Model save
Browse files- README.md +77 -0
- all_results.json +6 -6
- model-00001-of-00003.safetensors +1 -1
- model-00002-of-00003.safetensors +1 -1
- model-00003-of-00003.safetensors +1 -1
- train_results.json +6 -6
- trainer_state.json +778 -9
- training_args.bin +1 -1
README.md
ADDED
|
@@ -0,0 +1,77 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- dpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-7b-dpo-full
|
| 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-full
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on an unknown dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Logits/chosen: -0.9769
|
| 22 |
+
- Logits/rejected: -0.3233
|
| 23 |
+
- Logps/chosen: -378.2564
|
| 24 |
+
- Logps/rejected: -432.0113
|
| 25 |
+
- Loss: 0.5111
|
| 26 |
+
- Rewards/accuracies: 0.7758
|
| 27 |
+
- Rewards/chosen: -0.9629
|
| 28 |
+
- Rewards/margins: 0.7553
|
| 29 |
+
- Rewards/rejected: -1.7181
|
| 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-07
|
| 49 |
+
- train_batch_size: 8
|
| 50 |
+
- eval_batch_size: 8
|
| 51 |
+
- seed: 42
|
| 52 |
+
- distributed_type: multi-GPU
|
| 53 |
+
- num_devices: 4
|
| 54 |
+
- gradient_accumulation_steps: 4
|
| 55 |
+
- total_train_batch_size: 128
|
| 56 |
+
- total_eval_batch_size: 32
|
| 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 | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
|
| 65 |
+
|:-------------:|:------:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
|
| 66 |
+
| 0.6242 | 0.1047 | 100 | -2.4575 | -2.3901 | -306.4380 | -306.3153 | 0.6225 | 0.7044 | -0.2447 | 0.2165 | -0.4612 |
|
| 67 |
+
| 0.5567 | 0.2093 | 200 | -0.9274 | -0.6452 | -394.8172 | -443.9744 | 0.5513 | 0.7381 | -1.1285 | 0.7093 | -1.8378 |
|
| 68 |
+
| 0.5334 | 0.3140 | 300 | -0.1225 | 0.3983 | -366.7193 | -428.3155 | 0.5222 | 0.7639 | -0.8475 | 0.8337 | -1.6812 |
|
| 69 |
+
| 0.5426 | 0.4186 | 400 | -0.9769 | -0.3233 | -378.2564 | -432.0113 | 0.5111 | 0.7758 | -0.9629 | 0.7553 | -1.7181 |
|
| 70 |
+
|
| 71 |
+
|
| 72 |
+
### Framework versions
|
| 73 |
+
|
| 74 |
+
- Transformers 4.44.1
|
| 75 |
+
- Pytorch 2.1.2+cu121
|
| 76 |
+
- Datasets 2.21.0
|
| 77 |
+
- Tokenizers 0.19.1
|
all_results.json
CHANGED
|
@@ -1,9 +1,9 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
"total_flos": 0.0,
|
| 4 |
-
"train_loss": 0.
|
| 5 |
-
"train_runtime":
|
| 6 |
-
"train_samples":
|
| 7 |
-
"train_samples_per_second":
|
| 8 |
-
"train_steps_per_second":
|
| 9 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 0.9984301412872841,
|
| 3 |
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.08199021051514824,
|
| 5 |
+
"train_runtime": 2352.184,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 25.99,
|
| 8 |
+
"train_steps_per_second": 0.203
|
| 9 |
}
|
model-00001-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4943162336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b2826b6f9a491e0746f370f2551b16b684e285e8b4801583f8105755b72323dc
|
| 3 |
size 4943162336
|
model-00002-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4999819336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:df3c87a6c70c435b32e6f2a7eaf284d5fcab60acf7f98e91791e817507b0b5c0
|
| 3 |
size 4999819336
|
model-00003-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4540516344
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0e68ccec670108682d4b397715ba1074bb6d19dede2387e14a15baf4fb58a799
|
| 3 |
size 4540516344
|
train_results.json
CHANGED
|
@@ -1,9 +1,9 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
"total_flos": 0.0,
|
| 4 |
-
"train_loss": 0.
|
| 5 |
-
"train_runtime":
|
| 6 |
-
"train_samples":
|
| 7 |
-
"train_samples_per_second":
|
| 8 |
-
"train_steps_per_second":
|
| 9 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 0.9984301412872841,
|
| 3 |
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.08199021051514824,
|
| 5 |
+
"train_runtime": 2352.184,
|
| 6 |
+
"train_samples": 61134,
|
| 7 |
+
"train_samples_per_second": 25.99,
|
| 8 |
+
"train_steps_per_second": 0.203
|
| 9 |
}
|
trainer_state.json
CHANGED
|
@@ -1,9 +1,9 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch":
|
| 5 |
"eval_steps": 100,
|
| 6 |
-
"global_step":
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
|
@@ -24,17 +24,786 @@
|
|
| 24 |
"step": 1
|
| 25 |
},
|
| 26 |
{
|
| 27 |
-
"epoch":
|
| 28 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 29 |
"total_flos": 0.0,
|
| 30 |
-
"train_loss": 0.
