54,
|
| 409 |
+
"learning_rate": 2.5457665670441937e-06,
|
| 410 |
+
"logits/chosen": 0.7957627177238464,
|
| 411 |
+
"logits/rejected": 0.8751128911972046,
|
| 412 |
+
"logps/chosen": -99.27378845214844,
|
| 413 |
+
"logps/rejected": -88.92988586425781,
|
| 414 |
+
"loss": 0.6915,
|
| 415 |
+
"rewards/accuracies": 0.33125001192092896,
|
| 416 |
+
"rewards/chosen": -0.0027769659645855427,
|
| 417 |
+
"rewards/margins": 0.003127423347905278,
|
| 418 |
+
"rewards/rejected": -0.005904389079660177,
|
| 419 |
+
"step": 260
|
| 420 |
+
},
|
| 421 |
+
{
|
| 422 |
+
"epoch": 0.57,
|
| 423 |
+
"learning_rate": 2.3627616503391813e-06,
|
| 424 |
+
"logits/chosen": 0.8911062479019165,
|
| 425 |
+
"logits/rejected": 0.9384480714797974,
|
| 426 |
+
"logps/chosen": -114.53426361083984,
|
| 427 |
+
"logps/rejected": -83.79386901855469,
|
| 428 |
+
"loss": 0.6917,
|
| 429 |
+
"rewards/accuracies": 0.4124999940395355,
|
| 430 |
+
"rewards/chosen": -0.0030242924112826586,
|
| 431 |
+
"rewards/margins": 0.0038346436340361834,
|
| 432 |
+
"rewards/rejected": -0.006858936045318842,
|
| 433 |
+
"step": 270
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.59,
|
| 437 |
+
"learning_rate": 2.1804923757009885e-06,
|
| 438 |
+
"logits/chosen": 0.8408538699150085,
|
| 439 |
+
"logits/rejected": 0.839997410774231,
|
| 440 |
+
"logps/chosen": -121.15773010253906,
|
| 441 |
+
"logps/rejected": -109.73968505859375,
|
| 442 |
+
"loss": 0.6916,
|
| 443 |
+
"rewards/accuracies": 0.38749998807907104,
|
| 444 |
+
"rewards/chosen": -0.0034516516607254744,
|
| 445 |
+
"rewards/margins": 0.0034923548810184,
|
| 446 |
+
"rewards/rejected": -0.006944006774574518,
|
| 447 |
+
"step": 280
|
| 448 |
+
},
|
| 449 |
+
{
|
| 450 |
+
"epoch": 0.61,
|
| 451 |
+
"learning_rate": 1.9999357655598894e-06,
|
| 452 |
+
"logits/chosen": 0.825127124786377,
|
| 453 |
+
"logits/rejected": 0.8463200330734253,
|
| 454 |
+
"logps/chosen": -94.99199676513672,
|
| 455 |
+
"logps/rejected": -92.08692932128906,
|
| 456 |
+
"loss": 0.6919,
|
| 457 |
"rewards/accuracies": 0.2750000059604645,
|
| 458 |
+
"rewards/chosen": -0.005209151655435562,
|
| 459 |
+
"rewards/margins": 0.0012942428002133965,
|
| 460 |
+
"rewards/rejected": -0.006503394804894924,
|
| 461 |
+
"step": 290
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.63,
|
| 465 |
+
"learning_rate": 1.8220596619089576e-06,
|
| 466 |
+
"logits/chosen": 0.7896022796630859,
|
| 467 |
+
"logits/rejected": 0.8404830694198608,
|
| 468 |
+
"logps/chosen": -79.6166763305664,
|
| 469 |
+
"logps/rejected": -87.97811126708984,
|
| 470 |
+
"loss": 0.6917,
|
| 471 |
+
"rewards/accuracies": 0.3125,
|
| 472 |
+
"rewards/chosen": -0.002977000316604972,
|
| 473 |
+
"rewards/margins": 0.004023983143270016,
|
| 474 |
+
"rewards/rejected": -0.007000982761383057,
|
| 475 |
+
"step": 300
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 0.63,
