Commit
•
cec2911
1
Parent(s):
44236d0
Model save
Browse files- README.md +74 -0
- adapter_model.safetensors +1 -1
- all_results.json +21 -0
- eval_results.json +16 -0
- train_results.json +8 -0
- trainer_state.json +412 -0
README.md
ADDED
|
@@ -0,0 +1,74 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 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-gpo-test-longest-iter-v1-0
|
| 11 |
+
results: []
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
+
|
| 17 |
+
# phi-2-gpo-test-longest-iter-v1-0
|
| 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: 0.0004
|
| 22 |
+
- Rewards/chosen: 0.0012
|
| 23 |
+
- Rewards/rejected: 0.0010
|
| 24 |
+
- Rewards/accuracies: 0.4995
|
| 25 |
+
- Rewards/margins: 0.0002
|
| 26 |
+
- Logps/rejected: -233.4380
|
| 27 |
+
- Logps/chosen: -256.4973
|
| 28 |
+
- Logits/rejected: 0.8990
|
| 29 |
+
- Logits/chosen: 0.8417
|
| 30 |
+
|
| 31 |
+
## Model description
|
| 32 |
+
|
| 33 |
+
More information needed
|
| 34 |
+
|
| 35 |
+
## Intended uses & limitations
|
| 36 |
+
|
| 37 |
+
More information needed
|
| 38 |
+
|
| 39 |
+
## Training and evaluation data
|
| 40 |
+
|
| 41 |
+
More information needed
|
| 42 |
+
|
| 43 |
+
## Training procedure
|
| 44 |
+
|
| 45 |
+
### Training hyperparameters
|
| 46 |
+
|
| 47 |
+
The following hyperparameters were used during training:
|
| 48 |
+
- learning_rate: 5e-06
|
| 49 |
+
- train_batch_size: 4
|
| 50 |
+
- eval_batch_size: 4
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- gradient_accumulation_steps: 4
|
| 54 |
+
- total_train_batch_size: 16
|
| 55 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 56 |
+
- lr_scheduler_type: cosine
|
| 57 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 58 |
+
- num_epochs: 4
|
| 59 |
+
|
| 60 |
+
### Training results
|
| 61 |
+
|
| 62 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 63 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 64 |
+
| 0.0003 | 1.6 | 100 | 0.0004 | 0.0006 | 0.0004 | 0.4855 | 0.0002 | -233.5017 | -256.5565 | 0.8960 | 0.8387 |
|
| 65 |
+
| 0.0003 | 3.2 | 200 | 0.0004 | 0.0013 | 0.0009 | 0.5100 | 0.0004 | -233.4492 | -256.4811 | 0.8984 | 0.8412 |
|
| 66 |
+
|
| 67 |
+
|
| 68 |
+
### Framework versions
|
| 69 |
+
|
| 70 |
+
- PEFT 0.7.1
|
| 71 |
+
- Transformers 4.36.2
|
| 72 |
+
- Pytorch 2.2.1+cu121
|
| 73 |
+
- Datasets 2.14.6
|
| 74 |
+
- Tokenizers 0.15.2
|
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 sha256:6c6111481cb1a2c7bcd128ee87e2115a05b3817370404ebd778cc9f4acc408bd
|
| 3 |
size 41977616
|
all_results.json
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.97,
|
| 3 |
+
"eval_logits/chosen": 0.8417125940322876,
|
| 4 |
+
"eval_logits/rejected": 0.8990198373794556,
|
| 5 |
+
"eval_logps/chosen": -256.497314453125,
|
| 6 |
+
"eval_logps/rejected": -233.4380340576172,
|
| 7 |
+
"eval_loss": 0.0003914303961209953,
|
| 8 |
+
"eval_rewards/accuracies": 0.49950000643730164,
|
| 9 |
+
"eval_rewards/chosen": 0.001168824383057654,
|
| 10 |
+
