Model save
Browse files
README.md
ADDED
|
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: llama2
|
| 3 |
+
base_model: llama-2-nl/Llama-2-7b-hf-lora-original-sft
|
| 4 |
+
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- dpo
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
model-index:
|
| 9 |
+
- name: Llama-2-7b-hf-lora-original-it
|
| 10 |
+
results: []
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 14 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 15 |
+
|
| 16 |
+
# Llama-2-7b-hf-lora-original-it
|
| 17 |
+
|
| 18 |
+
This model is a fine-tuned version of [llama-2-nl/Llama-2-7b-hf-lora-original-sft](https://huggingface.co/llama-2-nl/Llama-2-7b-hf-lora-original-sft) on an unknown dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.3540
|
| 21 |
+
- Rewards/chosen: 0.1166
|
| 22 |
+
- Rewards/rejected: -0.9259
|
| 23 |
+
- Rewards/accuracies: 0.9396
|
| 24 |
+
- Rewards/margins: 1.0425
|
| 25 |
+
- Logps/rejected: -547.4224
|
| 26 |
+
- Logps/chosen: -600.8121
|
| 27 |
+
- Logits/rejected: -0.8736
|
| 28 |
+
- Logits/chosen: -0.9600
|
| 29 |
+
|
| 30 |
+
## Model description
|
| 31 |
+
|
| 32 |
+
More information needed
|
| 33 |
+
|
| 34 |
+
## Intended uses & limitations
|
| 35 |
+
|
| 36 |
+
More information needed
|
| 37 |
+
|
| 38 |
+
## Training and evaluation data
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Training procedure
|
| 43 |
+
|
| 44 |
+
### Training hyperparameters
|
| 45 |
+
|
| 46 |
+
The following hyperparameters were used during training:
|
| 47 |
+
- learning_rate: 5e-07
|
| 48 |
+
- train_batch_size: 4
|
| 49 |
+
- eval_batch_size: 4
|
| 50 |
+
- seed: 42
|
| 51 |
+
- distributed_type: multi-GPU
|
| 52 |
+
- num_devices: 4
|
| 53 |
+
- gradient_accumulation_steps: 4
|
| 54 |
+
- total_train_batch_size: 64
|
| 55 |
+
- total_eval_batch_size: 16
|
| 56 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 57 |
+
- lr_scheduler_type: cosine
|
| 58 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 59 |
+
- num_epochs: 1
|
| 60 |
+
|
| 61 |
+
### Training results
|
| 62 |
+
|
| 63 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 64 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 65 |
+
| 0.5984 | 0.1327 | 100 | 0.5904 | 0.0549 | -0.1735 | 0.9030 | 0.2283 | -539.8975 | -601.4293 | -1.1606 | -1.1395 |
|
| 66 |
+
| 0.4622 | 0.2653 | 200 | 0.4581 | 0.1134 | -0.4980 | 0.9351 | 0.6113 | -543.1426 | -600.8441 | -1.2714 | -1.2180 |
|
| 67 |
+
| 0.3934 | 0.3980 | 300 | 0.3959 | 0.1263 | -0.7212 | 0.9366 | 0.8475 | -545.3747 | -600.7144 | -1.0528 | -1.0755 |
|
| 68 |
+
| 0.3629 | 0.5307 | 400 | 0.3674 | 0.1170 | -0.8608 | 0.9381 | 0.9777 | -546.7705 | -600.8080 | -1.1109 | -1.1154 |
|
| 69 |
+
| 0.3556 | 0.6633 | 500 | 0.3561 | 0.1136 | -0.9146 | 0.9388 | 1.0282 | -547.3090 | -600.8419 | -0.8266 | -0.9289 |
|
| 70 |
+
| 0.3488 | 0.7960 | 600 | 0.3540 | 0.1104 | -0.9310 | 0.9410 | 1.0415 | -547.4734 | -600.8737 | -1.0676 | -1.0877 |
|
| 71 |
+
| 0.3563 | 0.9287 | 700 | 0.3540 | 0.1166 | -0.9259 | 0.9396 | 1.0425 | -547.4224 | -600.8121 | -0.8736 | -0.9600 |
|
| 72 |
+
|
| 73 |
+
|
| 74 |
+
### Framework versions
|
| 75 |
+
|
| 76 |
+
- Transformers 4.40.1
|
| 77 |
+
- Pytorch 2.1.2+cu121
|
| 78 |
+
- Datasets 2.19.0
|
| 79 |
+
- Tokenizers 0.19.1
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9990049751243781,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.4291752041731856,
|
| 5 |
+
"train_runtime": 22849.6211,
|
| 6 |
+
"train_samples": 48227,
|
| 7 |
+
"train_samples_per_second": 2.111,
|
| 8 |
+
"train_steps_per_second": 0.033
|
| 9 |
+
}
|
runs/May20_00-51-42_gpu530.dodrio.os/events.out.tfevents.1716160119.gpu530.dodrio.os.1926607.0
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 sha256:0d202074ab4ae56d82d901450a52c38c647e371ba23392ab84221fbd7bfab1d5
|
| 3 |
+
size 62846
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0.9990049751243781,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.4291752041731856,
|
| 5 |
+
"train_runtime": 22849.6211,
|
| 6 |
+
"train_samples": 48227,
|
| 7 |
+
"train_samples_per_second": 2.111,
|
| 8 |
+
"train_steps_per_second": 0.033
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1282 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9990049751243781,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 753,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0013266998341625207,
|
| 13 |
+
"grad_norm": 40.25,
|
| 14 |
+
"learning_rate": 6.578947368421052e-09,
|
| 15 |
+
"logits/chosen": -1.2802138328552246,
|
| 16 |
+
"logits/rejected": -1.3739961385726929,
|
| 17 |
+
"logps/chosen": -584.777587890625,
|
| 18 |
+
"logps/rejected": -533.882080078125,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.0,
|
| 21 |
+
"rewards/chosen": 0.0,
|
| 22 |
+
"rewards/margins": 0.0,
|
| 23 |
+
"rewards/rejected": 0.0,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.013266998341625208,
|
| 28 |
+
"grad_norm": 54.5,
|
| 29 |
+
"learning_rate": 6.578947368421052e-08,
|
| 30 |
+
"logits/chosen": -1.1525533199310303,
|
| 31 |
+
"logits/rejected": -1.1556764841079712,
|
| 32 |
+
"logps/chosen": -577.8804321289062,
|
| 33 |
+
"logps/rejected": -498.16986083984375,
|
| 34 |
+
"loss": 0.6903,
|
| 35 |
+
"rewards/accuracies": 0.4930555522441864,
|
| 36 |
+
"rewards/chosen": 0.00347831379622221,
|
| 37 |
+
"rewards/margins": 0.00828113965690136,
|
| 38 |
+
"rewards/rejected": -0.004802825395017862,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.026533996683250415,
|
| 43 |
+
"grad_norm": 44.75,
|
| 44 |
+
"learning_rate": 1.3157894736842104e-07,
|
| 45 |
+
"logits/chosen": -1.1788235902786255,
|
| 46 |
+
"logits/rejected": -1.2242963314056396,
|
| 47 |
+
"logps/chosen": -611.204833984375,
|
| 48 |
+
"logps/rejected": -543.866455078125,
|
| 49 |
+
"loss": 0.6928,
|
| 50 |
+
"rewards/accuracies": 0.4937500059604645,
|
| 51 |
+
"rewards/chosen": 0.0023336124140769243,
|
| 52 |
+
"rewards/margins": 0.003125081304460764,
|
| 53 |
+
"rewards/rejected": -0.0007914667949080467,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.03980099502487562,
|
| 58 |
+
"grad_norm": 36.25,
|
| 59 |
+
"learning_rate": 1.9736842105263157e-07,
|
| 60 |
+
"logits/chosen": -1.158151388168335,
|
| 61 |
+
"logits/rejected": -1.1601974964141846,
|
| 62 |
+
"logps/chosen": -633.5345458984375,
|
| 63 |
+
"logps/rejected": -536.1189575195312,
|
| 64 |
+
"loss": 0.6905,
|
| 65 |
+
"rewards/accuracies": 0.581250011920929,
|
| 66 |
+
"rewards/chosen": 0.013027493841946125,
|
| 67 |
+
"rewards/margins": 0.015905674546957016,
|
| 68 |
+