|
| 479 |
+
"eval_logits/chosen": 0.787030041217804,
|
| 480 |
+
"eval_logits/rejected": 0.8057210445404053,
|
| 481 |
+
"eval_logps/chosen": -92.21893310546875,
|
| 482 |
+
"eval_logps/rejected": -94.9659194946289,
|
| 483 |
+
"eval_loss": 0.6917064189910889,
|
| 484 |
+
"eval_rewards/accuracies": 0.3100000023841858,
|
| 485 |
+
"eval_rewards/chosen": -0.004915657918900251,
|
| 486 |
+
"eval_rewards/margins": 0.0027734539471566677,
|
| 487 |
+
"eval_rewards/rejected": -0.0076891109347343445,
|
| 488 |
+
"eval_runtime": 273.1314,
|
| 489 |
+
"eval_samples_per_second": 7.322,
|
| 490 |
+
"eval_steps_per_second": 0.458,
|
| 491 |
+
"step": 300
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.65,
|
| 495 |
+
"learning_rate": 1.647817538357072e-06,
|
| 496 |
+
"logits/chosen": 0.797222912311554,
|
| 497 |
+
"logits/rejected": 0.8298647999763489,
|
| 498 |
+
"logps/chosen": -85.97596740722656,
|
| 499 |
+
"logps/rejected": -87.59102630615234,
|
| 500 |
+
"loss": 0.6915,
|
| 501 |
+
"rewards/accuracies": 0.33125001192092896,
|
| 502 |
+
"rewards/chosen": -0.005664575379341841,
|
| 503 |
+
"rewards/margins": 0.002779053058475256,
|
| 504 |
+
"rewards/rejected": -0.008443629369139671,
|
| 505 |
+
"step": 310
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 0.67,
|
| 509 |
+
"learning_rate": 1.4781433892011132e-06,
|
| 510 |
+
"logits/chosen": 0.7955067753791809,
|
| 511 |
+
"logits/rejected": 0.8460105657577515,
|
| 512 |
+
"logps/chosen": -88.00541687011719,
|
| 513 |
+
"logps/rejected": -91.3482437133789,
|
| 514 |
+
"loss": 0.6916,
|
| 515 |
+
"rewards/accuracies": 0.3687500059604645,
|
| 516 |
+
"rewards/chosen": -0.005240852013230324,
|
| 517 |
+
"rewards/margins": 0.003471267642453313,
|
| 518 |
+
"rewards/rejected": -0.008712120354175568,
|
| 519 |
+
"step": 320
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"epoch": 0.69,
|
| 523 |
+
"learning_rate": 1.3139467229135999e-06,
|
| 524 |
+
"logits/chosen": 0.7453655004501343,
|
| 525 |
+
"logits/rejected": 0.7835742235183716,
|
| 526 |
+
"logps/chosen": -101.49515533447266,
|
| 527 |
+
"logps/rejected": -102.06614685058594,
|
| 528 |
+
"loss": 0.6914,
|
| 529 |
+
"rewards/accuracies": 0.33125001192092896,
|
| 530 |
+
"rewards/chosen": -0.00533133652061224,
|
| 531 |
+
"rewards/margins": 0.004351903218775988,
|
| 532 |
+
"rewards/rejected": -0.00968323927372694,
|
| 533 |
+
"step": 330
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 0.71,
|
| 537 |
+
"learning_rate": 1.1561076868822756e-06,
|
| 538 |
+
"logits/chosen": 0.7904757857322693,
|
| 539 |
+
"logits/rejected": 0.8207653760910034,
|
| 540 |
+
"logps/chosen": -83.08015441894531,
|
| 541 |
+
"logps/rejected": -73.13031768798828,
|
| 542 |
+
"loss": 0.6908,
|
| 543 |
+
"rewards/accuracies": 0.29374998807907104,
|
| 544 |
+
"rewards/chosen": -0.002843918278813362,