"eval_rewards/margins": 0.00016647118900436908,
|
| 11 |
+
"eval_rewards/rejected": 0.001002353266812861,
|
| 12 |
+
"eval_runtime": 561.9533,
|
| 13 |
+
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 3.559,
|
| 15 |
+
"eval_steps_per_second": 0.89,
|
| 16 |
+
"train_loss": 0.0003399105843263603,
|
| 17 |
+
"train_runtime": 4613.339,
|
| 18 |
+
"train_samples": 61135,
|
| 19 |
+
"train_samples_per_second": 0.867,
|
| 20 |
+
"train_steps_per_second": 0.054
|
| 21 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.97,
|
| 3 |
+
"eval_logits/chosen": 0.8417125940322876,
|
| 4 |
+
"eval_logits/rejected": 0.8990198373794556,
|
| 5 |
+
"eval_logps/chosen": -256.497314453125,
|
| 6 |
+
"eval_logps/rejected": -233.4380340576172,
|
| 7 |
+
"eval_loss": 0.0003914303961209953,
|
| 8 |
+
"eval_rewards/accuracies": 0.49950000643730164,
|
| 9 |
+
"eval_rewards/chosen": 0.001168824383057654,
|
| 10 |
+
"eval_rewards/margins": 0.00016647118900436908,
|
| 11 |
+
"eval_rewards/rejected": 0.001002353266812861,
|
| 12 |
+
"eval_runtime": 561.9533,
|
| 13 |
+
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 3.559,
|
| 15 |
+
"eval_steps_per_second": 0.89
|
| 16 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.97,
|
| 3 |
+
"train_loss": 0.0003399105843263603,
|
| 4 |
+
"train_runtime": 4613.339,
|
| 5 |
+
"train_samples": 61135,
|
| 6 |
+
"train_samples_per_second": 0.867,
|
| 7 |
+
"train_steps_per_second": 0.054
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,412 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.968,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 248,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.02,
|
| 13 |
+
"learning_rate": 2.0000000000000002e-07,
|
| 14 |
+
"logits/chosen": 0.9600980877876282,
|
| 15 |
+
"logits/rejected": 1.094868540763855,
|
| 16 |
+
"logps/chosen": -119.03445434570312,
|
| 17 |
+
"logps/rejected": -112.18000030517578,
|
| 18 |
+
"loss": 0.0003,
|
| 19 |
+
"rewards/accuracies": 0.0,
|
| 20 |
+
"rewards/chosen": 0.0,
|
| 21 |
+
"rewards/margins": 0.0,
|
| 22 |
+
"rewards/rejected": 0.0,
|
| 23 |
+
"step": 1
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.16,
|
| 27 |
+
"learning_rate": 2.0000000000000003e-06,
|
| 28 |
+
"logits/chosen": 0.8413593769073486,
|
| 29 |
+
"logits/rejected": 0.813139796257019,
|
| 30 |
+
"logps/chosen": -142.51600646972656,
|
| 31 |
+
"logps/rejected": -130.04800415039062,
|
| 32 |
+
"loss": 0.0004,
|
| 33 |
+
"rewards/accuracies": 0.3194444477558136,
|
| 34 |
+
"rewards/chosen": 0.0009210502030327916,
|
| 35 |
+
"rewards/margins": -8.903745765564963e-05,
|
| 36 |
+
"rewards/rejected": 0.0010100876679643989,
|
| 37 |
+
"step": 10
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.32,
|
| 41 |
+
"learning_rate": 4.000000000000001e-06,
|
| 42 |
+
"logits/chosen": 0.8089853525161743,
|
| 43 |
+
"logits/rejected": 0.8187138438224792,
|
| 44 |
+
"logps/chosen": -155.18431091308594,
|
| 45 |
+
"logps/rejected": -143.44085693359375,
|
| 46 |
+
"loss": 0.0004,
|
| 47 |
+