|
| 31 |
-
"train_runtime":
|
| 32 |
-
"train_samples_per_second":
|
| 33 |
-
"train_steps_per_second":
|
| 34 |
}
|
| 35 |
],
|
| 36 |
"logging_steps": 10,
|
| 37 |
-
"max_steps":
|
| 38 |
"num_input_tokens_seen": 0,
|
| 39 |
"num_train_epochs": 1,
|
| 40 |
"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,
|
|
|
|
| 24 |
"step": 1
|
| 25 |
},
|
| 26 |
{
|
| 27 |
+
"epoch": 0.010465724751439037,
|
| 28 |
+
"grad_norm": 11.253420838176064,
|
| 29 |
+
"learning_rate": 5.208333333333333e-08,
|
| 30 |
+
"logits/chosen": -2.5951685905456543,
|
| 31 |
+
"logits/rejected": -2.568584442138672,
|
| 32 |
+
"logps/chosen": -256.67633056640625,
|
| 33 |
+
"logps/rejected": -235.04104614257812,
|
| 34 |
+
"loss": 0.6929,
|
| 35 |
+
"rewards/accuracies": 0.5347222089767456,
|
| 36 |
+
"rewards/chosen": -0.0006383458967320621,
|
| 37 |
+
"rewards/margins": 0.0009785331785678864,
|
| 38 |
+
"rewards/rejected": -0.0016168789006769657,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.020931449502878074,
|
| 43 |
+
"grad_norm": 13.012899469929792,
|
| 44 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 45 |
+
"logits/chosen": -2.6123788356781006,
|
| 46 |
+
"logits/rejected": -2.574439764022827,
|
| 47 |
+
"logps/chosen": -283.15533447265625,
|
| 48 |
+
"logps/rejected": -282.46636962890625,
|
| 49 |
+
"loss": 0.6923,
|
| 50 |
+
"rewards/accuracies": 0.581250011920929,
|
| 51 |
+
"rewards/chosen": -0.0005093275103718042,
|
| 52 |
+
"rewards/margins": 0.000968976120930165,
|
| 53 |
+
"rewards/rejected": -0.0014783035730943084,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.03139717425431711,
|
| 58 |
+
"grad_norm": 12.332229893392334,
|
| 59 |
+
"learning_rate": 1.5624999999999999e-07,
|
| 60 |
+
"logits/chosen": -2.6900181770324707,
|
| 61 |
+
"logits/rejected": -2.665644884109497,
|
| 62 |
+
"logps/chosen": -270.2990417480469,
|
| 63 |
+
"logps/rejected": -277.00897216796875,
|
| 64 |
+
"loss": 0.6912,
|
| 65 |
+
"rewards/accuracies": 0.643750011920929,
|
| 66 |
+
"rewards/chosen": -0.0003606746322475374,
|
| 67 |
+
"rewards/margins": 0.0033442180138081312,
|
| 68 |
+
"rewards/rejected": -0.003704892937093973,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.04186289900575615,
|
| 73 |
+
"grad_norm": 11.842443150545527,
|
| 74 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 75 |
+
"logits/chosen": -2.6624841690063477,
|
| 76 |
+
"logits/rejected": -2.586958408355713,
|
| 77 |
+
"logps/chosen": -290.5253601074219,
|
| 78 |
+
"logps/rejected": -282.37347412109375,
|
| 79 |
+
"loss": 0.6894,
|
| 80 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 81 |
+
"rewards/chosen": 0.0042248861864209175,
|
| 82 |
+
"rewards/margins": 0.007044057361781597,
|
| 83 |
+
"rewards/rejected": -0.0028191714081913233,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.052328623757195186,
|
| 88 |
+
"grad_norm": 13.07201246671239,
|
| 89 |
+
"learning_rate": 2.604166666666667e-07,
|
| 90 |
+
"logits/chosen": -2.666559934616089,
|
| 91 |
+
"logits/rejected": -2.5819408893585205,
|
| 92 |
+
"logps/chosen": -266.31890869140625,
|
| 93 |
+
"logps/rejected": -236.9652862548828,
|
| 94 |
+
"loss": 0.6849,
|
| 95 |
+
"rewards/accuracies": 0.675000011920929,
|
| 96 |
+
"rewards/chosen": 0.01324030477553606,
|
| 97 |
+
"rewards/margins": 0.017407527193427086,
|
| 98 |
+
"rewards/rejected": -0.0041672238148748875,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.06279434850863422,
|
| 103 |
+
"grad_norm": 11.802730661940267,
|
| 104 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 105 |
+
"logits/chosen": -2.620361328125,
|
| 106 |
+
"logits/rejected": -2.5846078395843506,
|
| 107 |
+
"logps/chosen": -299.8506774902344,
|
| 108 |
+
"logps/rejected": -274.9012145996094,
|
| 109 |
+
"loss": 0.6779,
|
| 110 |
+
"rewards/accuracies": 0.675000011920929,
|
| 111 |
+
"rewards/chosen": 0.041565705090761185,
|
| 112 |
+
"rewards/margins": 0.0326404795050621,
|
| 113 |
+
"rewards/rejected": 0.008925220929086208,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.07326007326007326,
|
| 118 |
+
"grad_norm": 12.308212741153868,
|
| 119 |
+
"learning_rate": 3.645833333333333e-07,
|
| 120 |
+
"logits/chosen": -2.5244507789611816,
|
| 121 |
+
"logits/rejected": -2.5163016319274902,
|
| 122 |
+
"logps/chosen": -258.10455322265625,
|
| 123 |
+
"logps/rejected": -263.7887268066406,
|
| 124 |
+
"loss": 0.6686,
|
| 125 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 126 |
+
"rewards/chosen": 0.01530471257865429,
|
| 127 |
+
"rewards/margins": 0.07072288542985916,
|
| 128 |
+
"rewards/rejected": -0.05541817098855972,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.0837257980115123,