|
| 545 |
+
"rewards/margins": 0.005075609777122736,
|
| 546 |
+
"rewards/rejected": -0.007919528521597385,
|
| 547 |
+
"step": 340
|
| 548 |
+
},
|
| 549 |
+
{
|
| 550 |
+
"epoch": 0.73,
|
| 551 |
+
"learning_rate": 1.0054723495346484e-06,
|
| 552 |
+
"logits/chosen": 0.8648473620414734,
|
| 553 |
+
"logits/rejected": 0.88224858045578,
|
| 554 |
+
"logps/chosen": -88.79496002197266,
|
| 555 |
+
"logps/rejected": -82.97408294677734,
|
| 556 |
+
"loss": 0.6911,
|
| 557 |
+
"rewards/accuracies": 0.36250001192092896,
|
| 558 |
+
"rewards/chosen": -0.0072964876890182495,
|
| 559 |
+
"rewards/margins": 0.003556302282959223,
|
| 560 |
+
"rewards/rejected": -0.01085279043763876,
|
| 561 |
+
"step": 350
|
| 562 |
+
},
|
| 563 |
+
{
|
| 564 |
+
"epoch": 0.75,
|
| 565 |
+
"learning_rate": 8.628481651367876e-07,
|
| 566 |
+
"logits/chosen": 0.7940279841423035,
|
| 567 |
+
"logits/rejected": 0.808300793170929,
|
| 568 |
+
"logps/chosen": -95.09690856933594,
|
| 569 |
+
"logps/rejected": -98.6402816772461,
|
| 570 |
+
"loss": 0.6911,
|
| 571 |
+
"rewards/accuracies": 0.375,
|
| 572 |
+
"rewards/chosen": -0.007194930221885443,
|
| 573 |
+
"rewards/margins": 0.004204220604151487,
|
| 574 |
+
"rewards/rejected": -0.01139915082603693,
|
| 575 |
+
"step": 360
|
| 576 |
+
},
|
| 577 |
+
{
|
| 578 |
+
"epoch": 0.77,
|
| 579 |
+
"learning_rate": 7.289996455765749e-07,
|
| 580 |
+
"logits/chosen": 0.8749688863754272,
|
| 581 |
+
"logits/rejected": 0.8549054861068726,
|
| 582 |
+
"logps/chosen": -107.79057312011719,
|
| 583 |
+
"logps/rejected": -104.4919662475586,
|
| 584 |
+
"loss": 0.6913,
|
| 585 |
+
"rewards/accuracies": 0.33125001192092896,
|
| 586 |
+
"rewards/chosen": -0.005484831985086203,
|
| 587 |
+
"rewards/margins": 0.0049090199172496796,
|
| 588 |
+
"rewards/rejected": -0.01039385236799717,
|
| 589 |
+
"step": 370
|
| 590 |
+
},
|
| 591 |
+
{
|
| 592 |
+
"epoch": 0.8,
|
| 593 |
+
"learning_rate": 6.046442623320145e-07,
|
| 594 |
+
"logits/chosen": 0.8186184167861938,
|
| 595 |
+
"logits/rejected": 0.8605905771255493,
|
| 596 |
+
"logps/chosen": -102.05201721191406,
|
| 597 |
+
"logps/rejected": -88.45350646972656,
|
| 598 |
+
"loss": 0.6905,
|
| 599 |
+
"rewards/accuracies": 0.35624998807907104,
|
| 600 |
+
"rewards/chosen": -0.006084176246076822,
|
| 601 |
+
"rewards/margins": 0.004532010294497013,
|
| 602 |
+
"rewards/rejected": -0.010616186074912548,
|
| 603 |
+
"step": 380
|
| 604 |
+
},
|
| 605 |
+
{
|
| 606 |
+
"epoch": 0.82,
|
| 607 |
+
"learning_rate": 4.904486005914027e-07,
|
| 608 |
+
"logits/chosen": 0.8054735064506531,
|
| 609 |
+
"logits/rejected": 0.8569159507751465,
|
| 610 |
+
"logps/chosen": -110.53129577636719,
|
| 611 |
+
"logps/rejected": -99.7970199584961,
|
| 612 |
+
"loss": 0.6905,
|
| 613 |
+
"rewards/accuracies": 0.35624998807907104,
|
| 614 |
+
"rewards/chosen": -0.0049868240021169186,