"rewards/accuracies": 0.40625,
|
| 48 |
+
"rewards/chosen": 0.0007359608425758779,
|
| 49 |
+
"rewards/margins": -0.0004602448898367584,
|
| 50 |
+
"rewards/rejected": 0.001196205848827958,
|
| 51 |
+
"step": 20
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.48,
|
| 55 |
+
"learning_rate": 4.993800445762451e-06,
|
| 56 |
+
"logits/chosen": 0.8762779235839844,
|
| 57 |
+
"logits/rejected": 0.9211899638175964,
|
| 58 |
+
"logps/chosen": -154.8330535888672,
|
| 59 |
+
"logps/rejected": -141.16299438476562,
|
| 60 |
+
"loss": 0.0004,
|
| 61 |
+
"rewards/accuracies": 0.375,
|
| 62 |
+
"rewards/chosen": -1.3224826943769585e-05,
|
| 63 |
+
"rewards/margins": -0.00042751478031277657,
|
| 64 |
+
"rewards/rejected": 0.0004142899124417454,
|
| 65 |
+
"step": 30
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.64,
|
| 69 |
+
"learning_rate": 4.944388344834205e-06,
|
| 70 |
+
"logits/chosen": 0.8115983009338379,
|
| 71 |
+
"logits/rejected": 0.8376752734184265,
|
| 72 |
+
"logps/chosen": -172.55001831054688,
|
| 73 |
+
"logps/rejected": -157.711669921875,
|
| 74 |
+
"loss": 0.0004,
|
| 75 |
+
"rewards/accuracies": 0.4124999940395355,
|
| 76 |
+
"rewards/chosen": 0.00026369161787442863,
|
| 77 |
+
"rewards/margins": 4.611946133081801e-05,
|
| 78 |
+
"rewards/rejected": 0.0002175722038373351,
|
| 79 |
+
"step": 40
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.8,
|
| 83 |
+
"learning_rate": 4.8465431931347904e-06,
|
| 84 |
+
"logits/chosen": 0.8291877508163452,
|
| 85 |
+
"logits/rejected": 0.8113203048706055,
|
| 86 |
+
"logps/chosen": -160.0177764892578,
|
| 87 |
+
"logps/rejected": -146.55140686035156,
|
| 88 |
+
"loss": 0.0004,
|
| 89 |
+
"rewards/accuracies": 0.375,
|
| 90 |
+
"rewards/chosen": -5.383044481277466e-07,
|
| 91 |
+
"rewards/margins": -8.223028999054804e-05,
|
| 92 |
+
"rewards/rejected": 8.169189095497131e-05,
|
| 93 |
+
"step": 50
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.96,
|
| 97 |
+
"learning_rate": 4.702203692102539e-06,
|
| 98 |
+
"logits/chosen": 0.7359780073165894,
|
| 99 |
+
"logits/rejected": 0.7216537594795227,
|
| 100 |
+
"logps/chosen": -146.64401245117188,
|
| 101 |
+
"logps/rejected": -134.4973907470703,
|
| 102 |
+
"loss": 0.0003,
|
| 103 |
+
"rewards/accuracies": 0.5062500238418579,
|
| 104 |
+
"rewards/chosen": 0.0005743899382650852,
|
| 105 |
+
"rewards/margins": 0.00039931651554070413,
|
| 106 |
+
"rewards/rejected": 0.00017507340817246586,
|
| 107 |
+
"step": 60
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 1.12,
|
| 111 |
+
"learning_rate": 4.514229781074239e-06,
|
| 112 |
+
"logits/chosen": 0.8711406588554382,
|
| 113 |
+
"logits/rejected": 0.8447128534317017,
|
| 114 |
+
"logps/chosen": -146.93917846679688,
|
| 115 |
+
"logps/rejected": -137.2763214111328,
|
| 116 |
+
"loss": 0.0003,
|
| 117 |
+
"rewards/accuracies": 0.4625000059604645,
|
| 118 |
+
"rewards/chosen": -9.175995364785194e-05,
|
| 119 |
+
"rewards/margins": 0.00013136237976141274,
|
| 120 |
+
