Model save
Browse files- README.md +78 -0
- adapter_model.safetensors +1 -1
- all_results.json +21 -0
- eval_results.json +16 -0
- runs/Feb02_15-24-17_ip-26-0-165-38/events.out.tfevents.1706887790.ip-26-0-165-38.1953627.0 +2 -2
- runs/Feb02_15-24-17_ip-26-0-165-38/events.out.tfevents.1706892789.ip-26-0-165-38.1953627.1 +3 -0
- train_results.json +8 -0
- trainer_state.json +766 -0
README.md
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
library_name: peft
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
base_model: mistralai/Mistral-7B-v0.1
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-7b-dpo-qlora-fix
|
| 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 |
+
# zephyr-7b-dpo-qlora-fix
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the None dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.5279
|
| 22 |
+
- Rewards/chosen: -1.0268
|
| 23 |
+
- Rewards/rejected: -1.8204
|
| 24 |
+
- Rewards/accuracies: 0.7617
|
| 25 |
+
- Rewards/margins: 0.7936
|
| 26 |
+
- Logps/rejected: -429.5990
|
| 27 |
+
- Logps/chosen: -349.1275
|
| 28 |
+
- Logits/rejected: 1.1048
|
| 29 |
+
- Logits/chosen: 1.1977
|
| 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: 8
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 8
|
| 54 |
+
- gradient_accumulation_steps: 4
|
| 55 |
+
- total_train_batch_size: 128
|
| 56 |
+
- total_eval_batch_size: 64
|
| 57 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 58 |
+
- lr_scheduler_type: cosine
|
| 59 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 60 |
+
- num_epochs: 1
|
| 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 |
+
| 0.5985 | 0.21 | 100 | 0.6167 | -0.6622 | -0.9981 | 0.7031 | 0.3359 | -347.3664 | -312.6618 | -2.0061 | -1.9992 |
|
| 67 |
+
| 0.5302 | 0.42 | 200 | 0.5495 | -0.8758 | -1.5987 | 0.7461 | 0.7229 | -407.4292 | -334.0204 | 0.3116 | 0.4001 |
|
| 68 |
+
| 0.533 | 0.63 | 300 | 0.5384 | -0.8142 | -1.5157 | 0.7617 | 0.7016 | -399.1313 | -327.8605 | 0.5716 | 0.6809 |
|
| 69 |
+
| 0.518 | 0.84 | 400 | 0.5276 | -1.0554 | -1.8498 | 0.75 | 0.7944 | -432.5438 | -351.9892 | 1.1053 | 1.1955 |
|
| 70 |
+
|
| 71 |
+
|
| 72 |
+
### Framework versions
|
| 73 |
+
|
| 74 |
+
- PEFT 0.7.1
|
| 75 |
+
- Transformers 4.36.2
|
| 76 |
+
- Pytorch 2.1.2+cu121
|
| 77 |
+
- Datasets 2.14.6
|
| 78 |
+
- Tokenizers 0.15.1
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 671150064
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f5b8c0474916d9bca6c774bf485ef22df5555578bb9f71a82560b0ab9c539161
|
| 3 |
size 671150064
|
all_results.json
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"eval_logits/chosen": 1.1976563930511475,
|
| 4 |
+
"eval_logits/rejected": 1.104769229888916,
|
| 5 |
+
"eval_logps/chosen": -349.1275329589844,
|
| 6 |
+
"eval_logps/rejected": -429.5989685058594,
|
| 7 |
+
"eval_loss": 0.5278915762901306,
|
| 8 |
+
"eval_rewards/accuracies": 0.76171875,
|
| 9 |
+
"eval_rewards/chosen": -1.0268259048461914,
|
| 10 |
+
"eval_rewards/margins": 0.7935623526573181,
|
| 11 |
+
"eval_rewards/rejected": -1.8203881978988647,
|
| 12 |
+
"eval_runtime": 72.3078,
|
| 13 |
+
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 27.66,
|
| 15 |
+
"eval_steps_per_second": 0.443,
|
| 16 |
+
"train_loss": 0.5624207920498319,
|
| 17 |
+
"train_runtime": 4926.1363,
|
| 18 |
+
"train_samples": 61135,
|
| 19 |
+
"train_samples_per_second": 12.41,
|
| 20 |
+
"train_steps_per_second": 0.097
|
| 21 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"eval_logits/chosen": 1.1976563930511475,
|
| 4 |
+
"eval_logits/rejected": 1.104769229888916,
|
| 5 |
+
"eval_logps/chosen": -349.1275329589844,
|
| 6 |
+
"eval_logps/rejected": -429.5989685058594,
|
| 7 |
+
"eval_loss": 0.5278915762901306,
|
| 8 |
+
"eval_rewards/accuracies": 0.76171875,
|
| 9 |
+
"eval_rewards/chosen": -1.0268259048461914,
|
| 10 |
+
"eval_rewards/margins": 0.7935623526573181,
|
| 11 |
+
"eval_rewards/rejected": -1.8203881978988647,
|
| 12 |
+
"eval_runtime": 72.3078,
|
| 13 |
+
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 27.66,
|
| 15 |
+
"eval_steps_per_second": 0.443
|
| 16 |
+
}
|
runs/Feb02_15-24-17_ip-26-0-165-38/events.out.tfevents.1706887790.ip-26-0-165-38.1953627.0
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:bc6faf2dee1587b724e87ddf6a767e4d2fd4a4c803d204209137bf6fcdefeeec
|
| 3 |
+
size 38521
|