|
| 133 |
+
"grad_norm": 14.305090615187044,
|
| 134 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 135 |
+
"logits/chosen": -2.545693874359131,
|
| 136 |
+
"logits/rejected": -2.472573757171631,
|
| 137 |
+
"logps/chosen": -274.86029052734375,
|
| 138 |
+
"logps/rejected": -261.91314697265625,
|
| 139 |
+
"loss": 0.6485,
|
| 140 |
+
"rewards/accuracies": 0.731249988079071,
|
| 141 |
+
"rewards/chosen": 0.012162665836513042,
|
| 142 |
+
"rewards/margins": 0.11833341419696808,
|
| 143 |
+
"rewards/rejected": -0.10617075115442276,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.09419152276295134,
|
| 148 |
+
"grad_norm": 15.35099235694953,
|
| 149 |
+
"learning_rate": 4.6874999999999996e-07,
|
| 150 |
+
"logits/chosen": -2.4933650493621826,
|
| 151 |
+
"logits/rejected": -2.4577670097351074,
|
| 152 |
+
"logps/chosen": -262.3482971191406,
|
| 153 |
+
"logps/rejected": -279.24951171875,
|
| 154 |
+
"loss": 0.6415,
|
| 155 |
+
"rewards/accuracies": 0.637499988079071,
|
| 156 |
+
"rewards/chosen": -0.06777340173721313,
|
| 157 |
+
"rewards/margins": 0.07752426713705063,
|
| 158 |
+
"rewards/rejected": -0.14529766142368317,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.10465724751439037,
|
| 163 |
+
"grad_norm": 14.293066089274086,
|
| 164 |
+
"learning_rate": 4.999732492681437e-07,
|
| 165 |
+
"logits/chosen": -2.4715874195098877,
|
| 166 |
+
"logits/rejected": -2.3922314643859863,
|
| 167 |
+
"logps/chosen": -309.3290710449219,
|
| 168 |
+
"logps/rejected": -309.43463134765625,
|
| 169 |
+
"loss": 0.6242,
|
| 170 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 171 |
+
"rewards/chosen": -0.094756118953228,
|
| 172 |
+
"rewards/margins": 0.19953417778015137,
|
| 173 |
+
"rewards/rejected": -0.29429030418395996,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.10465724751439037,
|
| 178 |
+
"eval_logits/chosen": -2.457521677017212,
|
| 179 |
+
"eval_logits/rejected": -2.390087366104126,
|
| 180 |
+
"eval_logps/chosen": -306.4379577636719,
|
| 181 |
+
"eval_logps/rejected": -306.3153381347656,
|
| 182 |
+
"eval_loss": 0.6225137710571289,
|
| 183 |
+
"eval_rewards/accuracies": 0.704365074634552,
|
| 184 |
+
"eval_rewards/chosen": -0.244705930352211,
|
| 185 |
+
"eval_rewards/margins": 0.2164848893880844,
|
| 186 |
+
"eval_rewards/rejected": -0.4611908495426178,
|
| 187 |
+
"eval_runtime": 195.8672,
|
| 188 |
+
"eval_samples_per_second": 10.211,
|
| 189 |
+
"eval_steps_per_second": 0.322,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.1151229722658294,
|
| 194 |
+
"grad_norm": 18.62360509072815,
|
| 195 |
+
"learning_rate": 4.996723692767926e-07,
|
| 196 |
+
"logits/chosen": -2.4408485889434814,
|
| 197 |
+
"logits/rejected": -2.3859519958496094,
|
| 198 |
+
"logps/chosen": -269.1166687011719,
|
| 199 |
+
"logps/rejected": -283.4706726074219,
|
| 200 |
+
"loss": 0.6337,
|
| 201 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 202 |
+
"rewards/chosen": -0.2764732837677002,
|
| 203 |
+
"rewards/margins": 0.1853046417236328,
|
| 204 |
+
"rewards/rejected": -0.461777925491333,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.12558869701726844,
|
| 209 |
+
"grad_norm": 17.7031088497414,
|
| 210 |
+
"learning_rate": 4.990375746213598e-07,
|
| 211 |
+
"logits/chosen": -2.3903005123138428,
|
| 212 |
+
"logits/rejected": -2.3269896507263184,
|
| 213 |
+
"logps/chosen": -270.2724914550781,
|
| 214 |
+
"logps/rejected": -319.43701171875,
|
| 215 |
+
"loss": 0.6054,
|
| 216 |
+
"rewards/accuracies": 0.731249988079071,
|
| 217 |
+
"rewards/chosen": -0.07692138850688934,
|
| 218 |
+
"rewards/margins": 0.23280763626098633,
|
| 219 |
+
"rewards/rejected": -0.3097290098667145,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.1360544217687075,
|
| 224 |
+
"grad_norm": 22.37561070304176,
|
| 225 |
+
"learning_rate": 4.980697142834314e-07,
|
| 226 |
+
"logits/chosen": -2.3643040657043457,
|
| 227 |
+
"logits/rejected": -2.296745538711548,
|
| 228 |
+
"logps/chosen": -302.34283447265625,
|
| 229 |
+
"logps/rejected": -309.90008544921875,
|
| 230 |
+
"loss": 0.5952,
|
| 231 |
+
"rewards/accuracies": 0.706250011920929,
|
| 232 |
+
"rewards/chosen": -0.21061678230762482,
|
| 233 |
+
"rewards/margins": 0.3443460464477539,
|
| 234 |
+
"rewards/rejected": -0.5549627542495728,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.14652014652014653,
|
| 239 |
+
"grad_norm": 15.841010404827179,
|
| 240 |
+
"learning_rate": 4.967700826904229e-07,
|
| 241 |
+
"logits/chosen": -2.232334852218628,
|
| 242 |
+
"logits/rejected": -2.1791012287139893,
|
| 243 |
+