"rewards/rejected": -0.00022312237706501037,
|
| 121 |
+
"step": 70
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 1.28,
|
| 125 |
+
"learning_rate": 4.286345970517195e-06,
|
| 126 |
+
"logits/chosen": 0.8513940572738647,
|
| 127 |
+
"logits/rejected": 0.8943384289741516,
|
| 128 |
+
"logps/chosen": -156.4779510498047,
|
| 129 |
+
"logps/rejected": -141.64044189453125,
|
| 130 |
+
"loss": 0.0003,
|
| 131 |
+
"rewards/accuracies": 0.41874998807907104,
|
| 132 |
+
"rewards/chosen": -0.0007812186959199607,
|
| 133 |
+
"rewards/margins": 0.0002858097432181239,
|
| 134 |
+
"rewards/rejected": -0.0010670283809304237,
|
| 135 |
+
"step": 80
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 1.44,
|
| 139 |
+
"learning_rate": 4.023067544670082e-06,
|
| 140 |
+
"logits/chosen": 0.8046412467956543,
|
| 141 |
+
"logits/rejected": 0.8172038793563843,
|
| 142 |
+
"logps/chosen": -159.82638549804688,
|
| 143 |
+
"logps/rejected": -148.19137573242188,
|
| 144 |
+
"loss": 0.0003,
|
| 145 |
+
"rewards/accuracies": 0.4749999940395355,
|
| 146 |
+
"rewards/chosen": -0.0005336708854883909,
|
| 147 |
+
"rewards/margins": 0.000577736587729305,
|
| 148 |
+
"rewards/rejected": -0.0011114075314253569,
|
| 149 |
+
"step": 90
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 1.6,
|
| 153 |
+
"learning_rate": 3.7296110958116845e-06,
|
| 154 |
+
"logits/chosen": 0.7820717096328735,
|
| 155 |
+
"logits/rejected": 0.8074045181274414,
|
| 156 |
+
"logps/chosen": -149.77316284179688,
|
| 157 |
+
"logps/rejected": -137.23370361328125,
|
| 158 |
+
"loss": 0.0003,
|
| 159 |
+
"rewards/accuracies": 0.42500001192092896,
|
| 160 |
+
"rewards/chosen": -0.00042833946645259857,
|
| 161 |
+
"rewards/margins": 0.0002858256921172142,
|
| 162 |
+
"rewards/rejected": -0.0007141652167774737,
|
| 163 |
+
"step": 100
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 1.6,
|
| 167 |
+
"eval_logits/chosen": 0.8386867046356201,
|
| 168 |
+
"eval_logits/rejected": 0.8959885835647583,
|
| 169 |
+
"eval_logps/chosen": -256.5564880371094,
|
| 170 |
+
"eval_logps/rejected": -233.50169372558594,
|
| 171 |
+
"eval_loss": 0.0003867279156111181,
|
| 172 |
+
"eval_rewards/accuracies": 0.4855000078678131,
|
| 173 |
+
"eval_rewards/chosen": 0.0005771049181930721,
|
| 174 |
+
"eval_rewards/margins": 0.00021130235109012574,
|
| 175 |
+
"eval_rewards/rejected": 0.00036580252344720066,
|
| 176 |
+
"eval_runtime": 895.2516,
|
| 177 |
+
"eval_samples_per_second": 2.234,
|
| 178 |
+
"eval_steps_per_second": 0.559,
|
| 179 |
+
"step": 100
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 1.76,
|
| 183 |
+
"learning_rate": 3.4117911628292944e-06,
|
| 184 |
+
"logits/chosen": 0.8274463415145874,
|
| 185 |
+
"logits/rejected": 0.7994173765182495,
|
| 186 |
+
"logps/chosen": -160.32693481445312,
|
| 187 |
+
"logps/rejected": -144.13430786132812,
|
| 188 |
+
"loss": 0.0004,
|
| 189 |
+
"rewards/accuracies": 0.3687500059604645,
|
| 190 |
+
"rewards/chosen": -0.0019955369643867016,
|
| 191 |
+
"rewards/margins": -0.0002371647860854864,