runs/Feb02_15-24-17_ip-26-0-165-38/events.out.tfevents.1706892789.ip-26-0-165-38.1953627.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6855d5dcdc3a4e7ceca071704d97d4da0fe2b39d1331a42a9d82f1b7420090f9
|
| 3 |
+
size 828
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.5624207920498319,
|
| 4 |
+
"train_runtime": 4926.1363,
|
| 5 |
+
"train_samples": 61135,
|
| 6 |
+
"train_samples_per_second": 12.41,
|
| 7 |
+
"train_steps_per_second": 0.097
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,766 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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": -2.5889174938201904,
|
| 15 |
+
"logits/rejected": -2.4813222885131836,
|
| 16 |
+
"logps/chosen": -289.8450622558594,
|
| 17 |
+
"logps/rejected": -264.9564514160156,
|
| 18 |
+
"loss": 0.6931,
|
| 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": 0.02,
|
| 27 |
+
"learning_rate": 1.0416666666666667e-06,
|
| 28 |
+
"logits/chosen": -2.388092517852783,
|
| 29 |
+
"logits/rejected": -2.4257497787475586,
|
| 30 |
+
"logps/chosen": -260.3330078125,
|
| 31 |
+
"logps/rejected": -219.36460876464844,
|
| 32 |
+
"loss": 0.6925,
|
| 33 |
+
"rewards/accuracies": 0.5277777910232544,
|
| 34 |
+
"rewards/chosen": 0.006709899753332138,
|
| 35 |
+
"rewards/margins": 0.001456400495953858,
|
| 36 |
+
"rewards/rejected": 0.005253499373793602,
|
| 37 |
+
"step": 10
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.04,
|
| 41 |
+
"learning_rate": 2.0833333333333334e-06,
|
| 42 |
+
"logits/chosen": -2.4777274131774902,
|
| 43 |
+
"logits/rejected": -2.4627153873443604,
|
| 44 |
+
"logps/chosen": -269.85845947265625,
|
| 45 |
+
"logps/rejected": -241.0397186279297,
|
| 46 |
+
"loss": 0.6883,
|
| 47 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 48 |
+
"rewards/chosen": 0.036987464874982834,
|
| 49 |
+
"rewards/margins": 0.010376101359724998,
|
| 50 |
+
"rewards/rejected": 0.026611363515257835,
|
| 51 |
+
"step": 20
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.06,
|
| 55 |
+
"learning_rate": 3.125e-06,
|
| 56 |
+
"logits/chosen": -2.4684128761291504,
|
| 57 |
+
"logits/rejected": -2.4569170475006104,
|
| 58 |
+
"logps/chosen": -248.72634887695312,
|
| 59 |
+
"logps/rejected": -245.45999145507812,
|
| 60 |
+
"loss": 0.6814,
|
| 61 |
+
"rewards/accuracies": 0.625,
|
| 62 |
+
"rewards/chosen": 0.04978276044130325,
|
| 63 |
+
"rewards/margins": 0.02046312391757965,
|
| 64 |
+
"rewards/rejected": 0.029319632798433304,
|
| 65 |
+
"step": 30
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.08,
|
| 69 |
+
"learning_rate": 4.166666666666667e-06,
|
| 70 |
+
"logits/chosen": -2.350592613220215,
|
| 71 |
+
"logits/rejected": -2.3819007873535156,
|
| 72 |
+
"logps/chosen": -286.8092041015625,
|
| 73 |
+
"logps/rejected": -258.41522216796875,
|
| 74 |
+
"loss": 0.6638,
|
| 75 |
+
"rewards/accuracies": 0.699999988079071,
|
| 76 |
+
"rewards/chosen": 0.04874344915151596,
|
| 77 |
+
"rewards/margins": 0.06839489936828613,
|
| 78 |
+
"rewards/rejected": -0.019651446491479874,
|
| 79 |
+
"step": 40
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.1,
|
| 83 |
+
"learning_rate": 4.999731868769027e-06,
|
| 84 |
+
"logits/chosen": -2.371358871459961,
|
| 85 |
+
"logits/rejected": -2.3549230098724365,
|
| 86 |
+
"logps/chosen": -286.4510192871094,
|
| 87 |
+
"logps/rejected": -292.4720153808594,
|
| 88 |
+
"loss": 0.6532,
|
| 89 |
+
"rewards/accuracies": 0.675000011920929,
|
| 90 |
+
"rewards/chosen": -0.040428485721349716,
|
| 91 |
+
"rewards/margins": 0.11083565652370453,
|
| 92 |
+
"rewards/rejected": -0.15126413106918335,
|
| 93 |
+
"step": 50
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.13,
|
| 97 |
+
"learning_rate": 4.9903533134293035e-06,
|
| 98 |
+
"logits/chosen": -2.3133959770202637,
|
| 99 |
+
"logits/rejected": -2.2497916221618652,
|
| 100 |
+
"logps/chosen": -266.2447204589844,
|
| 101 |
+
"logps/rejected": -250.51513671875,
|
| 102 |
+
"loss": 0.6381,
|
| 103 |
+
"rewards/accuracies": 0.6875,
|
| 104 |
+
"rewards/chosen": -0.177861288189888,
|
| 105 |
+
"rewards/margins": 0.15977302193641663,
|
| 106 |
+
"rewards/rejected": -0.33763426542282104,
|
| 107 |
+
"step": 60
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.15,
|
| 111 |
+
"learning_rate": 4.967625656594782e-06,
|
| 112 |
+
"logits/chosen": -2.359027624130249,
|
| 113 |
+
"logits/rejected": -2.3221073150634766,
|
| 114 |
+