"logps/chosen": -310.8321838378906,
|
| 244 |
+
"logps/rejected": -345.97906494140625,
|
| 245 |
+
"loss": 0.5764,
|
| 246 |
+
"rewards/accuracies": 0.75,
|
| 247 |
+
"rewards/chosen": -0.19714145362377167,
|
| 248 |
+
"rewards/margins": 0.37270230054855347,
|
| 249 |
+
"rewards/rejected": -0.5698437690734863,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.15698587127158556,
|
| 254 |
+
"grad_norm": 29.527345852120696,
|
| 255 |
+
"learning_rate": 4.951404179843962e-07,
|
| 256 |
+
"logits/chosen": -2.416675090789795,
|
| 257 |
+
"logits/rejected": -2.345705270767212,
|
| 258 |
+
"logps/chosen": -322.52532958984375,
|
| 259 |
+
"logps/rejected": -315.2313537597656,
|
| 260 |
+
"loss": 0.5922,
|
| 261 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 262 |
+
"rewards/chosen": -0.3468259871006012,
|
| 263 |
+
"rewards/margins": 0.3354201018810272,
|
| 264 |
+
"rewards/rejected": -0.6822460889816284,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.1674515960230246,
|
| 269 |
+
"grad_norm": 18.994646147175615,
|
| 270 |
+
"learning_rate": 4.931828996974498e-07,
|
| 271 |
+
"logits/chosen": -2.336097478866577,
|
| 272 |
+
"logits/rejected": -2.1988613605499268,
|
| 273 |
+
"logps/chosen": -325.79058837890625,
|
| 274 |
+
"logps/rejected": -324.61297607421875,
|
| 275 |
+
"loss": 0.5769,
|
| 276 |
+
"rewards/accuracies": 0.706250011920929,
|
| 277 |
+
"rewards/chosen": -0.2996981739997864,
|
| 278 |
+
"rewards/margins": 0.3862423896789551,
|
| 279 |
+
"rewards/rejected": -0.6859406232833862,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.17791732077446362,
|
| 284 |
+
"grad_norm": 26.687259386090716,
|
| 285 |
+
"learning_rate": 4.909001458367866e-07,
|
| 286 |
+
"logits/chosen": -2.323286533355713,
|
| 287 |
+
"logits/rejected": -2.189216136932373,
|
| 288 |
+
"logps/chosen": -319.3229064941406,
|
| 289 |
+
"logps/rejected": -323.94879150390625,
|
| 290 |
+
"loss": 0.5739,
|
| 291 |
+
"rewards/accuracies": 0.6875,
|
| 292 |
+
"rewards/chosen": -0.40793466567993164,
|
| 293 |
+
"rewards/margins": 0.4229934811592102,
|
| 294 |
+
"rewards/rejected": -0.8309281468391418,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.18838304552590268,
|
| 299 |
+
"grad_norm": 29.067072686641563,
|
| 300 |
+
"learning_rate": 4.882952093833627e-07,
|
| 301 |
+
"logits/chosen": -1.956639289855957,
|
| 302 |
+
"logits/rejected": -1.8405163288116455,
|
| 303 |
+
"logps/chosen": -318.9150695800781,
|
| 304 |
+
"logps/rejected": -385.9713439941406,
|
| 305 |
+
"loss": 0.5384,
|
| 306 |
+
"rewards/accuracies": 0.737500011920929,
|
| 307 |
+
"rewards/chosen": -0.7896965146064758,
|
| 308 |
+
"rewards/margins": 0.6018267869949341,
|
| 309 |
+
"rewards/rejected": -1.3915233612060547,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.1988487702773417,
|
| 314 |
+
"grad_norm": 22.179665325246994,
|
| 315 |
+
"learning_rate": 4.853715742087946e-07,
|
| 316 |
+
"logits/chosen": -1.426253318786621,
|
| 317 |
+
"logits/rejected": -1.3614884614944458,
|
| 318 |
+
"logps/chosen": -310.2138671875,
|
| 319 |
+
"logps/rejected": -383.07489013671875,
|
| 320 |
+
"loss": 0.5462,
|
| 321 |
+
"rewards/accuracies": 0.706250011920929,
|
| 322 |
+
"rewards/chosen": -0.7427991032600403,
|
| 323 |
+
"rewards/margins": 0.5888546705245972,
|
| 324 |
+
"rewards/rejected": -1.3316537141799927,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.20931449502878074,
|
| 329 |
+
"grad_norm": 32.92359913670319,
|
| 330 |
+
"learning_rate": 4.821331504159906e-07,
|
| 331 |
+
"logits/chosen": -1.2346317768096924,
|
| 332 |
+
"logits/rejected": -0.8972011804580688,
|
| 333 |
+
"logps/chosen": -392.1202697753906,
|
| 334 |
+
"logps/rejected": -404.49041748046875,
|
| 335 |
+
"loss": 0.5567,
|
| 336 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 337 |
+
"rewards/chosen": -0.8150125741958618,
|
| 338 |
+
"rewards/margins": 0.6587424874305725,
|
| 339 |
+
"rewards/rejected": -1.473755121231079,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.20931449502878074,
|
| 344 |
+
"eval_logits/chosen": -0.9274371862411499,
|
| 345 |
+
"eval_logits/rejected": -0.6452000737190247,
|
| 346 |
+
"eval_logps/chosen": -394.81719970703125,
|
| 347 |
+
"eval_logps/rejected": -443.9743957519531,
|
| 348 |
+
"eval_loss": 0.551316499710083,
|
| 349 |
+
"eval_rewards/accuracies": 0.738095223903656,
|
| 350 |
+
"eval_rewards/chosen": -1.1284979581832886,
|
| 351 |
+
"eval_rewards/margins": 0.709282398223877,
|
| 352 |
+
"eval_rewards/rejected": -1.8377805948257446,
|
| 353 |
+
"eval_runtime": 189.7213,
|
| 354 |
+
"eval_samples_per_second": 10.542,
|
| 355 |
+