|
| 192 |
+
"rewards/rejected": -0.0017583720618858933,
|
| 193 |
+
"step": 110
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 1.92,
|
| 197 |
+
"learning_rate": 3.075905022087675e-06,
|
| 198 |
+
"logits/chosen": 0.8538025617599487,
|
| 199 |
+
"logits/rejected": 0.8362048268318176,
|
| 200 |
+
"logps/chosen": -144.97744750976562,
|
| 201 |
+
"logps/rejected": -132.49935913085938,
|
| 202 |
+
"loss": 0.0003,
|
| 203 |
+
"rewards/accuracies": 0.4375,
|
| 204 |
+
"rewards/chosen": -0.0011052186600863934,
|
| 205 |
+
"rewards/margins": 0.00014151711366139352,
|
| 206 |
+
"rewards/rejected": -0.0012467356864362955,
|
| 207 |
+
"step": 120
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 2.08,
|
| 211 |
+
"learning_rate": 2.728607913349464e-06,
|
| 212 |
+
"logits/chosen": 0.8091080784797668,
|
| 213 |
+
"logits/rejected": 0.7883086800575256,
|
| 214 |
+
"logps/chosen": -168.44651794433594,
|
| 215 |
+
"logps/rejected": -154.5172882080078,
|
| 216 |
+
"loss": 0.0003,
|
| 217 |
+
"rewards/accuracies": 0.4124999940395355,
|
| 218 |
+
"rewards/chosen": -0.001501282094977796,
|
| 219 |
+
"rewards/margins": 0.00047290659858845174,
|
| 220 |
+
"rewards/rejected": -0.001974188955500722,
|
| 221 |
+
"step": 130
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 2.24,
|
| 225 |
+
"learning_rate": 2.376781173017589e-06,
|
| 226 |
+
"logits/chosen": 0.8410237431526184,
|
| 227 |
+
"logits/rejected": 0.8125711679458618,
|
| 228 |
+
"logps/chosen": -135.33348083496094,
|
| 229 |
+
"logps/rejected": -125.95147705078125,
|
| 230 |
+
"loss": 0.0003,
|
| 231 |
+
"rewards/accuracies": 0.518750011920929,
|
| 232 |
+
"rewards/chosen": -0.0011694638524204493,
|
| 233 |
+
"rewards/margins": 0.0012481597950682044,
|
| 234 |
+
"rewards/rejected": -0.002417623531073332,
|
| 235 |
+
"step": 140
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 2.4,
|
| 239 |
+
"learning_rate": 2.0273958875043877e-06,
|
| 240 |
+
"logits/chosen": 0.8500027656555176,
|
| 241 |
+
"logits/rejected": 0.8616348505020142,
|
| 242 |
+
"logps/chosen": -163.2034454345703,
|
| 243 |
+
"logps/rejected": -147.33352661132812,
|
| 244 |
+
"loss": 0.0003,
|
| 245 |
+
"rewards/accuracies": 0.4312500059604645,
|
| 246 |
+
"rewards/chosen": -0.0014132513897493482,
|
| 247 |
+
"rewards/margins": 0.0005158716812729836,
|
| 248 |
+
"rewards/rejected": -0.0019291229546070099,
|
| 249 |
+
"step": 150
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 2.56,
|
| 253 |
+
"learning_rate": 1.6873747682962393e-06,
|
| 254 |
+
"logits/chosen": 0.7654728293418884,
|
| 255 |
+
"logits/rejected": 0.7762084007263184,
|
| 256 |
+
"logps/chosen": -164.34304809570312,
|
| 257 |
+
"logps/rejected": -151.01991271972656,
|
| 258 |
+
"loss": 0.0003,
|
| 259 |
+
"rewards/accuracies": 0.46875,
|
| 260 |
+
"rewards/chosen": -0.002124720485880971,
|
| 261 |
+
"rewards/margins": 0.0006703810067847371,
|
| 262 |
+
"rewards/rejected": -0.00279510160908103,
|
| 263 |
+
"step": 160