"logps/chosen": -293.37640380859375,
|
| 115 |
+
"logps/rejected": -272.8565979003906,
|
| 116 |
+
"loss": 0.6222,
|
| 117 |
+
"rewards/accuracies": 0.637499988079071,
|
| 118 |
+
"rewards/chosen": -0.28658825159072876,
|
| 119 |
+
"rewards/margins": 0.1545184701681137,
|
| 120 |
+
"rewards/rejected": -0.4411067068576813,
|
| 121 |
+
"step": 70
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.17,
|
| 125 |
+
"learning_rate": 4.93167072587771e-06,
|
| 126 |
+
"logits/chosen": -2.2765626907348633,
|
| 127 |
+
"logits/rejected": -2.2555174827575684,
|
| 128 |
+
"logps/chosen": -295.1297607421875,
|
| 129 |
+
"logps/rejected": -313.0910339355469,
|
| 130 |
+
"loss": 0.6196,
|
| 131 |
+
"rewards/accuracies": 0.706250011920929,
|
| 132 |
+
"rewards/chosen": -0.2227451503276825,
|
| 133 |
+
"rewards/margins": 0.2126256674528122,
|
| 134 |
+
"rewards/rejected": -0.43537086248397827,
|
| 135 |
+
"step": 80
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.19,
|
| 139 |
+
"learning_rate": 4.882681251368549e-06,
|
| 140 |
+
"logits/chosen": -2.245281219482422,
|
| 141 |
+
"logits/rejected": -2.227741241455078,
|
| 142 |
+
"logps/chosen": -269.5973205566406,
|
| 143 |
+
"logps/rejected": -291.8226318359375,
|
| 144 |
+
"loss": 0.5989,
|
| 145 |
+
"rewards/accuracies": 0.699999988079071,
|
| 146 |
+
"rewards/chosen": -0.3465212285518646,
|
| 147 |
+
"rewards/margins": 0.26312902569770813,
|
| 148 |
+
"rewards/rejected": -0.6096502542495728,
|
| 149 |
+
"step": 90
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.21,
|
| 153 |
+
"learning_rate": 4.8209198325401815e-06,
|
| 154 |
+
"logits/chosen": -2.2300758361816406,
|
| 155 |
+
"logits/rejected": -2.1660382747650146,
|
| 156 |
+
"logps/chosen": -315.1083068847656,
|
| 157 |
+
"logps/rejected": -333.6819763183594,
|
| 158 |
+
"loss": 0.5985,
|
| 159 |
+
"rewards/accuracies": 0.699999988079071,
|
| 160 |
+
"rewards/chosen": -0.5296173691749573,
|
| 161 |
+
"rewards/margins": 0.28275665640830994,
|
| 162 |
+
"rewards/rejected": -0.8123741149902344,
|
| 163 |
+
"step": 100
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 0.21,
|
| 167 |
+
"eval_logits/chosen": -1.999224066734314,
|
| 168 |
+
"eval_logits/rejected": -2.0060811042785645,
|
| 169 |
+
"eval_logps/chosen": -312.6618347167969,
|
| 170 |
+
"eval_logps/rejected": -347.36639404296875,
|
| 171 |
+
"eval_loss": 0.6166529059410095,
|
| 172 |
+
"eval_rewards/accuracies": 0.703125,
|
| 173 |
+
"eval_rewards/chosen": -0.6621690392494202,
|
| 174 |
+
"eval_rewards/margins": 0.3358937203884125,
|
| 175 |
+
"eval_rewards/rejected": -0.9980627298355103,
|
| 176 |
+
"eval_runtime": 73.6215,
|
| 177 |
+
"eval_samples_per_second": 27.166,
|
| 178 |
+
"eval_steps_per_second": 0.435,
|
| 179 |
+
"step": 100
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 0.23,
|
| 183 |
+
"learning_rate": 4.746717530629565e-06,
|
| 184 |
+
"logits/chosen": -2.0477728843688965,
|
| 185 |
+
"logits/rejected": -2.079465389251709,
|
| 186 |
+
"logps/chosen": -331.4808349609375,
|
| 187 |
+
"logps/rejected": -322.7523193359375,
|
| 188 |
+
"loss": 0.603,
|
| 189 |
+
"rewards/accuracies": 0.737500011920929,
|
| 190 |
+
"rewards/chosen": -0.47515854239463806,
|
| 191 |
+
"rewards/margins": 0.4070638120174408,
|
| 192 |
+
"rewards/rejected": -0.8822224736213684,
|
| 193 |
+
"step": 110
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"epoch": 0.25,
|
| 197 |
+
"learning_rate": 4.660472094042121e-06,
|
| 198 |
+
"logits/chosen": -1.9152177572250366,
|
| 199 |
+
"logits/rejected": -1.9233732223510742,
|
| 200 |
+
"logps/chosen": -306.1752014160156,
|
| 201 |
+
"logps/rejected": -293.25958251953125,
|
| 202 |
+
"loss": 0.5842,
|
| 203 |
+
"rewards/accuracies": 0.737500011920929,
|
| 204 |
+
"rewards/chosen": -0.44953522086143494,
|
| 205 |
+
"rewards/margins": 0.3475509285926819,
|
| 206 |
+
"rewards/rejected": -0.7970861196517944,
|
| 207 |
+
"step": 120
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 0.27,
|
| 211 |
+
"learning_rate": 4.5626458262912745e-06,
|
| 212 |
+
"logits/chosen": -1.70361328125,
|
| 213 |
+
"logits/rejected": -1.6728603839874268,
|
| 214 |
+
"logps/chosen": -318.77362060546875,
|
| 215 |
+
"logps/rejected": -358.8970642089844,
|
| 216 |
+
"loss": 0.5741,
|
| 217 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 218 |
+
"rewards/chosen": -0.5537757873535156,
|
| 219 |