"rewards/rejected": -0.0028781811706721783,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.05306799336650083,
|
| 73 |
+
"grad_norm": 42.5,
|
| 74 |
+
"learning_rate": 2.631578947368421e-07,
|
| 75 |
+
"logits/chosen": -1.19637930393219,
|
| 76 |
+
"logits/rejected": -1.22651207447052,
|
| 77 |
+
"logps/chosen": -540.5403442382812,
|
| 78 |
+
"logps/rejected": -552.4425048828125,
|
| 79 |
+
"loss": 0.687,
|
| 80 |
+
"rewards/accuracies": 0.518750011920929,
|
| 81 |
+
"rewards/chosen": 0.0009165151277557015,
|
| 82 |
+
"rewards/margins": 0.011455372907221317,
|
| 83 |
+
"rewards/rejected": -0.01053885743021965,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.06633499170812604,
|
| 88 |
+
"grad_norm": 48.25,
|
| 89 |
+
"learning_rate": 3.2894736842105264e-07,
|
| 90 |
+
"logits/chosen": -1.198271632194519,
|
| 91 |
+
"logits/rejected": -1.2518165111541748,
|
| 92 |
+
"logps/chosen": -579.4686279296875,
|
| 93 |
+
"logps/rejected": -571.4375,
|
| 94 |
+
"loss": 0.6818,
|
| 95 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 96 |
+
"rewards/chosen": 0.005964647978544235,
|
| 97 |
+
"rewards/margins": 0.019765758886933327,
|
| 98 |
+
"rewards/rejected": -0.013801109977066517,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.07960199004975124,
|
| 103 |
+
"grad_norm": 39.25,
|
| 104 |
+
"learning_rate": 3.9473684210526315e-07,
|
| 105 |
+
"logits/chosen": -1.1785616874694824,
|
| 106 |
+
"logits/rejected": -1.226552963256836,
|
| 107 |
+
"logps/chosen": -595.6961059570312,
|
| 108 |
+
"logps/rejected": -572.9510498046875,
|
| 109 |
+
"loss": 0.6691,
|
| 110 |
+
"rewards/accuracies": 0.731249988079071,
|
| 111 |
+
"rewards/chosen": 0.022173848003149033,
|
| 112 |
+
"rewards/margins": 0.06000928208231926,
|
| 113 |
+
"rewards/rejected": -0.03783543407917023,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.09286898839137644,
|
| 118 |
+
"grad_norm": 44.5,
|
| 119 |
+
"learning_rate": 4.6052631578947365e-07,
|
| 120 |
+
"logits/chosen": -1.1982749700546265,
|
| 121 |
+
"logits/rejected": -1.236537218093872,
|
| 122 |
+
"logps/chosen": -634.6461181640625,
|
| 123 |
+
"logps/rejected": -673.8761596679688,
|
| 124 |
+
"loss": 0.6543,
|
| 125 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 126 |
+
"rewards/chosen": 0.02825814113020897,
|
| 127 |
+
"rewards/margins": 0.08534505218267441,
|
| 128 |
+
"rewards/rejected": -0.05708691477775574,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.10613598673300166,
|
| 133 |
+
"grad_norm": 36.75,
|
| 134 |
+
"learning_rate": 4.999569334646955e-07,
|
| 135 |
+
"logits/chosen": -1.0726072788238525,
|
| 136 |
+
"logits/rejected": -1.1171576976776123,
|
| 137 |
+
"logps/chosen": -614.9038696289062,
|
| 138 |
+
"logps/rejected": -572.2459716796875,
|
| 139 |
+
"loss": 0.638,
|
| 140 |
+
"rewards/accuracies": 0.875,
|
| 141 |
+
"rewards/chosen": 0.04197516664862633,
|
| 142 |
+
"rewards/margins": 0.1271333396434784,
|
| 143 |
+
"rewards/rejected": -0.08515818417072296,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.11940298507462686,
|
| 148 |
+
"grad_norm": 38.25,
|
| 149 |
+
"learning_rate": 4.994726053293702e-07,
|
| 150 |
+
"logits/chosen": -1.1955012083053589,
|
| 151 |
+
"logits/rejected": -1.2350232601165771,
|
| 152 |
+
"logps/chosen": -590.61376953125,
|
| 153 |
+
"logps/rejected": -564.5113525390625,
|
| 154 |
+
"loss": 0.6237,
|
| 155 |
+
"rewards/accuracies": 0.8812500238418579,
|
| 156 |
+
"rewards/chosen": 0.031416941434144974,
|
| 157 |
+
"rewards/margins": 0.15016348659992218,
|
| 158 |
+
"rewards/rejected": -0.1187465563416481,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.13266998341625208,
|
| 163 |
+
"grad_norm": 39.0,
|
| 164 |
+
"learning_rate": 4.984511621268102e-07,
|
| 165 |
+
"logits/chosen": -1.162690281867981,
|
| 166 |
+
"logits/rejected": -1.1824209690093994,
|
| 167 |
+
"logps/chosen": -573.8049926757812,
|
| 168 |
+
"logps/rejected": -528.6422119140625,
|
| 169 |
+
"loss": 0.5984,
|
| 170 |
+
"rewards/accuracies": 0.90625,
|
| 171 |
+
"rewards/chosen": 0.038635846227407455,
|
| 172 |
+
"rewards/margins": 0.2009139508008957,
|
| 173 |
+
"rewards/rejected": -0.16227811574935913,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.13266998341625208,
|
| 178 |
+
"eval_logits/chosen": -1.139477252960205,
|
| 179 |
+
"eval_logits/rejected": -1.160577416419983,
|
| 180 |
+
"eval_logps/chosen": -601.4292602539062,
|
| 181 |
+
"eval_logps/rejected": -539.8974609375,
|
| 182 |
+
"eval_loss": 0.5903807878494263,
|
| 183 |
+
"eval_rewards/accuracies": 0.9029850959777832,
|
| 184 |
+
"eval_rewards/chosen": 0.05485348403453827,
|
| 185 |
+
"eval_rewards/margins": 0.22831708192825317,
|
| 186 |
+
"eval_rewards/rejected": -0.1734635829925537,
|
| 187 |
+
"eval_runtime": 685.2004,
|
| 188 |
+
"eval_samples_per_second": 7.821,
|
| 189 |
+
"eval_steps_per_second": 0.489,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.14593698175787728,
|
| 194 |
+
"grad_norm": 33.5,
|
| 195 |
+
"learning_rate": 4.968948030264742e-07,
|
| 196 |
+
"logits/chosen": -1.1363273859024048,
|
| 197 |
+
"logits/rejected": -1.1610157489776611,
|
| 198 |
+
"logps/chosen": -592.44482421875,
|
| 199 |
+
"logps/rejected": -573.2866821289062,
|
| 200 |
+
"loss": 0.582,
|
| 201 |
+
"rewards/accuracies": 0.918749988079071,
|
| 202 |
+
"rewards/chosen": 0.07824783772230148,
|
| 203 |
+
"rewards/margins": 0.2655600905418396,
|
| 204 |
+
"rewards/rejected": -0.1873122602701187,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.15920398009950248,
|
| 209 |
+
"grad_norm": 33.25,
|
| 210 |
+
"learning_rate": 4.948068788729238e-07,
|
| 211 |
+
"logits/chosen": -1.1630527973175049,
|
| 212 |
+
"logits/rejected": -1.202096939086914,
|
| 213 |
+
"logps/chosen": -579.3135375976562,
|
| 214 |
+
"logps/rejected": -578.4791259765625,
|
| 215 |
+
"loss": 0.5691,
|
| 216 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 217 |
+
"rewards/chosen": 0.07454140484333038,
|
| 218 |
+
"rewards/margins": 0.28431838750839233,
|
| 219 |
+
"rewards/rejected": -0.20977696776390076,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.1724709784411277,
|
| 224 |
+
"grad_norm": 35.25,
|
| 225 |
+
"learning_rate": 4.921918849714475e-07,
|
| 226 |
+
"logits/chosen": -1.185011625289917,
|
| 227 |
+
"logits/rejected": -1.1927728652954102,
|
| 228 |
+
"logps/chosen": -621.1232299804688,
|
| 229 |
+
"logps/rejected": -598.525634765625,
|
| 230 |
+
"loss": 0.5436,
|
| 231 |
+
"rewards/accuracies": 0.918749988079071,
|
| 232 |
+
"rewards/chosen": 0.07992725074291229,
|
| 233 |
+
"rewards/margins": 0.3480888307094574,
|
| 234 |
+
"rewards/rejected": -0.2681615948677063,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.1857379767827529,