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"epoch": 2.72,
|
| 267 |
+
"learning_rate": 1.363454985517803e-06,
|
| 268 |
+
"logits/chosen": 0.8323400616645813,
|
| 269 |
+
"logits/rejected": 0.8276194334030151,
|
| 270 |
+
"logps/chosen": -139.4966278076172,
|
| 271 |
+
"logps/rejected": -126.7676010131836,
|
| 272 |
+
"loss": 0.0003,
|
| 273 |
+
"rewards/accuracies": 0.45625001192092896,
|
| 274 |
+
"rewards/chosen": -0.002051023533567786,
|
| 275 |
+
"rewards/margins": 0.0008618654683232307,
|
| 276 |
+
"rewards/rejected": -0.002912889001891017,
|
| 277 |
+
"step": 170
|
| 278 |
+
},
|
| 279 |
+
{
|
| 280 |
+
"epoch": 2.88,
|
| 281 |
+
"learning_rate": 1.062054677808238e-06,
|
| 282 |
+
"logits/chosen": 0.8200883865356445,
|
| 283 |
+
"logits/rejected": 0.8247148394584656,
|
| 284 |
+
"logps/chosen": -147.3248291015625,
|
| 285 |
+
"logps/rejected": -131.50941467285156,
|
| 286 |
+
"loss": 0.0004,
|
| 287 |
+
"rewards/accuracies": 0.39375001192092896,
|
| 288 |
+
"rewards/chosen": -0.002400606404989958,
|
| 289 |
+
"rewards/margins": -0.0002633151598274708,
|
| 290 |
+
"rewards/rejected": -0.0021372910123318434,
|
| 291 |
+
"step": 180
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"epoch": 3.04,
|
| 295 |
+
"learning_rate": 7.891457834794711e-07,
|
| 296 |
+
"logits/chosen": 0.7988116145133972,
|
| 297 |
+
"logits/rejected": 0.7790786027908325,
|
| 298 |
+
"logps/chosen": -157.28366088867188,
|
| 299 |
+
"logps/rejected": -148.06373596191406,
|
| 300 |
+
"loss": 0.0003,
|
| 301 |
+
"rewards/accuracies": 0.48124998807907104,
|
| 302 |
+
"rewards/chosen": -0.0013300395803526044,
|
| 303 |
+
"rewards/margins": 0.0013687892351299524,
|
| 304 |
+
"rewards/rejected": -0.002698828699067235,
|
| 305 |
+
"step": 190
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 3.2,
|
| 309 |
+
"learning_rate": 5.501357126768117e-07,
|
| 310 |
+
"logits/chosen": 0.8291190266609192,
|
| 311 |
+
"logits/rejected": 0.8441616296768188,
|
| 312 |
+
"logps/chosen": -179.22891235351562,
|
| 313 |
+
"logps/rejected": -165.08668518066406,
|
| 314 |
+
"loss": 0.0003,
|
| 315 |
+
"rewards/accuracies": 0.36250001192092896,
|
| 316 |
+
"rewards/chosen": -0.0026950170285999775,
|
| 317 |
+
"rewards/margins": 0.0003695812192745507,
|
| 318 |
+
"rewards/rejected": -0.0030645981896668673,
|
| 319 |
+
"step": 200
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"epoch": 3.2,
|
| 323 |
+
"eval_logits/chosen": 0.8411728143692017,
|
| 324 |
+
"eval_logits/rejected": 0.8983786106109619,
|
| 325 |
+
"eval_logps/chosen": -256.48114013671875,
|
| 326 |
+
"eval_logps/rejected": -233.4491729736328,
|
| 327 |
+
"eval_loss": 0.0003846254840027541,
|
| 328 |
+
"eval_rewards/accuracies": 0.5099999904632568,
|
| 329 |
+
"eval_rewards/chosen": 0.0013305767206475139,
|
| 330 |
+
"eval_rewards/margins": 0.00043944927165284753,
|
| 331 |
+
"eval_rewards/rejected": 0.0008911274489946663,
|
| 332 |
+
"eval_runtime": 662.5153,
|
| 333 |
+
"eval_samples_per_second": 3.019,
|
| 334 |
+
"eval_steps_per_second": 0.755,