+
"rewards/margins": 0.4414924681186676,
|
| 220 |
+
"rewards/rejected": -0.9952683448791504,
|
| 221 |
+
"step": 130
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 0.29,
|
| 225 |
+
"learning_rate": 4.453763107901676e-06,
|
| 226 |
+
"logits/chosen": -1.389970302581787,
|
| 227 |
+
"logits/rejected": -1.3908889293670654,
|
| 228 |
+
"logps/chosen": -365.573974609375,
|
| 229 |
+
"logps/rejected": -389.08026123046875,
|
| 230 |
+
"loss": 0.5673,
|
| 231 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 232 |
+
"rewards/chosen": -0.8941513895988464,
|
| 233 |
+
"rewards/margins": 0.5348166823387146,
|
| 234 |
+
"rewards/rejected": -1.4289681911468506,
|
| 235 |
+
"step": 140
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.31,
|
| 239 |
+
"learning_rate": 4.33440758555951e-06,
|
| 240 |
+
"logits/chosen": -1.2829688787460327,
|
| 241 |
+
"logits/rejected": -1.1666593551635742,
|
| 242 |
+
"logps/chosen": -331.8448181152344,
|
| 243 |
+
"logps/rejected": -347.62164306640625,
|
| 244 |
+
"loss": 0.5587,
|
| 245 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 246 |
+
"rewards/chosen": -0.8844560384750366,
|
| 247 |
+
"rewards/margins": 0.41986608505249023,
|
| 248 |
+
"rewards/rejected": -1.304322361946106,
|
| 249 |
+
"step": 150
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.33,
|
| 253 |
+
"learning_rate": 4.205219043576955e-06,
|
| 254 |
+
"logits/chosen": -0.9949747323989868,
|
| 255 |
+
"logits/rejected": -0.5801770687103271,
|
| 256 |
+
"logps/chosen": -321.4789123535156,
|
| 257 |
+
"logps/rejected": -394.52911376953125,
|
| 258 |
+
"loss": 0.5409,
|
| 259 |
+
"rewards/accuracies": 0.75,
|
| 260 |
+
"rewards/chosen": -0.5671305060386658,
|
| 261 |
+
"rewards/margins": 0.7361670732498169,
|
| 262 |
+
"rewards/rejected": -1.3032976388931274,
|
| 263 |
+
"step": 160
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"epoch": 0.36,
|
| 267 |
+
"learning_rate": 4.066889974440757e-06,
|
| 268 |
+
"logits/chosen": -0.5876813530921936,
|
| 269 |
+
"logits/rejected": -0.3649616539478302,
|
| 270 |
+
"logps/chosen": -293.5026550292969,
|
| 271 |
+
"logps/rejected": -311.59423828125,
|
| 272 |
+
"loss": 0.5916,
|
| 273 |
+
"rewards/accuracies": 0.65625,
|
| 274 |
+
"rewards/chosen": -0.5926370024681091,
|
| 275 |
+
"rewards/margins": 0.5672849416732788,
|
| 276 |
+
"rewards/rejected": -1.1599220037460327,
|
| 277 |
+
"step": 170
|
| 278 |
+
},
|
| 279 |
+
{
|
| 280 |
+
"epoch": 0.38,
|
| 281 |
+
"learning_rate": 3.92016186682789e-06,
|
| 282 |
+
"logits/chosen": -0.8265250325202942,
|
| 283 |
+
"logits/rejected": -0.7326392531394958,
|
| 284 |
+
"logps/chosen": -324.5079650878906,
|
| 285 |
+
"logps/rejected": -388.0706481933594,
|
| 286 |
+
"loss": 0.5533,
|
| 287 |
+
"rewards/accuracies": 0.6875,
|
| 288 |
+
"rewards/chosen": -0.7355499863624573,
|
| 289 |
+
"rewards/margins": 0.6323290467262268,
|
| 290 |
+
"rewards/rejected": -1.3678789138793945,
|
| 291 |
+
"step": 180
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"epoch": 0.4,
|
| 295 |
+
"learning_rate": 3.7658212309857576e-06,
|
| 296 |
+
"logits/chosen": -0.7887569665908813,
|
| 297 |
+
"logits/rejected": -0.5450983643531799,
|
| 298 |
+
"logps/chosen": -341.79412841796875,
|
| 299 |
+
"logps/rejected": -386.400146484375,
|
| 300 |
+
"loss": 0.5368,
|
| 301 |
+
"rewards/accuracies": 0.731249988079071,
|
| 302 |
+
"rewards/chosen": -0.7941572666168213,
|
| 303 |
+
"rewards/margins": 0.5156592130661011,
|
| 304 |
+
"rewards/rejected": -1.309816598892212,
|
| 305 |
+
"step": 190
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 0.42,
|
| 309 |
+
"learning_rate": 3.604695382782159e-06,
|
| 310 |
+
"logits/chosen": -0.2513345181941986,
|
| 311 |
+
"logits/rejected": -0.26383644342422485,
|
| 312 |
+
"logps/chosen": -360.3321228027344,
|
| 313 |
+
"logps/rejected": -372.03326416015625,
|
| 314 |
+
"loss": 0.5302,
|
| 315 |
+
"rewards/accuracies": 0.699999988079071,
|
| 316 |
+
"rewards/chosen": -0.8325196504592896,
|
| 317 |
+
"rewards/margins": 0.6745045185089111,
|
| 318 |
+
"rewards/rejected": -1.5070240497589111,
|
| 319 |
+
"step": 200
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"epoch": 0.42,
|
| 323 |
+
"eval_logits/chosen": 0.4000808000564575,
|
| 324 |
+
"eval_logits/rejected": 0.31158187985420227,
|
| 325 |
+
"eval_logps/chosen": -334.02044677734375,
|
| 326 |
+