|
| 239 |
+
"grad_norm": 31.375,
|
| 240 |
+
"learning_rate": 4.890554514096591e-07,
|
| 241 |
+
"logits/chosen": -1.1601734161376953,
|
| 242 |
+
"logits/rejected": -1.1828594207763672,
|
| 243 |
+
"logps/chosen": -589.795654296875,
|
| 244 |
+
"logps/rejected": -544.245849609375,
|
| 245 |
+
"loss": 0.5303,
|
| 246 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 247 |
+
"rewards/chosen": 0.06376471370458603,
|
| 248 |
+
"rewards/margins": 0.3611634075641632,
|
| 249 |
+
"rewards/rejected": -0.29739871621131897,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.19900497512437812,
|
| 254 |
+
"grad_norm": 28.25,
|
| 255 |
+
"learning_rate": 4.854043309359063e-07,
|
| 256 |
+
"logits/chosen": -1.2025436162948608,
|
| 257 |
+
"logits/rejected": -1.2480312585830688,
|
| 258 |
+
"logps/chosen": -550.0576782226562,
|
| 259 |
+
"logps/rejected": -483.212646484375,
|
| 260 |
+
"loss": 0.5173,
|
| 261 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 262 |
+
"rewards/chosen": 0.08968071639537811,
|
| 263 |
+
"rewards/margins": 0.39301761984825134,
|
| 264 |
+
"rewards/rejected": -0.30333688855171204,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.21227197346600332,
|
| 269 |
+
"grad_norm": 28.125,
|
| 270 |
+
"learning_rate": 4.812463844205884e-07,
|
| 271 |
+
"logits/chosen": -1.1907384395599365,
|
| 272 |
+
"logits/rejected": -1.218056321144104,
|
| 273 |
+
"logps/chosen": -612.6536865234375,
|
| 274 |
+
"logps/rejected": -552.7808837890625,
|
| 275 |
+
"loss": 0.5032,
|
| 276 |
+
"rewards/accuracies": 0.9375,
|
| 277 |
+
"rewards/chosen": 0.12153647094964981,
|
| 278 |
+
"rewards/margins": 0.4728039801120758,
|
| 279 |
+
"rewards/rejected": -0.3512675166130066,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.22553897180762852,
|
| 284 |
+
"grad_norm": 27.125,
|
| 285 |
+
"learning_rate": 4.7659056393168604e-07,
|
| 286 |
+
"logits/chosen": -1.2418904304504395,
|
| 287 |
+
"logits/rejected": -1.3286497592926025,
|
| 288 |
+
"logps/chosen": -532.3972778320312,
|
| 289 |
+
"logps/rejected": -481.52459716796875,
|
| 290 |
+
"loss": 0.5031,
|
| 291 |
+
"rewards/accuracies": 0.887499988079071,
|
| 292 |
+
"rewards/chosen": 0.10282758623361588,
|
| 293 |
+
"rewards/margins": 0.44313424825668335,
|
| 294 |
+
"rewards/rejected": -0.3403066396713257,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.23880597014925373,
|
| 299 |
+
"grad_norm": 27.0,
|
| 300 |
+
"learning_rate": 4.714468934609381e-07,
|
| 301 |
+
"logits/chosen": -1.2157742977142334,
|
| 302 |
+
"logits/rejected": -1.2327635288238525,
|
| 303 |
+
"logps/chosen": -541.45361328125,
|
| 304 |
+
"logps/rejected": -482.06103515625,
|
| 305 |
+
"loss": 0.4824,
|
| 306 |
+
"rewards/accuracies": 0.96875,
|
| 307 |
+
"rewards/chosen": 0.1126769408583641,
|
| 308 |
+
"rewards/margins": 0.5232059359550476,
|
| 309 |
+
"rewards/rejected": -0.4105289876461029,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.25207296849087896,
|
| 314 |
+
"grad_norm": 23.875,
|
| 315 |
+
"learning_rate": 4.658264473421659e-07,
|
| 316 |
+
"logits/chosen": -1.2209118604660034,
|
| 317 |
+
"logits/rejected": -1.2742892503738403,
|
| 318 |
+
"logps/chosen": -600.0318603515625,
|
| 319 |
+
"logps/rejected": -533.5223388671875,
|
| 320 |
+
"loss": 0.4874,
|
| 321 |
+
"rewards/accuracies": 0.893750011920929,
|
| 322 |
+
"rewards/chosen": 0.09032230079174042,
|
| 323 |
+
"rewards/margins": 0.5253477096557617,
|
| 324 |
+
"rewards/rejected": -0.4350253939628601,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.26533996683250416,
|
| 329 |
+
"grad_norm": 30.125,
|
| 330 |
+
"learning_rate": 4.597413264082086e-07,
|
| 331 |
+
"logits/chosen": -1.2209361791610718,
|
| 332 |
+
"logits/rejected": -1.295668125152588,
|
| 333 |
+
"logps/chosen": -573.05859375,
|
| 334 |
+
"logps/rejected": -572.3590087890625,
|
| 335 |
+
"loss": 0.4622,
|
| 336 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 337 |
+
"rewards/chosen": 0.08147934824228287,
|
| 338 |
+
"rewards/margins": 0.5580233931541443,
|
| 339 |
+
"rewards/rejected": -0.4765440821647644,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.26533996683250416,
|
| 344 |
+
"eval_logits/chosen": -1.2179902791976929,
|
| 345 |
+
"eval_logits/rejected": -1.271428108215332,
|
| 346 |
+
"eval_logps/chosen": -600.8441162109375,
|
| 347 |
+
"eval_logps/rejected": -543.1426391601562,
|
| 348 |
+
"eval_loss": 0.45805710554122925,
|
| 349 |
+
"eval_rewards/accuracies": 0.9350746273994446,
|
| 350 |
+
"eval_rewards/chosen": 0.11337064951658249,
|
| 351 |
+
"eval_rewards/margins": 0.6113449931144714,
|
| 352 |
+
"eval_rewards/rejected": -0.49797430634498596,
|
| 353 |
+
"eval_runtime": 676.1331,
|
| 354 |
+
"eval_samples_per_second": 7.926,
|
| 355 |
+
"eval_steps_per_second": 0.495,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.27860696517412936,
|
| 360 |
+
"grad_norm": 25.25,
|
| 361 |
+
"learning_rate": 4.5320463193780256e-07,
|
| 362 |
+
"logits/chosen": -1.1879713535308838,
|
| 363 |
+
"logits/rejected": -1.234440565109253,
|
| 364 |
+
"logps/chosen": -570.4061279296875,
|
| 365 |
+
"logps/rejected": -540.0577392578125,
|
| 366 |
+
"loss": 0.4485,
|
| 367 |
+
"rewards/accuracies": 0.918749988079071,
|
| 368 |
+
"rewards/chosen": 0.1254458725452423,
|
| 369 |
+
"rewards/margins": 0.5999458432197571,
|
| 370 |
+
"rewards/rejected": -0.4744999408721924,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.29187396351575456,
|
| 375 |
+
"grad_norm": 22.625,
|
| 376 |
+
"learning_rate": 4.4623043744850044e-07,
|
| 377 |
+
"logits/chosen": -1.1740987300872803,
|
| 378 |
+
"logits/rejected": -1.2075783014297485,
|
| 379 |
+
"logps/chosen": -588.6426391601562,
|
| 380 |
+
"logps/rejected": -534.1394653320312,
|
| 381 |
+
"loss": 0.4551,
|
| 382 |
+
"rewards/accuracies": 0.925000011920929,
|
| 383 |
+
"rewards/chosen": 0.11971279233694077,
|
| 384 |
+
"rewards/margins": 0.642697811126709,
|
| 385 |
+
"rewards/rejected": -0.5229849815368652,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.30514096185737977,
|
| 390 |
+
"grad_norm": 24.875,
|
| 391 |
+
"learning_rate": 4.388337583963563e-07,
|
| 392 |
+
"logits/chosen": -1.1734439134597778,
|
| 393 |
+
"logits/rejected": -1.1642463207244873,
|
| 394 |
+
"logps/chosen": -630.2869873046875,
|
| 395 |
+
"logps/rejected": -650.8348388671875,
|
| 396 |
+
"loss": 0.4326,
|
| 397 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 398 |
+
"rewards/chosen": 0.11086304485797882,
|
| 399 |
+
"rewards/margins": 0.7283642888069153,
|
| 400 |
+
"rewards/rejected": -0.6175012588500977,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.31840796019900497,