|
| 615 |
+
"rewards/margins": 0.005900658201426268,
|
| 616 |
+
"rewards/rejected": -0.01088748313486576,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 0.84,
|
| 621 |
+
"learning_rate": 3.8702478614051353e-07,
|
| 622 |
+
"logits/chosen": 0.8212454915046692,
|
| 623 |
+
"logits/rejected": 0.8515514135360718,
|
| 624 |
+
"logps/chosen": -100.35879516601562,
|
| 625 |
+
"logps/rejected": -103.1037368774414,
|
| 626 |
+
"loss": 0.6905,
|
| 627 |
+
"rewards/accuracies": 0.3375000059604645,
|
| 628 |
+
"rewards/chosen": -0.006732915993779898,
|
| 629 |
+
"rewards/margins": 0.004675927106291056,
|
| 630 |
+
"rewards/rejected": -0.011408843100070953,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 0.84,
|
| 635 |
+
"eval_logits/chosen": 0.7826990485191345,
|
| 636 |
+
"eval_logits/rejected": 0.8012421727180481,
|
| 637 |
+
"eval_logps/chosen": -92.42469787597656,
|
| 638 |
+
"eval_logps/rejected": -95.25089263916016,
|
| 639 |
+
"eval_loss": 0.6913294792175293,
|
| 640 |
+
"eval_rewards/accuracies": 0.328000009059906,
|
| 641 |
+
"eval_rewards/chosen": -0.00697335647419095,
|
| 642 |
+
"eval_rewards/margins": 0.0035653903614729643,
|
| 643 |
+
"eval_rewards/rejected": -0.010538745671510696,
|
| 644 |
+
"eval_runtime": 273.6535,
|
| 645 |
+
"eval_samples_per_second": 7.309,
|
| 646 |
+
"eval_steps_per_second": 0.457,
|
| 647 |
+
"step": 400
|
| 648 |
+
},
|
| 649 |
+
{
|
| 650 |
+
"epoch": 0.86,
|
| 651 |
+
"learning_rate": 2.9492720416985004e-07,
|
| 652 |
+
"logits/chosen": 0.8186850547790527,
|
| 653 |
+
"logits/rejected": 0.8151271939277649,
|
| 654 |
+
"logps/chosen": -95.82102966308594,
|
| 655 |
+
"logps/rejected": -80.586669921875,
|
| 656 |
+
"loss": 0.6916,
|
| 657 |
+
"rewards/accuracies": 0.3375000059604645,
|
| 658 |
+
"rewards/chosen": -0.00794359389692545,
|
| 659 |
+
"rewards/margins": 0.003056485904380679,
|
| 660 |
+
"rewards/rejected": -0.011000080034136772,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.88,
|
| 665 |
+
"learning_rate": 2.1464952759020857e-07,
|
| 666 |
+
"logits/chosen": 0.7571959495544434,
|
| 667 |
+
"logits/rejected": 0.8163139224052429,
|
| 668 |
+
"logps/chosen": -95.65677642822266,
|
| 669 |
+
"logps/rejected": -97.86299896240234,
|
| 670 |
+
"loss": 0.6915,
|
| 671 |
+
"rewards/accuracies": 0.3125,
|
| 672 |
+
"rewards/chosen": -0.00859010498970747,
|
| 673 |
+
"rewards/margins": 0.0028384437318891287,
|
| 674 |
+
"rewards/rejected": -0.011428548023104668,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 0.9,
|
| 679 |
+
"learning_rate": 1.4662207078575685e-07,
|
| 680 |
+
"logits/chosen": 0.8645572662353516,
|
| 681 |
+
"logits/rejected": 0.887597918510437,
|
| 682 |
+
"logps/chosen": -83.41182708740234,
|
| 683 |
+
"logps/rejected": -90.17640686035156,
|
| 684 |
+
"loss": 0.6913,
|
| 685 |
+
"rewards/accuracies": 0.39375001192092896,
|
| 686 |