"eval_logps/rejected": -407.42919921875,
|
| 327 |
+
"eval_loss": 0.54950350522995,
|
| 328 |
+
"eval_rewards/accuracies": 0.74609375,
|
| 329 |
+
"eval_rewards/chosen": -0.8757554292678833,
|
| 330 |
+
"eval_rewards/margins": 0.7229353189468384,
|
| 331 |
+
"eval_rewards/rejected": -1.5986907482147217,
|
| 332 |
+
"eval_runtime": 72.2579,
|
| 333 |
+
"eval_samples_per_second": 27.679,
|
| 334 |
+
"eval_steps_per_second": 0.443,
|
| 335 |
+
"step": 200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 0.44,
|
| 339 |
+
"learning_rate": 3.437648009023905e-06,
|
| 340 |
+
"logits/chosen": -0.355979859828949,
|
| 341 |
+
"logits/rejected": -0.13378120958805084,
|
| 342 |
+
"logps/chosen": -354.21832275390625,
|
| 343 |
+
"logps/rejected": -387.26171875,
|
| 344 |
+
"loss": 0.5542,
|
| 345 |
+
"rewards/accuracies": 0.637499988079071,
|
| 346 |
+
"rewards/chosen": -0.7991288304328918,
|
| 347 |
+
"rewards/margins": 0.5489404201507568,
|
| 348 |
+
"rewards/rejected": -1.348069190979004,
|
| 349 |
+
"step": 210
|
| 350 |
+
},
|
| 351 |
+
{
|
| 352 |
+
"epoch": 0.46,
|
| 353 |
+
"learning_rate": 3.265574537815398e-06,
|
| 354 |
+
"logits/chosen": -0.2700692415237427,
|
| 355 |
+
"logits/rejected": -0.08166289329528809,
|
| 356 |
+
"logps/chosen": -290.8014831542969,
|
| 357 |
+
"logps/rejected": -332.3136901855469,
|
| 358 |
+
"loss": 0.569,
|
| 359 |
+
"rewards/accuracies": 0.668749988079071,
|
| 360 |
+
"rewards/chosen": -0.608197033405304,
|
| 361 |
+
"rewards/margins": 0.507323145866394,
|
| 362 |
+
"rewards/rejected": -1.1155202388763428,
|
| 363 |
+
"step": 220
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 0.48,
|
| 367 |
+
"learning_rate": 3.089397338773569e-06,
|
| 368 |
+
"logits/chosen": -0.06517831236124039,
|
| 369 |
+
"logits/rejected": 0.004061543848365545,
|
| 370 |
+
"logps/chosen": -317.51495361328125,
|
| 371 |
+
"logps/rejected": -371.9200439453125,
|
| 372 |
+
"loss": 0.5576,
|
| 373 |
+
"rewards/accuracies": 0.71875,
|
| 374 |
+
"rewards/chosen": -0.7464595437049866,
|
| 375 |
+
"rewards/margins": 0.5365883111953735,
|
| 376 |
+
"rewards/rejected": -1.2830479145050049,
|
| 377 |
+
"step": 230
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 0.5,
|
| 381 |
+
"learning_rate": 2.9100607788275547e-06,
|
| 382 |
+
"logits/chosen": -0.13817472755908966,
|
| 383 |
+
"logits/rejected": -0.04675758630037308,
|
| 384 |
+
"logps/chosen": -334.0376281738281,
|
| 385 |
+
"logps/rejected": -356.2500915527344,
|
| 386 |
+
"loss": 0.5592,
|
| 387 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 388 |
+
"rewards/chosen": -0.908129096031189,
|
| 389 |
+
"rewards/margins": 0.4950867295265198,
|
| 390 |
+
"rewards/rejected": -1.403215765953064,
|
| 391 |
+
"step": 240
|
| 392 |
+
},
|
| 393 |
+
{
|
| 394 |
+
"epoch": 0.52,
|
| 395 |
+
"learning_rate": 2.72852616010567e-06,
|
| 396 |
+
"logits/chosen": -0.0858619436621666,
|
| 397 |
+
"logits/rejected": 0.23496215045452118,
|
| 398 |
+
"logps/chosen": -347.7973327636719,
|
| 399 |
+
"logps/rejected": -415.5159606933594,
|
| 400 |
+
"loss": 0.539,
|
| 401 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 402 |
+
"rewards/chosen": -0.9283957481384277,
|
| 403 |
+
"rewards/margins": 0.7460072636604309,
|
| 404 |
+
"rewards/rejected": -1.6744029521942139,
|
| 405 |
+
"step": 250
|
| 406 |
+
},
|
| 407 |
+
{
|
| 408 |
+
"epoch": 0.54,
|
| 409 |
+
"learning_rate": 2.5457665670441937e-06,
|
| 410 |
+
"logits/chosen": 0.35799893736839294,
|
| 411 |
+
"logits/rejected": 0.7192636728286743,
|
| 412 |
+
"logps/chosen": -364.86138916015625,
|
| 413 |
+
"logps/rejected": -400.7781066894531,
|
| 414 |
+
"loss": 0.5156,
|
| 415 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 416 |
+
"rewards/chosen": -1.0014389753341675,
|
| 417 |
+
"rewards/margins": 0.67198646068573,
|
| 418 |
+
"rewards/rejected": -1.673425316810608,
|
| 419 |
+
"step": 260
|
| 420 |
+
},
|
| 421 |
+
{
|
| 422 |
+
"epoch": 0.57,
|
| 423 |
+
"learning_rate": 2.3627616503391813e-06,
|
| 424 |
+
"logits/chosen": 0.7610759735107422,
|
| 425 |
+
"logits/rejected": 0.6359414458274841,
|
| 426 |
+
"logps/chosen": -394.61541748046875,
|
| 427 |
+
"logps/rejected": -451.08563232421875,
|
| 428 |
+
"loss": 0.5183,
|
| 429 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 430 |
+
"rewards/chosen": -1.1308982372283936,
|
| 431 |
+
"rewards/margins": 0.7238563299179077,
|