|
| 405 |
+
"grad_norm": 28.375,
|
| 406 |
+
"learning_rate": 4.31030519847616e-07,
|
| 407 |
+
"logits/chosen": -1.170459508895874,
|
| 408 |
+
"logits/rejected": -1.1728956699371338,
|
| 409 |
+
"logps/chosen": -568.0635986328125,
|
| 410 |
+
"logps/rejected": -495.91094970703125,
|
| 411 |
+
"loss": 0.4293,
|
| 412 |
+
"rewards/accuracies": 0.956250011920929,
|
| 413 |
+
"rewards/chosen": 0.09963791817426682,
|
| 414 |
+
"rewards/margins": 0.654728889465332,
|
| 415 |
+
"rewards/rejected": -0.5550910234451294,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.33167495854063017,
|
| 420 |
+
"grad_norm": 22.875,
|
| 421 |
+
"learning_rate": 4.2283752219201464e-07,
|
| 422 |
+
"logits/chosen": -1.0885179042816162,
|
| 423 |
+
"logits/rejected": -1.133748173713684,
|
| 424 |
+
"logps/chosen": -556.9310913085938,
|
| 425 |
+
"logps/rejected": -509.1935119628906,
|
| 426 |
+
"loss": 0.4244,
|
| 427 |
+
"rewards/accuracies": 0.9375,
|
| 428 |
+
"rewards/chosen": 0.10411699116230011,
|
| 429 |
+
"rewards/margins": 0.6922882795333862,
|
| 430 |
+
"rewards/rejected": -0.5881712436676025,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.3449419568822554,
|
| 435 |
+
"grad_norm": 24.75,
|
| 436 |
+
"learning_rate": 4.1427240497150047e-07,
|
| 437 |
+
"logits/chosen": -1.132869839668274,
|
| 438 |
+
"logits/rejected": -1.1097866296768188,
|
| 439 |
+
"logps/chosen": -604.4608154296875,
|
| 440 |
+
"logps/rejected": -561.1189575195312,
|
| 441 |
+
"loss": 0.4112,
|
| 442 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 443 |
+
"rewards/chosen": 0.10400505363941193,
|
| 444 |
+
"rewards/margins": 0.708010196685791,
|
| 445 |
+
"rewards/rejected": -0.6040050983428955,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.3582089552238806,
|
| 450 |
+
"grad_norm": 27.125,
|
| 451 |
+
"learning_rate": 4.053536089022623e-07,
|
| 452 |
+
"logits/chosen": -1.1613821983337402,
|
| 453 |
+
"logits/rejected": -1.195441722869873,
|
| 454 |
+
"logps/chosen": -502.41607666015625,
|
| 455 |
+
"logps/rejected": -479.5301208496094,
|
| 456 |
+
"loss": 0.4202,
|
| 457 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 458 |
+
"rewards/chosen": 0.11941847950220108,
|
| 459 |
+
"rewards/margins": 0.7337124347686768,
|
| 460 |
+
"rewards/rejected": -0.6142939329147339,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.3714759535655058,
|
| 465 |
+
"grad_norm": 27.625,
|
| 466 |
+
"learning_rate": 3.9610033617182715e-07,
|
| 467 |
+
"logits/chosen": -1.1407119035720825,
|
| 468 |
+
"logits/rejected": -1.1306835412979126,
|
| 469 |
+
"logps/chosen": -632.7389526367188,
|
| 470 |
+
"logps/rejected": -621.2080078125,
|
| 471 |
+
"loss": 0.4085,
|
| 472 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 473 |
+
"rewards/chosen": 0.15029093623161316,
|
| 474 |
+
"rewards/margins": 0.8855623006820679,
|
| 475 |
+
"rewards/rejected": -0.7352713346481323,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.38474295190713104,
|
| 480 |
+
"grad_norm": 20.75,
|
| 481 |
+
"learning_rate": 3.865325090967081e-07,
|
| 482 |
+
"logits/chosen": -1.1717865467071533,
|
| 483 |
+
"logits/rejected": -1.1872893571853638,
|
| 484 |
+
"logps/chosen": -561.0062255859375,
|
| 485 |
+
"logps/rejected": -548.43603515625,
|
| 486 |
+
"loss": 0.4078,
|
| 487 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 488 |
+
"rewards/chosen": 0.11881165206432343,
|
| 489 |
+
"rewards/margins": 0.7872866988182068,
|
| 490 |
+
"rewards/rejected": -0.6684750318527222,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.39800995024875624,
|
| 495 |
+
"grad_norm": 23.25,
|
| 496 |
+
"learning_rate": 3.7667072722961357e-07,
|
| 497 |
+
"logits/chosen": -1.1743600368499756,
|
| 498 |
+
"logits/rejected": -1.1946338415145874,
|
| 499 |
+
"logps/chosen": -590.4188232421875,
|
| 500 |
+
"logps/rejected": -515.8514404296875,
|
| 501 |
+
"loss": 0.3934,
|
| 502 |
+
"rewards/accuracies": 0.925000011920929,
|
| 503 |
+
"rewards/chosen": 0.14561176300048828,
|
| 504 |
+
"rewards/margins": 0.7691020369529724,
|
| 505 |
+
"rewards/rejected": -0.6234902739524841,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.39800995024875624,
|
| 510 |
+
"eval_logits/chosen": -1.0754693746566772,
|
| 511 |
+
"eval_logits/rejected": -1.0528287887573242,
|
| 512 |
+
"eval_logps/chosen": -600.71435546875,
|
| 513 |
+
"eval_logps/rejected": -545.3746948242188,
|
| 514 |
+
"eval_loss": 0.39592820405960083,
|
| 515 |
+
"eval_rewards/accuracies": 0.9365671873092651,
|
| 516 |
+
"eval_rewards/chosen": 0.12633956968784332,
|
| 517 |
+
"eval_rewards/margins": 0.8475195169448853,
|
| 518 |
+
"eval_rewards/rejected": -0.7211799621582031,
|
| 519 |
+
"eval_runtime": 694.2423,
|
| 520 |
+
"eval_samples_per_second": 7.719,
|
| 521 |
+
"eval_steps_per_second": 0.483,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.41127694859038144,
|
| 526 |
+
"grad_norm": 22.375,
|
| 527 |
+
"learning_rate": 3.6653622300856457e-07,
|
| 528 |
+
"logits/chosen": -1.154953956604004,
|
| 529 |
+
"logits/rejected": -1.1766210794448853,
|
| 530 |
+
"logps/chosen": -573.0366821289062,
|
| 531 |
+
"logps/rejected": -538.422119140625,
|
| 532 |
+
"loss": 0.3901,
|
| 533 |
+
"rewards/accuracies": 0.925000011920929,
|
| 534 |
+
"rewards/chosen": 0.10485813766717911,
|
| 535 |
+
"rewards/margins": 0.8158906698226929,
|
| 536 |
+
"rewards/rejected": -0.711032509803772,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.42454394693200664,
|
| 541 |
+
"grad_norm": 22.0,
|
| 542 |
+
"learning_rate": 3.5615081604340903e-07,
|
| 543 |
+
"logits/chosen": -1.196800708770752,
|
| 544 |
+
"logits/rejected": -1.248241662979126,
|
| 545 |
+
"logps/chosen": -629.1947631835938,
|
| 546 |
+
"logps/rejected": -599.6900634765625,
|
| 547 |
+
"loss": 0.3898,
|
| 548 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 549 |
+
"rewards/chosen": 0.13255654275417328,
|
| 550 |
+
"rewards/margins": 0.8788650631904602,
|
| 551 |
+
"rewards/rejected": -0.7463085055351257,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.43781094527363185,
|
| 556 |
+
"grad_norm": 22.125,
|
| 557 |
+
"learning_rate": 3.455368661381543e-07,
|
| 558 |
+
"logits/chosen": -1.1678781509399414,
|
| 559 |
+
"logits/rejected": -1.178554892539978,
|
| 560 |
+
"logps/chosen": -493.3885803222656,
|
| 561 |
+
"logps/rejected": -474.68402099609375,
|
| 562 |
+
"loss": 0.3899,
|
| 563 |
+
"rewards/accuracies": 0.925000011920929,
|
| 564 |
+
"rewards/chosen": 0.07870842516422272,
|
| 565 |
+
"rewards/margins": 0.7927115559577942,
|
| 566 |
+