"eval_steps_per_second": 0.332,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.21978021978021978,
|
| 360 |
+
"grad_norm": 32.04286337283456,
|
| 361 |
+
"learning_rate": 4.785842691097342e-07,
|
| 362 |
+
"logits/chosen": -0.9299179911613464,
|
| 363 |
+
"logits/rejected": -0.6753123998641968,
|
| 364 |
+
"logps/chosen": -400.389892578125,
|
| 365 |
+
"logps/rejected": -444.14349365234375,
|
| 366 |
+
"loss": 0.5785,
|
| 367 |
+
"rewards/accuracies": 0.668749988079071,
|
| 368 |
+
"rewards/chosen": -1.1756278276443481,
|
| 369 |
+
"rewards/margins": 0.5428754687309265,
|
| 370 |
+
"rewards/rejected": -1.7185035943984985,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.2302459445316588,
|
| 375 |
+
"grad_norm": 28.16699768317943,
|
| 376 |
+
"learning_rate": 4.7472967660421603e-07,
|
| 377 |
+
"logits/chosen": -1.2778313159942627,
|
| 378 |
+
"logits/rejected": -0.8387069702148438,
|
| 379 |
+
"logps/chosen": -362.0044250488281,
|
| 380 |
+
"logps/rejected": -370.3143310546875,
|
| 381 |
+
"loss": 0.5346,
|
| 382 |
+
"rewards/accuracies": 0.706250011920929,
|
| 383 |
+
"rewards/chosen": -0.7484822869300842,
|
| 384 |
+
"rewards/margins": 0.5931913256645203,
|
| 385 |
+
"rewards/rejected": -1.341673731803894,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.24071166928309787,
|
| 390 |
+
"grad_norm": 34.9569009505112,
|
| 391 |
+
"learning_rate": 4.705745280752585e-07,
|
| 392 |
+
"logits/chosen": -0.8671897649765015,
|
| 393 |
+
"logits/rejected": -0.5648606419563293,
|
| 394 |
+
"logps/chosen": -372.58831787109375,
|
| 395 |
+
"logps/rejected": -411.037841796875,
|
| 396 |
+
"loss": 0.5499,
|
| 397 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 398 |
+
"rewards/chosen": -0.8385046124458313,
|
| 399 |
+
"rewards/margins": 0.5863326787948608,
|
| 400 |
+
"rewards/rejected": -1.4248372316360474,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.25117739403453687,
|
| 405 |
+
"grad_norm": 31.66255501363827,
|
| 406 |
+
"learning_rate": 4.6612438066572555e-07,
|
| 407 |
+
"logits/chosen": -0.5366261601448059,
|
| 408 |
+
"logits/rejected": -0.20873865485191345,
|
| 409 |
+
"logps/chosen": -345.1943359375,
|
| 410 |
+
"logps/rejected": -403.20513916015625,
|
| 411 |
+
"loss": 0.5247,
|
| 412 |
+
"rewards/accuracies": 0.737500011920929,
|
| 413 |
+
"rewards/chosen": -0.7760641574859619,
|
| 414 |
+
"rewards/margins": 0.7140566110610962,
|
| 415 |
+
"rewards/rejected": -1.490120768547058,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.2616431187859759,
|
| 420 |
+
"grad_norm": 30.28299029089666,
|
| 421 |
+
"learning_rate": 4.6138518605333664e-07,
|
| 422 |
+
"logits/chosen": -0.47556072473526,
|
| 423 |
+
"logits/rejected": -0.13702349364757538,
|
| 424 |
+
"logps/chosen": -366.315673828125,
|
| 425 |
+
"logps/rejected": -408.03656005859375,
|
| 426 |
+
"loss": 0.5139,
|
| 427 |
+
"rewards/accuracies": 0.699999988079071,
|
| 428 |
+
"rewards/chosen": -0.8853791952133179,
|
| 429 |
+
"rewards/margins": 0.6188973188400269,
|
| 430 |
+
"rewards/rejected": -1.5042766332626343,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.272108843537415,
|
| 435 |
+
"grad_norm": 25.299094246215923,
|
| 436 |
+
"learning_rate": 4.5636328249082514e-07,
|
| 437 |
+
"logits/chosen": -0.2890857756137848,
|
| 438 |
+
"logits/rejected": 0.1314472258090973,
|
| 439 |
+
"logps/chosen": -392.33758544921875,
|
| 440 |
+
"logps/rejected": -448.05511474609375,
|
| 441 |
+
"loss": 0.5298,
|
| 442 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 443 |
+
"rewards/chosen": -1.033733606338501,
|
| 444 |
+
"rewards/margins": 0.6238042116165161,
|
| 445 |
+
"rewards/rejected": -1.657537817955017,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.282574568288854,
|
| 450 |
+
"grad_norm": 34.91396082026137,
|
| 451 |
+
"learning_rate": 4.510653863290871e-07,
|
| 452 |
+
"logits/chosen": -0.4680868089199066,
|
| 453 |
+
"logits/rejected": -0.04308897256851196,
|
| 454 |
+
"logps/chosen": -368.75616455078125,
|
| 455 |
+
"logps/rejected": -397.94854736328125,
|
| 456 |
+
"loss": 0.5079,
|
| 457 |
+
"rewards/accuracies": 0.731249988079071,
|
| 458 |
+
"rewards/chosen": -0.6907842755317688,
|
| 459 |
+
"rewards/margins": 0.7531805038452148,
|
| 460 |
+
"rewards/rejected": -1.4439646005630493,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.29304029304029305,
|
| 465 |
+
"grad_norm": 26.47125396812231,
|
| 466 |
+
"learning_rate": 4.4549858303465737e-07,
|
| 467 |
+
"logits/chosen": -0.11131812632083893,
|
| 468 |
+
"logits/rejected": 0.286846786737442,
|
| 469 |
+
"logps/chosen": -380.48907470703125,
|
| 470 |
+