+
"rewards/chosen": -0.006618577986955643,
|
| 687 |
+
"rewards/margins": 0.004637080244719982,
|
| 688 |
+
"rewards/rejected": -0.011255658231675625,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 0.92,
|
| 693 |
+
"learning_rate": 9.120948298936422e-08,
|
| 694 |
+
"logits/chosen": 0.7825466394424438,
|
| 695 |
+
"logits/rejected": 0.8328782916069031,
|
| 696 |
+
"logps/chosen": -109.79942321777344,
|
| 697 |
+
"logps/rejected": -82.07637023925781,
|
| 698 |
+
"loss": 0.6911,
|
| 699 |
+
"rewards/accuracies": 0.34375,
|
| 700 |
+
"rewards/chosen": -0.005336672533303499,
|
| 701 |
+
"rewards/margins": 0.005987245589494705,
|
| 702 |
+
"rewards/rejected": -0.011323917657136917,
|
| 703 |
+
"step": 440
|
| 704 |
},
|
| 705 |
{
|
| 706 |
"epoch": 0.94,
|
| 707 |
+
"learning_rate": 4.870879364444109e-08,
|
| 708 |
+
"logits/chosen": 0.774901270866394,
|
| 709 |
+
"logits/rejected": 0.8499285578727722,
|
| 710 |
+
"logps/chosen": -81.24304962158203,
|
| 711 |
+
"logps/rejected": -71.88755798339844,
|
| 712 |
+
"loss": 0.6911,
|
| 713 |
+
"rewards/accuracies": 0.29374998807907104,
|
| 714 |
+
"rewards/chosen": -0.005921828560531139,
|
| 715 |
+
"rewards/margins": 0.0035197760444134474,
|
| 716 |
+
"rewards/rejected": -0.00944160483777523,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 0.96,
|
| 721 |
+
"learning_rate": 1.93478202307823e-08,
|
| 722 |
+
"logits/chosen": 0.8049672842025757,
|
| 723 |
+
"logits/rejected": 0.8570533990859985,
|
| 724 |
+
"logps/chosen": -105.7918930053711,
|
| 725 |
+
"logps/rejected": -87.53938293457031,
|
| 726 |
+
"loss": 0.6907,
|
| 727 |
+
"rewards/accuracies": 0.3187499940395355,
|
| 728 |
+
"rewards/chosen": -0.005085950251668692,
|
| 729 |
+
"rewards/margins": 0.0068057505413889885,
|
| 730 |
+
"rewards/rejected": -0.011891700327396393,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 0.98,
|
| 735 |
+
"learning_rate": 3.283947088983663e-09,
|
| 736 |
+
"logits/chosen": 0.8561135530471802,
|
| 737 |
+
"logits/rejected": 0.8520036935806274,
|
| 738 |
+
"logps/chosen": -115.6650390625,
|
| 739 |
+
"logps/rejected": -110.88753509521484,
|
| 740 |
+
"loss": 0.6903,
|
| 741 |
+
"rewards/accuracies": 0.39375001192092896,
|
| 742 |
+
"rewards/chosen": -0.006030657794326544,
|
| 743 |
+
"rewards/margins": 0.00688832625746727,
|
| 744 |
+
"rewards/rejected": -0.012918984517455101,
|
| 745 |
+
"step": 470
|
| 746 |
},
|
| 747 |
{
|
| 748 |
+
"epoch": 1.0,
|
| 749 |
+
"step": 477,
|
| 750 |
"total_flos": 0.0,
|
| 751 |
+
"train_loss": 0.6920521804121805,
|
| 752 |
+
"train_runtime": 6813.2628,
|
| 753 |
+
"train_samples_per_second": 4.486,
|
| 754 |
+
"train_steps_per_second": 0.07
|
| 755 |
}
|
| 756 |
],
|
| 757 |
"logging_steps": 10,
|
| 758 |
+
"max_steps": 477,
|
| 759 |
"num_input_tokens_seen": 0,
|
| 760 |
"num_train_epochs": 1,
|
| 761 |
"save_steps": 100,
|