"rewards/rejected": -0.7140030860900879,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.45107794361525705,
|
| 571 |
+
"grad_norm": 22.625,
|
| 572 |
+
"learning_rate": 3.347172251502598e-07,
|
| 573 |
+
"logits/chosen": -1.1612517833709717,
|
| 574 |
+
"logits/rejected": -1.1822433471679688,
|
| 575 |
+
"logps/chosen": -608.53271484375,
|
| 576 |
+
"logps/rejected": -532.9712524414062,
|
| 577 |
+
"loss": 0.3763,
|
| 578 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 579 |
+
"rewards/chosen": 0.12583430111408234,
|
| 580 |
+
"rewards/margins": 0.9212196469306946,
|
| 581 |
+
"rewards/rejected": -0.7953853607177734,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.46434494195688225,
|
| 586 |
+
"grad_norm": 21.25,
|
| 587 |
+
"learning_rate": 3.2371518779053744e-07,
|
| 588 |
+
"logits/chosen": -1.101665735244751,
|
| 589 |
+
"logits/rejected": -1.0791598558425903,
|
| 590 |
+
"logps/chosen": -651.3994750976562,
|
| 591 |
+
"logps/rejected": -633.3742065429688,
|
| 592 |
+
"loss": 0.3678,
|
| 593 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 594 |
+
"rewards/chosen": 0.12759700417518616,
|
| 595 |
+
"rewards/margins": 0.9944127798080444,
|
| 596 |
+
"rewards/rejected": -0.8668158650398254,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.47761194029850745,
|
| 601 |
+
"grad_norm": 24.75,
|
| 602 |
+
"learning_rate": 3.1255444146958844e-07,
|
| 603 |
+
"logits/chosen": -1.1323697566986084,
|
| 604 |
+
"logits/rejected": -1.118276596069336,
|
| 605 |
+
"logps/chosen": -565.9261474609375,
|
| 606 |
+
"logps/rejected": -552.2655639648438,
|
| 607 |
+
"loss": 0.3684,
|
| 608 |
+
"rewards/accuracies": 0.90625,
|
| 609 |
+
"rewards/chosen": 0.06775705516338348,
|
| 610 |
+
"rewards/margins": 0.8904596567153931,
|
| 611 |
+
"rewards/rejected": -0.8227025270462036,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.49087893864013266,
|
| 616 |
+
"grad_norm": 20.875,
|
| 617 |
+
"learning_rate": 3.012590152987561e-07,
|
| 618 |
+
"logits/chosen": -1.1285905838012695,
|
| 619 |
+
"logits/rejected": -1.1064956188201904,
|
| 620 |
+
"logps/chosen": -601.677734375,
|
| 621 |
+
"logps/rejected": -547.2380981445312,
|
| 622 |
+
"loss": 0.3663,
|
| 623 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 624 |
+
"rewards/chosen": 0.12407927215099335,
|
| 625 |
+
"rewards/margins": 0.9496960639953613,
|
| 626 |
+
"rewards/rejected": -0.825616717338562,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.5041459369817579,
|
| 631 |
+
"grad_norm": 21.125,
|
| 632 |
+
"learning_rate": 2.8985322835539626e-07,
|
| 633 |
+
"logits/chosen": -1.0900777578353882,
|
| 634 |
+
"logits/rejected": -1.0679134130477905,
|
| 635 |
+
"logps/chosen": -622.616455078125,
|
| 636 |
+
"logps/rejected": -570.1490478515625,
|
| 637 |
+
"loss": 0.3705,
|
| 638 |
+
"rewards/accuracies": 0.956250011920929,
|
| 639 |
+
"rewards/chosen": 0.15115661919116974,
|
| 640 |
+
"rewards/margins": 0.9384227991104126,
|
| 641 |
+
"rewards/rejected": -0.7872661352157593,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.5174129353233831,
|
| 646 |
+
"grad_norm": 21.375,
|
| 647 |
+
"learning_rate": 2.7836163732385063e-07,
|
| 648 |
+
"logits/chosen": -1.146226167678833,
|
| 649 |
+
"logits/rejected": -1.131203293800354,
|
| 650 |
+
"logps/chosen": -637.7056274414062,
|
| 651 |
+
"logps/rejected": -580.0550537109375,
|
| 652 |
+
"loss": 0.3568,
|
| 653 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 654 |
+
"rewards/chosen": 0.09901280701160431,
|
| 655 |
+
"rewards/margins": 0.9571696519851685,
|
| 656 |
+
"rewards/rejected": -0.8581568598747253,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.5306799336650083,
|
| 661 |
+
"grad_norm": 21.125,
|
| 662 |
+
"learning_rate": 2.6680898362485124e-07,
|
| 663 |
+
"logits/chosen": -1.0712168216705322,
|
| 664 |
+
"logits/rejected": -1.0648881196975708,
|
| 665 |
+
"logps/chosen": -538.2913208007812,
|
| 666 |
+
"logps/rejected": -526.31689453125,
|
| 667 |
+
"loss": 0.3629,
|
| 668 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 669 |
+
"rewards/chosen": 0.10081170499324799,
|
| 670 |
+
"rewards/margins": 0.889560341835022,
|
| 671 |
+
"rewards/rejected": -0.788748562335968,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.5306799336650083,
|
| 676 |
+
"eval_logits/chosen": -1.1154277324676514,
|
| 677 |
+
"eval_logits/rejected": -1.110862135887146,
|
| 678 |
+
"eval_logps/chosen": -600.8080444335938,
|
| 679 |
+
"eval_logps/rejected": -546.7705078125,
|
| 680 |
+
"eval_loss": 0.3673515021800995,
|
| 681 |
+
"eval_rewards/accuracies": 0.9380596876144409,
|
| 682 |
+
"eval_rewards/chosen": 0.11697468906641006,
|
| 683 |
+
"eval_rewards/margins": 0.9777337312698364,
|
| 684 |
+
"eval_rewards/rejected": -0.8607590198516846,
|
| 685 |
+
"eval_runtime": 712.8834,
|
| 686 |
+
"eval_samples_per_second": 7.517,
|
| 687 |
+
"eval_steps_per_second": 0.47,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.5439469320066335,
|
| 692 |
+
"grad_norm": 19.625,
|
| 693 |
+
"learning_rate": 2.5522014014718697e-07,
|
| 694 |
+
"logits/chosen": -1.0688056945800781,
|
| 695 |
+
"logits/rejected": -1.0452687740325928,
|
| 696 |
+
"logps/chosen": -548.5653076171875,
|
| 697 |
+
"logps/rejected": -500.2899475097656,
|
| 698 |
+
"loss": 0.3686,
|
| 699 |
+
"rewards/accuracies": 0.9375,
|
| 700 |
+
"rewards/chosen": 0.09097462147474289,
|
| 701 |
+
"rewards/margins": 0.8869258761405945,
|
| 702 |
+
"rewards/rejected": -0.7959513664245605,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.5572139303482587,
|
| 707 |
+
"grad_norm": 20.75,
|
| 708 |
+
"learning_rate": 2.436200576963198e-07,
|
| 709 |
+
"logits/chosen": -1.1284773349761963,
|
| 710 |
+
"logits/rejected": -1.0750479698181152,
|
| 711 |
+
"logps/chosen": -570.462890625,
|
| 712 |
+
"logps/rejected": -484.501708984375,
|
| 713 |
+
"loss": 0.3638,
|
| 714 |
+
"rewards/accuracies": 0.956250011920929,
|
| 715 |
+
"rewards/chosen": 0.1417674571275711,
|
| 716 |
+
"rewards/margins": 0.9661204218864441,
|
| 717 |
+
"rewards/rejected": -0.8243529200553894,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.5704809286898839,
|
| 722 |
+
"grad_norm": 26.625,
|
| 723 |
+
"learning_rate": 2.3203371127524588e-07,
|
| 724 |
+
"logits/chosen": -1.142064094543457,
|
| 725 |
+
"logits/rejected": -1.0941470861434937,
|
| 726 |
+
"logps/chosen": -540.7872314453125,
|
| 727 |
+
"logps/rejected": -460.7454528808594,
|
| 728 |
+
"loss": 0.3751,
|
| 729 |
+
"rewards/accuracies": 0.90625,
|
| 730 |
+
"rewards/chosen": 0.09556931257247925,
|
| 731 |
+
"rewards/margins": 0.9223111271858215,
|
| 732 |
+