| 432 |
+
"rewards/rejected": -1.8547546863555908,
|
| 433 |
+
"step": 270
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"epoch": 0.59,
|
| 437 |
+
"learning_rate": 2.1804923757009885e-06,
|
| 438 |
+
"logits/chosen": 1.2029451131820679,
|
| 439 |
+
"logits/rejected": 1.3649919033050537,
|
| 440 |
+
"logps/chosen": -380.61932373046875,
|
| 441 |
+
"logps/rejected": -396.6190490722656,
|
| 442 |
+
"loss": 0.5485,
|
| 443 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 444 |
+
"rewards/chosen": -1.2345253229141235,
|
| 445 |
+
"rewards/margins": 0.6180087327957153,
|
| 446 |
+
"rewards/rejected": -1.8525340557098389,
|
| 447 |
+
"step": 280
|
| 448 |
+
},
|
| 449 |
+
{
|
| 450 |
+
"epoch": 0.61,
|
| 451 |
+
"learning_rate": 1.9999357655598894e-06,
|
| 452 |
+
"logits/chosen": 0.7018269300460815,
|
| 453 |
+
"logits/rejected": 0.5022421479225159,
|
| 454 |
+
"logps/chosen": -329.9044189453125,
|
| 455 |
+
"logps/rejected": -380.2184143066406,
|
| 456 |
+
"loss": 0.5464,
|
| 457 |
+
"rewards/accuracies": 0.71875,
|
| 458 |
+
"rewards/chosen": -0.9144694209098816,
|
| 459 |
+
"rewards/margins": 0.6528812050819397,
|
| 460 |
+
"rewards/rejected": -1.5673506259918213,
|
| 461 |
+
"step": 290
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.63,
|
| 465 |
+
"learning_rate": 1.8220596619089576e-06,
|
| 466 |
+
"logits/chosen": 0.12925171852111816,
|
| 467 |
+
"logits/rejected": 0.4913701117038727,
|
| 468 |
+
"logps/chosen": -334.60748291015625,
|
| 469 |
+
"logps/rejected": -398.9501647949219,
|
| 470 |
+
"loss": 0.533,
|
| 471 |
+
"rewards/accuracies": 0.762499988079071,
|
| 472 |
+
"rewards/chosen": -0.7113243341445923,
|
| 473 |
+
"rewards/margins": 0.6202197670936584,
|
| 474 |
+
"rewards/rejected": -1.3315439224243164,
|
| 475 |
+
"step": 300
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 0.63,
|
| 479 |
+
"eval_logits/chosen": 0.6809147000312805,
|
| 480 |
+
"eval_logits/rejected": 0.5716233849525452,
|
| 481 |
+
"eval_logps/chosen": -327.86053466796875,
|
| 482 |
+
"eval_logps/rejected": -399.1313171386719,
|
| 483 |
+
"eval_loss": 0.5384255051612854,
|
| 484 |
+
"eval_rewards/accuracies": 0.76171875,
|
| 485 |
+
"eval_rewards/chosen": -0.8141559362411499,
|
| 486 |
+
"eval_rewards/margins": 0.7015555500984192,
|
| 487 |
+
"eval_rewards/rejected": -1.5157114267349243,
|
| 488 |
+
"eval_runtime": 72.3068,
|
| 489 |
+
"eval_samples_per_second": 27.66,
|
| 490 |
+
"eval_steps_per_second": 0.443,
|
| 491 |
+
"step": 300
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.65,
|
| 495 |
+
"learning_rate": 1.647817538357072e-06,
|
| 496 |
+
"logits/chosen": 0.2949855327606201,
|
| 497 |
+
"logits/rejected": 0.6500759124755859,
|
| 498 |
+
"logps/chosen": -364.088134765625,
|
| 499 |
+
"logps/rejected": -378.37872314453125,
|
| 500 |
+
"loss": 0.5199,
|
| 501 |
+
"rewards/accuracies": 0.737500011920929,
|
| 502 |
+
"rewards/chosen": -1.0496848821640015,
|
| 503 |
+
"rewards/margins": 0.6011860966682434,
|
| 504 |
+
"rewards/rejected": -1.6508712768554688,
|
| 505 |
+
"step": 310
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 0.67,
|
| 509 |
+
"learning_rate": 1.4781433892011132e-06,
|
| 510 |
+
"logits/chosen": 1.0437233448028564,
|
| 511 |
+
"logits/rejected": 0.9824169278144836,
|
| 512 |
+
"logps/chosen": -368.3329772949219,
|
| 513 |
+
"logps/rejected": -443.03125,
|
| 514 |
+
"loss": 0.5141,
|
| 515 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 516 |
+
"rewards/chosen": -1.1911169290542603,
|
| 517 |
+
"rewards/margins": 0.9159032702445984,
|
| 518 |
+
"rewards/rejected": -2.107020139694214,
|
| 519 |
+
"step": 320
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"epoch": 0.69,
|
| 523 |
+
"learning_rate": 1.3139467229135999e-06,
|
| 524 |
+
"logits/chosen": 0.43222084641456604,
|
| 525 |
+
"logits/rejected": 0.5403100252151489,
|
| 526 |
+
"logps/chosen": -406.03656005859375,
|
| 527 |
+
"logps/rejected": -415.65667724609375,
|
| 528 |
+
"loss": 0.5332,
|
| 529 |
+
"rewards/accuracies": 0.706250011920929,
|
| 530 |
+
"rewards/chosen": -1.136765718460083,
|
| 531 |
+
"rewards/margins": 0.6447176337242126,
|
| 532 |
+
"rewards/rejected": -1.7814832925796509,
|
| 533 |
+
"step": 330
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 0.71,
|
| 537 |
+
"learning_rate": 1.1561076868822756e-06,
|
| 538 |
+
"logits/chosen": 0.783063530921936,
|