|
| 335 |
+
"step": 200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 3.36,
|
| 339 |
+
"learning_rate": 3.4976020508682345e-07,
|
| 340 |
+
"logits/chosen": 0.7520900964736938,
|
| 341 |
+
"logits/rejected": 0.7521581053733826,
|
| 342 |
+
"logps/chosen": -142.36843872070312,
|
| 343 |
+
"logps/rejected": -131.4374542236328,
|
| 344 |
+
"loss": 0.0003,
|
| 345 |
+
"rewards/accuracies": 0.4749999940395355,
|
| 346 |
+
"rewards/chosen": -0.0021503984462469816,
|
| 347 |
+
"rewards/margins": 0.0008300153422169387,
|
| 348 |
+
"rewards/rejected": -0.0029804136138409376,
|
| 349 |
+
"step": 210
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 3.52,
|
| 353 |
+
"learning_rate": 1.9198949610721273e-07,
|
| 354 |
+
"logits/chosen": 0.8309981226921082,
|
| 355 |
+
"logits/rejected": 0.811602771282196,
|
| 356 |
+
"logps/chosen": -150.01573181152344,
|
| 357 |
+
"logps/rejected": -136.91067504882812,
|
| 358 |
+
"loss": 0.0003,
|
| 359 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 360 |
+
"rewards/chosen": -0.00250506354495883,
|
| 361 |
+
"rewards/margins": 0.0008244045311585069,
|
| 362 |
+
"rewards/rejected": -0.0033294681925326586,
|
| 363 |
+
"step": 220
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 3.68,
|
| 367 |
+
"learning_rate": 7.994965069994143e-08,
|
| 368 |
+
"logits/chosen": 0.8669425249099731,
|
| 369 |
+
"logits/rejected": 0.880296528339386,
|
| 370 |
+
"logps/chosen": -159.811767578125,
|
| 371 |
+
"logps/rejected": -145.57418823242188,
|
| 372 |
+
"loss": 0.0003,
|
| 373 |
+
"rewards/accuracies": 0.512499988079071,
|
| 374 |
+
"rewards/chosen": -0.0020190321374684572,
|
| 375 |
+
"rewards/margins": 0.0008607100462540984,
|
| 376 |
+
"rewards/rejected": -0.002879742067307234,
|
| 377 |
+
"step": 230
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 3.84,
|
| 381 |
+
"learning_rate": 1.5860623616664183e-08,
|
| 382 |
+
"logits/chosen": 0.7746256589889526,
|
| 383 |
+
"logits/rejected": 0.7747436761856079,
|
| 384 |
+
"logps/chosen": -154.1355743408203,
|
| 385 |
+
"logps/rejected": -141.05926513671875,
|
| 386 |
+
"loss": 0.0003,
|
| 387 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 388 |
+
"rewards/chosen": -0.001991902943700552,
|
| 389 |
+
"rewards/margins": 0.000743478478398174,
|
| 390 |
+
"rewards/rejected": -0.0027353812474757433,
|
| 391 |
+
"step": 240
|
| 392 |
+
},
|
| 393 |
+
{
|
| 394 |
+
"epoch": 3.97,
|
| 395 |
+
"step": 248,
|
| 396 |
+
"total_flos": 0.0,
|
| 397 |
+
"train_loss": 0.0003399105843263603,
|
| 398 |
+
"train_runtime": 4613.339,
|
| 399 |
+
"train_samples_per_second": 0.867,
|
| 400 |
+
"train_steps_per_second": 0.054
|
| 401 |
+
}
|
| 402 |
+
],
|
| 403 |
+
"logging_steps": 10,
|
| 404 |
+
"max_steps": 248,
|
| 405 |
+
"num_input_tokens_seen": 0,
|
| 406 |
+
"num_train_epochs": 4,
|
| 407 |
+
"save_steps": 100,
|
| 408 |
+
"total_flos": 0.0,
|
| 409 |
+
"train_batch_size": 4,
|
| 410 |
+
"trial_name": null,
|
| 411 |
+
"trial_params": null
|
| 412 |
+
}
|