"rewards/rejected": -0.8267418742179871,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.5837479270315091,
|
| 737 |
+
"grad_norm": 21.75,
|
| 738 |
+
"learning_rate": 2.2048604631325892e-07,
|
| 739 |
+
"logits/chosen": -1.0380522012710571,
|
| 740 |
+
"logits/rejected": -1.036592721939087,
|
| 741 |
+
"logps/chosen": -568.5135498046875,
|
| 742 |
+
"logps/rejected": -558.4591064453125,
|
| 743 |
+
"loss": 0.3722,
|
| 744 |
+
"rewards/accuracies": 0.9375,
|
| 745 |
+
"rewards/chosen": 0.05709138512611389,
|
| 746 |
+
"rewards/margins": 0.89503014087677,
|
| 747 |
+
"rewards/rejected": -0.837938666343689,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.5970149253731343,
|
| 752 |
+
"grad_norm": 20.5,
|
| 753 |
+
"learning_rate": 2.0900192495838615e-07,
|
| 754 |
+
"logits/chosen": -1.0975573062896729,
|
| 755 |
+
"logits/rejected": -1.0408273935317993,
|
| 756 |
+
"logps/chosen": -531.0595703125,
|
| 757 |
+
"logps/rejected": -478.84222412109375,
|
| 758 |
+
"loss": 0.3591,
|
| 759 |
+
"rewards/accuracies": 0.925000011920929,
|
| 760 |
+
"rewards/chosen": 0.07921108603477478,
|
| 761 |
+
"rewards/margins": 0.9134254455566406,
|
| 762 |
+
"rewards/rejected": -0.8342143893241882,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.6102819237147595,
|
| 767 |
+
"grad_norm": 18.5,
|
| 768 |
+
"learning_rate": 1.9760607254912926e-07,
|
| 769 |
+
"logits/chosen": -1.0876163244247437,
|
| 770 |
+
"logits/rejected": -1.0495961904525757,
|
| 771 |
+
"logps/chosen": -616.9581298828125,
|
| 772 |
+
"logps/rejected": -538.6895751953125,
|
| 773 |
+
"loss": 0.3562,
|
| 774 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 775 |
+
"rewards/chosen": 0.09119832515716553,
|
| 776 |
+
"rewards/margins": 0.9315482974052429,
|
| 777 |
+
"rewards/rejected": -0.8403499722480774,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.6235489220563848,
|
| 782 |
+
"grad_norm": 20.75,
|
| 783 |
+
"learning_rate": 1.8632302438075613e-07,
|
| 784 |
+
"logits/chosen": -1.1088566780090332,
|
| 785 |
+
"logits/rejected": -1.1191766262054443,
|
| 786 |
+
"logps/chosen": -589.3919067382812,
|
| 787 |
+
"logps/rejected": -567.9808349609375,
|
| 788 |
+
"loss": 0.3629,
|
| 789 |
+
"rewards/accuracies": 0.956250011920929,
|
| 790 |
+
"rewards/chosen": 0.1344141662120819,
|
| 791 |
+
"rewards/margins": 1.0434377193450928,
|
| 792 |
+
"rewards/rejected": -0.9090234637260437,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 0.6368159203980099,
|
| 797 |
+
"grad_norm": 18.625,
|
| 798 |
+
"learning_rate": 1.7517707288075614e-07,
|
| 799 |
+
"logits/chosen": -1.106209635734558,
|
| 800 |
+
"logits/rejected": -1.1109434366226196,
|
| 801 |
+
"logps/chosen": -565.6036987304688,
|
| 802 |
+
"logps/rejected": -519.3636474609375,
|
| 803 |
+
"loss": 0.3464,
|
| 804 |
+
"rewards/accuracies": 0.9375,
|
| 805 |
+
"rewards/chosen": 0.09805373102426529,
|
| 806 |
+
"rewards/margins": 0.9947841763496399,
|
| 807 |
+
"rewards/rejected": -0.8967304229736328,
|
| 808 |
+
"step": 480
|
| 809 |
+
},
|
| 810 |
+
{
|
| 811 |
+
"epoch": 0.6500829187396352,
|
| 812 |
+
"grad_norm": 20.5,
|
| 813 |
+
"learning_rate": 1.641922153071906e-07,
|
| 814 |
+
"logits/chosen": -1.0548484325408936,
|
| 815 |
+
"logits/rejected": -1.0250844955444336,
|
| 816 |
+
"logps/chosen": -573.2247924804688,
|
| 817 |
+
"logps/rejected": -566.44970703125,
|
| 818 |
+
"loss": 0.365,
|
| 819 |
+
"rewards/accuracies": 0.956250011920929,
|
| 820 |
+
"rewards/chosen": 0.09521574527025223,
|
| 821 |
+
"rewards/margins": 1.005110740661621,
|
| 822 |
+
"rewards/rejected": -0.9098949432373047,
|
| 823 |
+
"step": 490
|
| 824 |
+
},
|
| 825 |
+
{
|
| 826 |
+
"epoch": 0.6633499170812603,
|
| 827 |
+
"grad_norm": 19.125,
|
| 828 |
+
"learning_rate": 1.5339210208254344e-07,
|
| 829 |
+
"logits/chosen": -1.060248613357544,
|
| 830 |
+
"logits/rejected": -1.0458314418792725,
|
| 831 |
+
"logps/chosen": -541.8770751953125,
|
| 832 |
+
"logps/rejected": -513.5958251953125,
|
| 833 |
+
"loss": 0.3556,
|
| 834 |
+
"rewards/accuracies": 0.925000011920929,
|
| 835 |
+
"rewards/chosen": 0.11755643784999847,
|
| 836 |
+
"rewards/margins": 0.9871570467948914,
|
| 837 |
+
"rewards/rejected": -0.8696004748344421,
|
| 838 |
+
"step": 500
|
| 839 |
+
},
|
| 840 |
+
{
|
| 841 |
+
"epoch": 0.6633499170812603,
|
| 842 |
+
"eval_logits/chosen": -0.9289145469665527,
|
| 843 |
+
"eval_logits/rejected": -0.8265557885169983,
|
| 844 |
+
"eval_logps/chosen": -600.8418579101562,
|
| 845 |
+
"eval_logps/rejected": -547.3089599609375,
|
| 846 |
+
"eval_loss": 0.3561394512653351,
|
| 847 |
+
"eval_rewards/accuracies": 0.9388059973716736,
|
| 848 |
+
"eval_rewards/chosen": 0.11359576135873795,
|
| 849 |
+
"eval_rewards/margins": 1.0282028913497925,
|
| 850 |
+
"eval_rewards/rejected": -0.9146071672439575,
|
| 851 |
+
"eval_runtime": 731.2224,
|
| 852 |
+
"eval_samples_per_second": 7.329,
|
| 853 |
+
"eval_steps_per_second": 0.458,
|
| 854 |
+
"step": 500
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"epoch": 0.6766169154228856,
|
| 858 |
+
"grad_norm": 19.75,
|
| 859 |
+
"learning_rate": 1.4279998587430943e-07,
|
| 860 |
+
"logits/chosen": -1.0720138549804688,
|
| 861 |
+
"logits/rejected": -1.0440585613250732,
|
| 862 |
+
"logps/chosen": -576.8155517578125,
|
| 863 |
+
"logps/rejected": -474.7138671875,
|
| 864 |
+
"loss": 0.3397,
|
| 865 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 866 |
+
"rewards/chosen": 0.12137231975793839,
|
| 867 |
+
"rewards/margins": 1.0447041988372803,
|
| 868 |
+
"rewards/rejected": -0.9233318567276001,
|
| 869 |
+
"step": 510
|
| 870 |
+
},
|
| 871 |
+
{
|
| 872 |
+
"epoch": 0.6898839137645107,
|
| 873 |
+
"grad_norm": 19.375,
|
| 874 |
+
"learning_rate": 1.324386715319503e-07,
|
| 875 |
+
"logits/chosen": -1.0745595693588257,
|
| 876 |
+
"logits/rejected": -1.0517549514770508,
|
| 877 |
+
"logps/chosen": -547.5264282226562,
|
| 878 |
+
"logps/rejected": -513.98974609375,
|
| 879 |
+
"loss": 0.3567,
|
| 880 |
+
"rewards/accuracies": 0.949999988079071,
|
| 881 |
+
"rewards/chosen": 0.12890011072158813,
|
| 882 |
+
"rewards/margins": 0.9725991487503052,
|
| 883 |
+
"rewards/rejected": -0.8436989784240723,
|
| 884 |
+
"step": 520
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"epoch": 0.703150912106136,
|
| 888 |
+
"grad_norm": 20.125,
|
| 889 |
+
"learning_rate": 1.2233046698800343e-07,
|
| 890 |
+
"logits/chosen": -1.0820659399032593,
|
| 891 |
+
"logits/rejected": -1.0528825521469116,
|
| 892 |
+
"logps/chosen": -623.875,
|
| 893 |
+
"logps/rejected": -623.9129638671875,
|
| 894 |
+
"loss": 0.3541,
|
| 895 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 896 |
+