| 539 |
+
"logits/rejected": 0.8185106515884399,
|
| 540 |
+
"logps/chosen": -352.68560791015625,
|
| 541 |
+
"logps/rejected": -389.525390625,
|
| 542 |
+
"loss": 0.5327,
|
| 543 |
+
"rewards/accuracies": 0.6875,
|
| 544 |
+
"rewards/chosen": -1.0253931283950806,
|
| 545 |
+
"rewards/margins": 0.48898547887802124,
|
| 546 |
+
"rewards/rejected": -1.514378547668457,
|
| 547 |
+
"step": 340
|
| 548 |
+
},
|
| 549 |
+
{
|
| 550 |
+
"epoch": 0.73,
|
| 551 |
+
"learning_rate": 1.0054723495346484e-06,
|
| 552 |
+
"logits/chosen": 0.8168965578079224,
|
| 553 |
+
"logits/rejected": 0.7234944105148315,
|
| 554 |
+
"logps/chosen": -320.1710510253906,
|
| 555 |
+
"logps/rejected": -381.6952209472656,
|
| 556 |
+
"loss": 0.5137,
|
| 557 |
+
"rewards/accuracies": 0.768750011920929,
|
| 558 |
+
"rewards/chosen": -1.005994200706482,
|
| 559 |
+
"rewards/margins": 0.8249969482421875,
|
| 560 |
+
"rewards/rejected": -1.8309911489486694,
|
| 561 |
+
"step": 350
|
| 562 |
+
},
|
| 563 |
+
{
|
| 564 |
+
"epoch": 0.75,
|
| 565 |
+
"learning_rate": 8.628481651367876e-07,
|
| 566 |
+
"logits/chosen": 0.688895046710968,
|
| 567 |
+
"logits/rejected": 1.0163129568099976,
|
| 568 |
+
"logps/chosen": -380.68157958984375,
|
| 569 |
+
"logps/rejected": -419.84234619140625,
|
| 570 |
+
"loss": 0.5228,
|
| 571 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 572 |
+
"rewards/chosen": -1.1236674785614014,
|
| 573 |
+
"rewards/margins": 0.698872447013855,
|
| 574 |
+
"rewards/rejected": -1.8225399255752563,
|
| 575 |
+
"step": 360
|
| 576 |
+
},
|
| 577 |
+
{
|
| 578 |
+
"epoch": 0.77,
|
| 579 |
+
"learning_rate": 7.289996455765749e-07,
|
| 580 |
+
"logits/chosen": 0.6197929382324219,
|
| 581 |
+
"logits/rejected": 0.7506722807884216,
|
| 582 |
+
"logps/chosen": -420.45263671875,
|
| 583 |
+
"logps/rejected": -426.29107666015625,
|
| 584 |
+
"loss": 0.5353,
|
| 585 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 586 |
+
"rewards/chosen": -1.1160409450531006,
|
| 587 |
+
"rewards/margins": 0.6991242170333862,
|
| 588 |
+
"rewards/rejected": -1.8151648044586182,
|
| 589 |
+
"step": 370
|
| 590 |
+
},
|
| 591 |
+
{
|
| 592 |
+
"epoch": 0.8,
|
| 593 |
+
"learning_rate": 6.046442623320145e-07,
|
| 594 |
+
"logits/chosen": 0.5390521287918091,
|
| 595 |
+
"logits/rejected": 0.5552736520767212,
|
| 596 |
+
"logps/chosen": -365.60577392578125,
|
| 597 |
+
"logps/rejected": -421.53564453125,
|
| 598 |
+
"loss": 0.5076,
|
| 599 |
+
"rewards/accuracies": 0.737500011920929,
|
| 600 |
+
"rewards/chosen": -1.0088932514190674,
|
| 601 |
+
"rewards/margins": 0.6299774050712585,
|
| 602 |
+
"rewards/rejected": -1.6388708353042603,
|
| 603 |
+
"step": 380
|
| 604 |
+
},
|
| 605 |
+
{
|
| 606 |
+
"epoch": 0.82,
|
| 607 |
+
"learning_rate": 4.904486005914027e-07,
|
| 608 |
+
"logits/chosen": 0.7070841789245605,
|
| 609 |
+
"logits/rejected": 0.8798855543136597,
|
| 610 |
+
"logps/chosen": -369.456787109375,
|
| 611 |
+
"logps/rejected": -406.3539123535156,
|
| 612 |
+
"loss": 0.5334,
|
| 613 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 614 |
+
"rewards/chosen": -1.0536937713623047,
|
| 615 |
+
"rewards/margins": 0.5137845277786255,
|
| 616 |
+
"rewards/rejected": -1.5674784183502197,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 0.84,
|
| 621 |
+
"learning_rate": 3.8702478614051353e-07,
|
| 622 |
+
"logits/chosen": 0.5031794905662537,
|
| 623 |
+
"logits/rejected": 0.47197189927101135,
|
| 624 |
+
"logps/chosen": -409.0977478027344,
|
| 625 |
+
"logps/rejected": -434.04693603515625,
|
| 626 |
+
"loss": 0.518,
|
| 627 |
+
"rewards/accuracies": 0.668749988079071,
|
| 628 |
+
"rewards/chosen": -1.1779779195785522,
|
| 629 |
+
"rewards/margins": 0.5962620973587036,
|
| 630 |
+
"rewards/rejected": -1.7742401361465454,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 0.84,
|
| 635 |
+
"eval_logits/chosen": 1.195478081703186,
|
| 636 |
+
"eval_logits/rejected": 1.1053041219711304,
|
| 637 |
+
"eval_logps/chosen": -351.9891662597656,
|
| 638 |
+
"eval_logps/rejected": -432.54376220703125,
|
| 639 |
+
"eval_loss": 0.5275784730911255,
|
| 640 |
+
"eval_rewards/accuracies": 0.75,
|
| 641 |
+
"eval_rewards/chosen": -1.0554425716400146,
|
| 642 |
+
"eval_rewards/margins": 0.7943933010101318,
|
| 643 |
+
"eval_rewards/rejected": -1.849835753440857,
|
| 644 |
+
"eval_runtime": 72.3032,
|
| 645 |
+
"eval_samples_per_second": 27.661,