"logps/rejected": -445.55841064453125,
|
| 471 |
+
"loss": 0.5407,
|
| 472 |
+
"rewards/accuracies": 0.75,
|
| 473 |
+
"rewards/chosen": -0.8424412608146667,
|
| 474 |
+
"rewards/margins": 0.7808824181556702,
|
| 475 |
+
"rewards/rejected": -1.623323678970337,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.3035060177917321,
|
| 480 |
+
"grad_norm": 41.351717296494634,
|
| 481 |
+
"learning_rate": 4.396703177135261e-07,
|
| 482 |
+
"logits/chosen": -0.018317412585020065,
|
| 483 |
+
"logits/rejected": 0.5203948020935059,
|
| 484 |
+
"logps/chosen": -367.64129638671875,
|
| 485 |
+
"logps/rejected": -404.89007568359375,
|
| 486 |
+
"loss": 0.528,
|
| 487 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 488 |
+
"rewards/chosen": -0.9490032196044922,
|
| 489 |
+
"rewards/margins": 0.7295158505439758,
|
| 490 |
+
"rewards/rejected": -1.6785190105438232,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.3139717425431711,
|
| 495 |
+
"grad_norm": 32.926368051569824,
|
| 496 |
+
"learning_rate": 4.335883851539693e-07,
|
| 497 |
+
"logits/chosen": -0.0854339599609375,
|
| 498 |
+
"logits/rejected": 0.22072991728782654,
|
| 499 |
+
"logps/chosen": -367.1238708496094,
|
| 500 |
+
"logps/rejected": -435.1971130371094,
|
| 501 |
+
"loss": 0.5334,
|
| 502 |
+
"rewards/accuracies": 0.706250011920929,
|
| 503 |
+
"rewards/chosen": -0.981222927570343,
|
| 504 |
+
"rewards/margins": 0.6932328939437866,
|
| 505 |
+
"rewards/rejected": -1.674456000328064,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.3139717425431711,
|
| 510 |
+
"eval_logits/chosen": -0.12252414971590042,
|
| 511 |
+
"eval_logits/rejected": 0.39825233817100525,
|
| 512 |
+
"eval_logps/chosen": -366.71929931640625,
|
| 513 |
+
"eval_logps/rejected": -428.3155212402344,
|
| 514 |
+
"eval_loss": 0.522176206111908,
|
| 515 |
+
"eval_rewards/accuracies": 0.7638888955116272,
|
| 516 |
+
"eval_rewards/chosen": -0.8475193977355957,
|
| 517 |
+
"eval_rewards/margins": 0.8336725234985352,
|
| 518 |
+
"eval_rewards/rejected": -1.6811920404434204,
|
| 519 |
+
"eval_runtime": 189.9081,
|
| 520 |
+
"eval_samples_per_second": 10.531,
|
| 521 |
+
"eval_steps_per_second": 0.332,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.32443746729461015,
|
| 526 |
+
"grad_norm": 40.413802800720795,
|
| 527 |
+
"learning_rate": 4.272609194017105e-07,
|
| 528 |
+
"logits/chosen": -0.33145222067832947,
|
| 529 |
+
"logits/rejected": 0.6672115921974182,
|
| 530 |
+
"logps/chosen": -373.406005859375,
|
| 531 |
+
"logps/rejected": -407.01348876953125,
|
| 532 |
+
"loss": 0.5137,
|
| 533 |
+
"rewards/accuracies": 0.737500011920929,
|
| 534 |
+
"rewards/chosen": -0.8969880938529968,
|
| 535 |
+
"rewards/margins": 0.8461712002754211,
|
| 536 |
+
"rewards/rejected": -1.7431590557098389,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.3349031920460492,
|
| 541 |
+
"grad_norm": 31.162326677270173,
|
| 542 |
+
"learning_rate": 4.2069638288135547e-07,
|
| 543 |
+
"logits/chosen": 0.1917075216770172,
|
| 544 |
+
"logits/rejected": 0.7021361589431763,
|
| 545 |
+
"logps/chosen": -353.34674072265625,
|
| 546 |
+
"logps/rejected": -432.170654296875,
|
| 547 |
+
"loss": 0.5424,
|
| 548 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 549 |
+
"rewards/chosen": -1.0030566453933716,
|
| 550 |
+
"rewards/margins": 0.8127084970474243,
|
| 551 |
+
"rewards/rejected": -1.8157650232315063,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.3453689167974882,
|
| 556 |
+
"grad_norm": 41.70557008646781,
|
| 557 |
+
"learning_rate": 4.139035550786494e-07,
|
| 558 |
+
"logits/chosen": 0.1427779197692871,
|
| 559 |
+
"logits/rejected": 0.5713220834732056,
|
| 560 |
+
"logps/chosen": -361.56280517578125,
|
| 561 |
+
"logps/rejected": -405.6873779296875,
|
| 562 |
+
"loss": 0.523,
|
| 563 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 564 |
+
"rewards/chosen": -1.0515334606170654,
|
| 565 |
+
"rewards/margins": 0.6843333840370178,
|
| 566 |
+
"rewards/rejected": -1.7358665466308594,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.35583464154892724,
|
| 571 |
+
"grad_norm": 31.85346501060009,
|
| 572 |
+
"learning_rate": 4.0689152079869306e-07,
|
| 573 |
+
"logits/chosen": 0.33731141686439514,
|
| 574 |
+
"logits/rejected": 0.7817382216453552,
|
| 575 |
+
"logps/chosen": -391.77117919921875,
|
| 576 |
+
"logps/rejected": -430.62860107421875,
|
| 577 |
+
"loss": 0.5412,
|
| 578 |
+
"rewards/accuracies": 0.699999988079071,
|
| 579 |
+
"rewards/chosen": -1.2225059270858765,
|
| 580 |
+
"rewards/margins": 0.723491370677948,
|
| 581 |
+
"rewards/rejected": -1.9459972381591797,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.3663003663003663,