"rewards/chosen": 0.141106516122818,
|
| 897 |
+
"rewards/margins": 1.074299931526184,
|
| 898 |
+
"rewards/rejected": -0.9331933856010437,
|
| 899 |
+
"step": 530
|
| 900 |
+
},
|
| 901 |
+
{
|
| 902 |
+
"epoch": 0.7164179104477612,
|
| 903 |
+
"grad_norm": 20.625,
|
| 904 |
+
"learning_rate": 1.124971352290545e-07,
|
| 905 |
+
"logits/chosen": -1.108722448348999,
|
| 906 |
+
"logits/rejected": -1.0873199701309204,
|
| 907 |
+
"logps/chosen": -593.7824096679688,
|
| 908 |
+
"logps/rejected": -559.232421875,
|
| 909 |
+
"loss": 0.337,
|
| 910 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 911 |
+
"rewards/chosen": 0.1328240931034088,
|
| 912 |
+
"rewards/margins": 1.019162654876709,
|
| 913 |
+
"rewards/rejected": -0.8863385915756226,
|
| 914 |
+
"step": 540
|
| 915 |
+
},
|
| 916 |
+
{
|
| 917 |
+
"epoch": 0.7296849087893864,
|
| 918 |
+
"grad_norm": 21.75,
|
| 919 |
+
"learning_rate": 1.0295984743997909e-07,
|
| 920 |
+
"logits/chosen": -1.085311770439148,
|
| 921 |
+
"logits/rejected": -1.0750799179077148,
|
| 922 |
+
"logps/chosen": -602.3040161132812,
|
| 923 |
+
"logps/rejected": -558.0685424804688,
|
| 924 |
+
"loss": 0.3544,
|
| 925 |
+
"rewards/accuracies": 0.9375,
|
| 926 |
+
"rewards/chosen": 0.08248591423034668,
|
| 927 |
+
"rewards/margins": 1.015809178352356,
|
| 928 |
+
"rewards/rejected": -0.9333232641220093,
|
| 929 |
+
"step": 550
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 0.7429519071310116,
|
| 933 |
+
"grad_norm": 20.375,
|
| 934 |
+
"learning_rate": 9.37391374223355e-08,
|
| 935 |
+
"logits/chosen": -1.1596343517303467,
|
| 936 |
+
"logits/rejected": -1.1693814992904663,
|
| 937 |
+
"logps/chosen": -583.8175659179688,
|
| 938 |
+
"logps/rejected": -574.1356811523438,
|
| 939 |
+
"loss": 0.3416,
|
| 940 |
+
"rewards/accuracies": 0.9375,
|
| 941 |
+
"rewards/chosen": 0.1034403070807457,
|
| 942 |
+
"rewards/margins": 1.0356991291046143,
|
| 943 |
+
"rewards/rejected": -0.9322587251663208,
|
| 944 |
+
"step": 560
|
| 945 |
+
},
|
| 946 |
+
{
|
| 947 |
+
"epoch": 0.7562189054726368,
|
| 948 |
+
"grad_norm": 17.875,
|
| 949 |
+
"learning_rate": 8.485485738504488e-08,
|
| 950 |
+
"logits/chosen": -1.1387842893600464,
|
| 951 |
+
"logits/rejected": -1.108246922492981,
|
| 952 |
+
"logps/chosen": -627.6776123046875,
|
| 953 |
+
"logps/rejected": -469.2587890625,
|
| 954 |
+
"loss": 0.3468,
|
| 955 |
+
"rewards/accuracies": 0.956250011920929,
|
| 956 |
+
"rewards/chosen": 0.1544940024614334,
|
| 957 |
+
"rewards/margins": 1.0551023483276367,
|
| 958 |
+
"rewards/rejected": -0.900608241558075,
|
| 959 |
+
"step": 570
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 0.7694859038142621,
|
| 963 |
+
"grad_norm": 19.25,
|
| 964 |
+
"learning_rate": 7.632613520254158e-08,
|
| 965 |
+
"logits/chosen": -1.0649652481079102,
|
| 966 |
+
"logits/rejected": -1.064888834953308,
|
| 967 |
+
"logps/chosen": -601.84033203125,
|
| 968 |
+
"logps/rejected": -529.4444580078125,
|
| 969 |
+
"loss": 0.359,
|
| 970 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 971 |
+
"rewards/chosen": 0.10679004341363907,
|
| 972 |
+
"rewards/margins": 1.0746941566467285,
|
| 973 |
+
"rewards/rejected": -0.9679039716720581,
|
| 974 |
+
"step": 580
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 0.7827529021558872,
|
| 978 |
+
"grad_norm": 20.75,
|
| 979 |
+
"learning_rate": 6.817133323241755e-08,
|
| 980 |
+
"logits/chosen": -1.1302725076675415,
|
| 981 |
+
"logits/rejected": -1.1106232404708862,
|
| 982 |
+
"logps/chosen": -675.71533203125,
|
| 983 |
+
"logps/rejected": -525.1083984375,
|
| 984 |
+
"loss": 0.3478,
|
| 985 |
+
"rewards/accuracies": 0.918749988079071,
|
| 986 |
+
"rewards/chosen": 0.09434429556131363,
|
| 987 |
+
"rewards/margins": 1.0678186416625977,
|
| 988 |
+
"rewards/rejected": -0.9734743237495422,
|
| 989 |
+
"step": 590
|
| 990 |
+
},
|
| 991 |
+
{
|
| 992 |
+
"epoch": 0.7960199004975125,
|
| 993 |
+
"grad_norm": 22.25,
|
| 994 |
+
"learning_rate": 6.040800878122654e-08,
|
| 995 |
+
"logits/chosen": -1.1192970275878906,
|
| 996 |
+
"logits/rejected": -1.1343142986297607,
|
| 997 |
+
"logps/chosen": -567.078125,
|
| 998 |
+
"logps/rejected": -525.439697265625,
|
| 999 |
+
"loss": 0.3488,
|
| 1000 |
+
"rewards/accuracies": 0.9375,
|
| 1001 |
+
"rewards/chosen": 0.09638460725545883,
|
| 1002 |
+
"rewards/margins": 0.9892334938049316,
|
| 1003 |
+
"rewards/rejected": -0.8928488492965698,
|
| 1004 |
+
"step": 600
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"epoch": 0.7960199004975125,
|
| 1008 |
+
"eval_logits/chosen": -1.0876879692077637,
|
| 1009 |
+
"eval_logits/rejected": -1.067589521408081,
|
| 1010 |
+
"eval_logps/chosen": -600.8737182617188,
|
| 1011 |
+
"eval_logps/rejected": -547.4734497070312,
|
| 1012 |
+
"eval_loss": 0.35399559140205383,
|
| 1013 |
+
"eval_rewards/accuracies": 0.9410447478294373,
|
| 1014 |
+
"eval_rewards/chosen": 0.11041063815355301,
|
| 1015 |
+
"eval_rewards/margins": 1.0414601564407349,
|
| 1016 |
+
"eval_rewards/rejected": -0.9310495853424072,
|
| 1017 |
+
"eval_runtime": 680.93,
|
| 1018 |
+
"eval_samples_per_second": 7.87,
|
| 1019 |
+
"eval_steps_per_second": 0.492,
|
| 1020 |
+
"step": 600
|
| 1021 |
+
},
|
| 1022 |
+
{
|
| 1023 |
+
"epoch": 0.8092868988391376,
|
| 1024 |
+
"grad_norm": 20.625,
|
| 1025 |
+
"learning_rate": 5.305287630356362e-08,
|
| 1026 |
+
"logits/chosen": -1.1514161825180054,
|
| 1027 |
+
"logits/rejected": -1.1489306688308716,
|
| 1028 |
+
"logps/chosen": -603.4745483398438,
|
| 1029 |
+
"logps/rejected": -562.2152709960938,
|
| 1030 |
+
"loss": 0.3556,
|
| 1031 |
+
"rewards/accuracies": 0.9375,
|
| 1032 |
+
"rewards/chosen": 0.08816297352313995,
|
| 1033 |
+
"rewards/margins": 1.0151373147964478,
|
| 1034 |
+
"rewards/rejected": -0.9269744157791138,
|
| 1035 |
+
"step": 610
|
| 1036 |
+
},
|
| 1037 |
+
{
|
| 1038 |
+
"epoch": 0.8225538971807629,
|
| 1039 |
+
"grad_norm": 23.375,
|
| 1040 |
+
"learning_rate": 4.612177141580875e-08,
|
| 1041 |
+
"logits/chosen": -1.0614503622055054,
|
| 1042 |
+
"logits/rejected": -1.0462639331817627,
|
| 1043 |
+
"logps/chosen": -600.1735229492188,
|
| 1044 |
+
"logps/rejected": -563.9249877929688,
|
| 1045 |
+
"loss": 0.3555,
|
| 1046 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1047 |
+
"rewards/chosen": 0.1291816681623459,
|
| 1048 |
+
"rewards/margins": 1.0133174657821655,
|
| 1049 |
+
"rewards/rejected": -0.8841358423233032,
|
| 1050 |
+
"step": 620
|
| 1051 |
+
},
|
| 1052 |
+
{
|
| 1053 |
+
"epoch": 0.835820895522388,
|
| 1054 |
+
"grad_norm": 19.25,
|
| 1055 |
+
"learning_rate": 3.962961680200927e-08,
|
| 1056 |
+
"logits/chosen": -1.154007911682129,