|
| 646 |
+
"eval_steps_per_second": 0.443,
|
| 647 |
+
"step": 400
|
| 648 |
+
},
|
| 649 |
+
{
|
| 650 |
+
"epoch": 0.86,
|
| 651 |
+
"learning_rate": 2.9492720416985004e-07,
|
| 652 |
+
"logits/chosen": 0.8437727093696594,
|
| 653 |
+
"logits/rejected": 0.800951361656189,
|
| 654 |
+
"logps/chosen": -389.89471435546875,
|
| 655 |
+
"logps/rejected": -402.897705078125,
|
| 656 |
+
"loss": 0.536,
|
| 657 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 658 |
+
"rewards/chosen": -1.1922531127929688,
|
| 659 |
+
"rewards/margins": 0.5932148694992065,
|
| 660 |
+
"rewards/rejected": -1.7854681015014648,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.88,
|
| 665 |
+
"learning_rate": 2.1464952759020857e-07,
|
| 666 |
+
"logits/chosen": 0.8531819581985474,
|
| 667 |
+
"logits/rejected": 0.7011052966117859,
|
| 668 |
+
"logps/chosen": -365.4083557128906,
|
| 669 |
+
"logps/rejected": -429.77392578125,
|
| 670 |
+
"loss": 0.5293,
|
| 671 |
+
"rewards/accuracies": 0.706250011920929,
|
| 672 |
+
"rewards/chosen": -1.1621034145355225,
|
| 673 |
+
"rewards/margins": 0.5474345684051514,
|
| 674 |
+
"rewards/rejected": -1.7095378637313843,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 0.9,
|
| 679 |
+
"learning_rate": 1.4662207078575685e-07,
|
| 680 |
+
"logits/chosen": 0.9001744389533997,
|
| 681 |
+
"logits/rejected": 0.7595891356468201,
|
| 682 |
+
"logps/chosen": -376.25653076171875,
|
| 683 |
+
"logps/rejected": -414.7420959472656,
|
| 684 |
+
"loss": 0.5296,
|
| 685 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 686 |
+
"rewards/chosen": -1.0551135540008545,
|
| 687 |
+
"rewards/margins": 0.7705513834953308,
|
| 688 |
+
"rewards/rejected": -1.825664758682251,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 0.92,
|
| 693 |
+
"learning_rate": 9.120948298936422e-08,
|
| 694 |
+
"logits/chosen": 0.6783544421195984,
|
| 695 |
+
"logits/rejected": 0.9758931994438171,
|
| 696 |
+
"logps/chosen": -374.05902099609375,
|
| 697 |
+
"logps/rejected": -429.77825927734375,
|
| 698 |
+
"loss": 0.5249,
|
| 699 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 700 |
+
"rewards/chosen": -1.049117088317871,
|
| 701 |
+
"rewards/margins": 0.7034090161323547,
|
| 702 |
+
"rewards/rejected": -1.752526044845581,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.94,
|
| 707 |
+
"learning_rate": 4.870879364444109e-08,
|
| 708 |
+
"logits/chosen": 0.8478315472602844,
|
| 709 |
+
"logits/rejected": 0.7550174593925476,
|
| 710 |
+
"logps/chosen": -369.22027587890625,
|
| 711 |
+
"logps/rejected": -430.88873291015625,
|
| 712 |
+
"loss": 0.5406,
|
| 713 |
+
"rewards/accuracies": 0.71875,
|
| 714 |
+
"rewards/chosen": -1.088853120803833,
|
| 715 |
+
"rewards/margins": 0.6529593467712402,
|
| 716 |
+
"rewards/rejected": -1.7418124675750732,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 0.96,
|
| 721 |
+
"learning_rate": 1.93478202307823e-08,
|
| 722 |
+
"logits/chosen": 0.6956031918525696,
|
| 723 |
+
"logits/rejected": 0.911687970161438,
|
| 724 |
+
"logps/chosen": -368.9820251464844,
|
| 725 |
+
"logps/rejected": -417.018310546875,
|
| 726 |
+
"loss": 0.516,
|
| 727 |
+
"rewards/accuracies": 0.78125,
|
| 728 |
+
"rewards/chosen": -1.0993984937667847,
|
| 729 |
+
"rewards/margins": 0.6686335206031799,
|
| 730 |
+
"rewards/rejected": -1.7680320739746094,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 0.98,
|
| 735 |
+
"learning_rate": 3.283947088983663e-09,
|
| 736 |
+
"logits/chosen": 0.7104119658470154,
|
| 737 |
+
"logits/rejected": 0.7404820322990417,
|
| 738 |
+
"logps/chosen": -359.8403625488281,
|
| 739 |
+
"logps/rejected": -435.29803466796875,
|
| 740 |
+
"loss": 0.509,
|
| 741 |
+
"rewards/accuracies": 0.8125,
|
| 742 |
+
"rewards/chosen": -1.0412371158599854,
|
| 743 |
+
"rewards/margins": 0.8249354362487793,
|
| 744 |
+
"rewards/rejected": -1.8661725521087646,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 1.0,
|
| 749 |
+
"step": 477,
|
| 750 |
+
"total_flos": 0.0,
|
| 751 |
+
"train_loss": 0.5624207920498319,
|
| 752 |
+
"train_runtime": 4926.1363,
|
| 753 |
+
"train_samples_per_second": 12.41,
|
| 754 |
+
"train_steps_per_second": 0.097
|
| 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,
|
| 762 |
+
"total_flos": 0.0,
|
| 763 |
+
"train_batch_size": 4,
|
| 764 |
+
"trial_name": null,
|
| 765 |
+
"trial_params": null
|
| 766 |
+
}
|