|
| 586 |
+
"grad_norm": 30.001320137405774,
|
| 587 |
+
"learning_rate": 3.99669658015821e-07,
|
| 588 |
+
"logits/chosen": -0.038751162588596344,
|
| 589 |
+
"logits/rejected": 0.41907891631126404,
|
| 590 |
+
"logps/chosen": -384.5907287597656,
|
| 591 |
+
"logps/rejected": -420.61883544921875,
|
| 592 |
+
"loss": 0.5049,
|
| 593 |
+
"rewards/accuracies": 0.737500011920929,
|
| 594 |
+
"rewards/chosen": -0.9339005351066589,
|
| 595 |
+
"rewards/margins": 0.7113012075424194,
|
| 596 |
+
"rewards/rejected": -1.6452019214630127,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.37676609105180536,
|
| 601 |
+
"grad_norm": 37.96262753583076,
|
| 602 |
+
"learning_rate": 3.92247625331392e-07,
|
| 603 |
+
"logits/chosen": -0.024645987898111343,
|
| 604 |
+
"logits/rejected": 0.5247767567634583,
|
| 605 |
+
"logps/chosen": -380.2080993652344,
|
| 606 |
+
"logps/rejected": -438.3382263183594,
|
| 607 |
+
"loss": 0.4811,
|
| 608 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 609 |
+
"rewards/chosen": -0.9434358477592468,
|
| 610 |
+
"rewards/margins": 0.8126322627067566,
|
| 611 |
+
"rewards/rejected": -1.7560679912567139,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.3872318158032444,
|
| 616 |
+
"grad_norm": 36.460849539184395,
|
| 617 |
+
"learning_rate": 3.846353490562664e-07,
|
| 618 |
+
"logits/chosen": 0.21033847332000732,
|
| 619 |
+
"logits/rejected": 0.6614584922790527,
|
| 620 |
+
"logps/chosen": -401.28472900390625,
|
| 621 |
+
"logps/rejected": -445.6495666503906,
|
| 622 |
+
"loss": 0.5156,
|
| 623 |
+
"rewards/accuracies": 0.706250011920929,
|
| 624 |
+
"rewards/chosen": -1.3794898986816406,
|
| 625 |
+
"rewards/margins": 0.720534086227417,
|
| 626 |
+
"rewards/rejected": -2.1000239849090576,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.3976975405546834,
|
| 631 |
+
"grad_norm": 39.504972287950515,
|
| 632 |
+
"learning_rate": 3.768430099352445e-07,
|
| 633 |
+
"logits/chosen": -0.20532044768333435,
|
| 634 |
+
"logits/rejected": 0.3994311988353729,
|
| 635 |
+
"logps/chosen": -399.9379577636719,
|
| 636 |
+
"logps/rejected": -469.12432861328125,
|
| 637 |
+
"loss": 0.5166,
|
| 638 |
+
"rewards/accuracies": 0.737500011920929,
|
| 639 |
+
"rewards/chosen": -1.4527246952056885,
|
| 640 |
+
"rewards/margins": 0.7748172879219055,
|
| 641 |
+
"rewards/rejected": -2.2275421619415283,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.40816326530612246,
|
| 646 |
+
"grad_norm": 32.70806505808847,
|
| 647 |
+
"learning_rate": 3.6888102953122304e-07,
|
| 648 |
+
"logits/chosen": -0.5273795127868652,
|
| 649 |
+
"logits/rejected": -0.03774275258183479,
|
| 650 |
+
"logps/chosen": -406.19122314453125,
|
| 651 |
+
"logps/rejected": -465.33441162109375,
|
| 652 |
+
"loss": 0.5367,
|
| 653 |
+
"rewards/accuracies": 0.731249988079071,
|
| 654 |
+
"rewards/chosen": -1.2459166049957275,
|
| 655 |
+
"rewards/margins": 0.7241197228431702,
|
| 656 |
+
"rewards/rejected": -1.970036268234253,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.4186289900575615,
|
| 661 |
+
"grad_norm": 94.38815430243706,
|
| 662 |
+
"learning_rate": 3.607600562872785e-07,
|
| 663 |
+
"logits/chosen": -0.8146815299987793,
|
| 664 |
+
"logits/rejected": -0.4568367004394531,
|
| 665 |
+
"logps/chosen": -379.3426818847656,
|
| 666 |
+
"logps/rejected": -437.07220458984375,
|
| 667 |
+
"loss": 0.5426,
|
| 668 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 669 |
+
"rewards/chosen": -1.0935862064361572,
|
| 670 |
+
"rewards/margins": 0.6003715991973877,
|
| 671 |
+
"rewards/rejected": -1.6939579248428345,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.4186289900575615,
|
| 676 |
+
"eval_logits/chosen": -0.9768523573875427,
|
| 677 |
+
"eval_logits/rejected": -0.3232794404029846,
|
| 678 |
+
"eval_logps/chosen": -378.2563781738281,
|
| 679 |
+
"eval_logps/rejected": -432.01129150390625,
|
| 680 |
+
"eval_loss": 0.5111254453659058,
|
| 681 |
+
"eval_rewards/accuracies": 0.7757936716079712,
|
| 682 |
+
"eval_rewards/chosen": -0.9628898501396179,
|
| 683 |
+
"eval_rewards/margins": 0.755260169506073,
|
| 684 |
+
"eval_rewards/rejected": -1.7181499004364014,
|
| 685 |
+
"eval_runtime": 196.1755,
|
| 686 |
+
"eval_samples_per_second": 10.195,
|
| 687 |
+
"eval_steps_per_second": 0.321,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.858189429618001,
|
| 692 |
+
"grad_norm": 24.669643752510265,
|
| 693 |
+
"learning_rate": 2.9492720416985e-08,
|
| 694 |
+
"logits/chosen": -0.8584449887275696,
|
| 695 |
+
"logits/rejected": -0.14207491278648376,
|
| 696 |
+
"logps/chosen": -397.1089782714844,
|
| 697 |
+
"logps/rejected": -437.30596923828125,
|
| 698 |
+
"loss": 0.5181,
|
| 699 |
+
"rewards/accuracies": 0.753125011920929,
|
| 700 |
+
"rewards/chosen": -0.9909969568252563,
|
| 701 |
+
"rewards/margins": 0.7339824438095093,
|
| 702 |
+
"rewards/rejected": -1.7249791622161865,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.8791208791208791,
|
| 707 |
+
"grad_norm": 22.57001710277683,
|
| 708 |
+
"learning_rate": 2.1464952759020856e-08,
|
| 709 |
+
"logits/chosen": -0.6151202321052551,
|
| 710 |
+
"logits/rejected": 0.10300643742084503,
|
| 711 |
+
"logps/chosen": -381.1826171875,
|
| 712 |
+
"logps/rejected": -446.4559020996094,
|
| 713 |
+
"loss": 0.5086,
|
| 714 |
+
"rewards/accuracies": 0.7281249761581421,
|
| 715 |
+
"rewards/chosen": -1.1097198724746704,
|
| 716 |
+
"rewards/margins": 0.7756140232086182,
|
| 717 |
+
"rewards/rejected": -1.8853340148925781,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.9000523286237572,
|
| 722 |
+
"grad_norm": 22.069784459721443,
|
| 723 |
+
"learning_rate": 1.4662207078575684e-08,
|
| 724 |
+
"logits/chosen": -0.37786802649497986,
|
| 725 |
+
"logits/rejected": 0.4924679398536682,
|
| 726 |
+
"logps/chosen": -396.33514404296875,
|
| 727 |
+
"logps/rejected": -442.93133544921875,
|
| 728 |
+
"loss": 0.4889,
|
| 729 |
+
"rewards/accuracies": 0.78125,
|
| 730 |
+
"rewards/chosen": -1.1174886226654053,
|
| 731 |
+
"rewards/margins": 0.8376533389091492,
|
| 732 |
+
"rewards/rejected": -1.9551417827606201,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9209837781266352,
|
| 737 |
+
"grad_norm": 22.352532911361152,
|
| 738 |
+
"learning_rate": 9.12094829893642e-09,
|
| 739 |
+
"logits/chosen": -0.554622232913971,
|
| 740 |
+
"logits/rejected": 0.22521421313285828,
|
| 741 |
+
"logps/chosen": -395.46380615234375,
|
| 742 |
+
"logps/rejected": -459.3907165527344,
|
| 743 |
+
"loss": 0.5052,
|
| 744 |
+
"rewards/accuracies": 0.762499988079071,
|
| 745 |
+
"rewards/chosen": -1.075073480606079,
|
| 746 |
+
"rewards/margins": 0.8130553960800171,
|
| 747 |
+
"rewards/rejected": -1.8881288766860962,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.9419152276295133,
|
| 752 |
+
"grad_norm": 25.721981849294604,
|
| 753 |
+
"learning_rate": 4.8708793644441086e-09,
|
| 754 |
+
"logits/chosen": -0.4537356495857239,
|
| 755 |
+
"logits/rejected": 0.10888534784317017,
|
| 756 |
+
"logps/chosen": -398.93768310546875,
|
| 757 |
+
"logps/rejected": -474.32183837890625,
|
| 758 |
+
"loss": 0.5042,
|
| 759 |
+
"rewards/accuracies": 0.746874988079071,
|
| 760 |
+
"rewards/chosen": -1.1490033864974976,
|
| 761 |
+
"rewards/margins": 0.7859469056129456,
|
| 762 |
+
"rewards/rejected": -1.9349502325057983,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.9628466771323915,
|
| 767 |
+
"grad_norm": 23.216240925245994,
|
| 768 |
+
"learning_rate": 1.9347820230782295e-09,
|
| 769 |
+
"logits/chosen": -0.3090813159942627,
|
| 770 |
+
"logits/rejected": 0.3994906544685364,
|
| 771 |
+
"logps/chosen": -367.4607849121094,
|
| 772 |
+
"logps/rejected": -421.9041442871094,
|
| 773 |
+
"loss": 0.5189,
|
| 774 |
+
"rewards/accuracies": 0.746874988079071,
|
| 775 |
+
"rewards/chosen": -1.123751163482666,
|
| 776 |
+
"rewards/margins": 0.7729546427726746,
|
| 777 |
+
"rewards/rejected": -1.8967056274414062,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.9837781266352695,
|
| 782 |
+
"grad_norm": 23.107361527734533,
|
| 783 |
+
"learning_rate": 3.2839470889836627e-10,
|
| 784 |
+
"logits/chosen": -0.514792799949646,
|
| 785 |
+
"logits/rejected": 0.3938598036766052,
|
| 786 |
+
"logps/chosen": -395.8327941894531,
|
| 787 |
+
"logps/rejected": -452.728515625,
|
| 788 |
+
"loss": 0.499,
|
| 789 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 790 |
+
"rewards/chosen": -1.0682015419006348,
|
| 791 |
+
"rewards/margins": 0.8392928838729858,
|
| 792 |
+
"rewards/rejected": -1.9074945449829102,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.9984301412872841,
|
| 797 |
+
"step": 477,
|
| 798 |
"total_flos": 0.0,
|
| 799 |
+
"train_loss": 0.08199021051514824,
|
| 800 |
+
"train_runtime": 2352.184,
|
| 801 |
+
"train_samples_per_second": 25.99,
|
| 802 |
+
"train_steps_per_second": 0.203
|
| 803 |
}
|
| 804 |
],
|
| 805 |
"logging_steps": 10,
|
| 806 |
+
"max_steps": 477,
|
| 807 |
"num_input_tokens_seen": 0,
|
| 808 |
"num_train_epochs": 1,
|
| 809 |
"save_steps": 100,
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7480
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:48e654a4d8c13baba448e98ee4024b51a157268d98e1173701b71f5a317dc8f5
|
| 3 |
size 7480
|