|
| 1057 |
+
"logits/rejected": -1.1668691635131836,
|
| 1058 |
+
"logps/chosen": -587.5554809570312,
|
| 1059 |
+
"logps/rejected": -579.425537109375,
|
| 1060 |
+
"loss": 0.3591,
|
| 1061 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 1062 |
+
"rewards/chosen": 0.1071285754442215,
|
| 1063 |
+
"rewards/margins": 1.01383376121521,
|
| 1064 |
+
"rewards/rejected": -0.9067050814628601,
|
| 1065 |
+
"step": 630
|
| 1066 |
+
},
|
| 1067 |
+
{
|
| 1068 |
+
"epoch": 0.8490878938640133,
|
| 1069 |
+
"grad_norm": 19.625,
|
| 1070 |
+
"learning_rate": 3.359039008530845e-08,
|
| 1071 |
+
"logits/chosen": -1.1280542612075806,
|
| 1072 |
+
"logits/rejected": -1.1074917316436768,
|
| 1073 |
+
"logps/chosen": -638.248779296875,
|
| 1074 |
+
"logps/rejected": -570.1997680664062,
|
| 1075 |
+
"loss": 0.3504,
|
| 1076 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1077 |
+
"rewards/chosen": 0.10942580550909042,
|
| 1078 |
+
"rewards/margins": 1.0872418880462646,
|
| 1079 |
+
"rewards/rejected": -0.9778162240982056,
|
| 1080 |
+
"step": 640
|
| 1081 |
+
},
|
| 1082 |
+
{
|
| 1083 |
+
"epoch": 0.8623548922056384,
|
| 1084 |
+
"grad_norm": 21.125,
|
| 1085 |
+
"learning_rate": 2.8017093734092474e-08,
|
| 1086 |
+
"logits/chosen": -1.0559157133102417,
|
| 1087 |
+
"logits/rejected": -0.9880287051200867,
|
| 1088 |
+
"logps/chosen": -617.8060302734375,
|
| 1089 |
+
"logps/rejected": -542.3763427734375,
|
| 1090 |
+
"loss": 0.3563,
|
| 1091 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1092 |
+
"rewards/chosen": 0.06925593316555023,
|
| 1093 |
+
"rewards/margins": 1.0108643770217896,
|
| 1094 |
+
"rewards/rejected": -0.9416083097457886,
|
| 1095 |
+
"step": 650
|
| 1096 |
+
},
|
| 1097 |
+
{
|
| 1098 |
+
"epoch": 0.8756218905472637,
|
| 1099 |
+
"grad_norm": 20.25,
|
| 1100 |
+
"learning_rate": 2.292172706764703e-08,
|
| 1101 |
+
"logits/chosen": -1.0475237369537354,
|
| 1102 |
+
"logits/rejected": -1.0124943256378174,
|
| 1103 |
+
"logps/chosen": -625.7453002929688,
|
| 1104 |
+
"logps/rejected": -627.2447509765625,
|
| 1105 |
+
"loss": 0.3567,
|
| 1106 |
+
"rewards/accuracies": 0.9375,
|
| 1107 |
+
"rewards/chosen": 0.11004464328289032,
|
| 1108 |
+
"rewards/margins": 1.0592918395996094,
|
| 1109 |
+
"rewards/rejected": -0.9492471814155579,
|
| 1110 |
+
"step": 660
|
| 1111 |
+
},
|
| 1112 |
+
{
|
| 1113 |
+
"epoch": 0.8888888888888888,
|
| 1114 |
+
"grad_norm": 19.875,
|
| 1115 |
+
"learning_rate": 1.8315260421596924e-08,
|
| 1116 |
+
"logits/chosen": -1.16936457157135,
|
| 1117 |
+
"logits/rejected": -1.1426036357879639,
|
| 1118 |
+
"logps/chosen": -555.6038818359375,
|
| 1119 |
+
"logps/rejected": -494.2486267089844,
|
| 1120 |
+
"loss": 0.3509,
|
| 1121 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 1122 |
+
"rewards/chosen": 0.12268821895122528,
|
| 1123 |
+
"rewards/margins": 0.9934048652648926,
|
| 1124 |
+
"rewards/rejected": -0.8707167506217957,
|
| 1125 |
+
"step": 670
|
| 1126 |
+
},
|
| 1127 |
+
{
|
| 1128 |
+
"epoch": 0.9021558872305141,
|
| 1129 |
+
"grad_norm": 20.125,
|
| 1130 |
+
"learning_rate": 1.4207611528748997e-08,
|
| 1131 |
+
"logits/chosen": -1.122236967086792,
|
| 1132 |
+
"logits/rejected": -1.0928575992584229,
|
| 1133 |
+
"logps/chosen": -568.34765625,
|
| 1134 |
+
"logps/rejected": -548.4368896484375,
|
| 1135 |
+
"loss": 0.3548,
|
| 1136 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 1137 |
+
"rewards/chosen": 0.10156118869781494,
|
| 1138 |
+
"rewards/margins": 1.01285719871521,
|
| 1139 |
+
"rewards/rejected": -0.9112960696220398,
|
| 1140 |
+
"step": 680
|
| 1141 |
+
},
|
| 1142 |
+
{
|
| 1143 |
+
"epoch": 0.9154228855721394,
|
| 1144 |
+
"grad_norm": 20.875,
|
| 1145 |
+
"learning_rate": 1.0607624166191958e-08,
|
| 1146 |
+
"logits/chosen": -1.102480411529541,
|
| 1147 |
+
"logits/rejected": -1.097570776939392,
|
| 1148 |
+
"logps/chosen": -670.6092529296875,
|
| 1149 |
+
"logps/rejected": -724.0338134765625,
|
| 1150 |
+
"loss": 0.3508,
|
| 1151 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 1152 |
+
"rewards/chosen": 0.122245192527771,
|
| 1153 |
+
"rewards/margins": 1.0727375745773315,
|
| 1154 |
+
"rewards/rejected": -0.9504923820495605,
|
| 1155 |
+
"step": 690
|
| 1156 |
+
},
|
| 1157 |
+
{
|
| 1158 |
+
"epoch": 0.9286898839137645,
|
| 1159 |
+
"grad_norm": 20.75,
|
| 1160 |
+
"learning_rate": 7.523049114624647e-09,
|
| 1161 |
+
"logits/chosen": -1.067058801651001,
|
| 1162 |
+
"logits/rejected": -1.0042006969451904,
|
| 1163 |
+
"logps/chosen": -610.7342529296875,
|
| 1164 |
+
"logps/rejected": -569.3170776367188,
|
| 1165 |
+
"loss": 0.3563,
|
| 1166 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1167 |
+
"rewards/chosen": 0.13328425586223602,
|
| 1168 |
+
"rewards/margins": 1.0865224599838257,
|
| 1169 |
+
"rewards/rejected": -0.9532381296157837,
|
| 1170 |
+
"step": 700
|
| 1171 |
+
},
|
| 1172 |
+
{
|
| 1173 |
+
"epoch": 0.9286898839137645,
|
| 1174 |
+
"eval_logits/chosen": -0.9600119590759277,
|
| 1175 |
+
"eval_logits/rejected": -0.8735809922218323,
|
| 1176 |
+
"eval_logps/chosen": -600.8121337890625,
|
| 1177 |
+
"eval_logps/rejected": -547.42236328125,
|
| 1178 |
+
"eval_loss": 0.3540438711643219,
|
| 1179 |
+
"eval_rewards/accuracies": 0.9395522475242615,
|
| 1180 |
+
"eval_rewards/chosen": 0.11656844615936279,
|
| 1181 |
+
"eval_rewards/margins": 1.0425076484680176,
|
| 1182 |
+
"eval_rewards/rejected": -0.92593914270401,
|
| 1183 |
+
"eval_runtime": 733.9957,
|
| 1184 |
+
"eval_samples_per_second": 7.301,
|
| 1185 |
+
"eval_steps_per_second": 0.456,
|
| 1186 |
+
"step": 700
|
| 1187 |
+
},
|
| 1188 |
+
{
|
| 1189 |
+
"epoch": 0.9419568822553898,
|
| 1190 |
+
"grad_norm": 20.25,
|
| 1191 |
+
"learning_rate": 4.960527470908277e-09,
|
| 1192 |
+
"logits/chosen": -0.9644180536270142,
|
| 1193 |
+
"logits/rejected": -0.860200047492981,
|
| 1194 |
+
"logps/chosen": -622.1219482421875,
|
| 1195 |
+
"logps/rejected": -567.1380615234375,
|
| 1196 |
+
"loss": 0.3555,
|
| 1197 |
+
"rewards/accuracies": 0.956250011920929,
|
| 1198 |
+
"rewards/chosen": 0.11110372841358185,
|
| 1199 |
+
"rewards/margins": 1.0618221759796143,
|
| 1200 |
+
"rewards/rejected": -0.9507185220718384,
|
| 1201 |
+
"step": 710
|
| 1202 |
+
},
|
| 1203 |
+
{
|
| 1204 |
+
"epoch": 0.9552238805970149,
|
| 1205 |
+
"grad_norm": 19.75,
|
| 1206 |
+
"learning_rate": 2.925576349770337e-09,
|
| 1207 |
+
"logits/chosen": -0.9986553192138672,
|
| 1208 |
+
"logits/rejected": -0.8984715342521667,
|
| 1209 |
+
"logps/chosen": -605.7318725585938,
|
| 1210 |
+
"logps/rejected": -542.7632446289062,
|
| 1211 |
+
"loss": 0.359,
|
| 1212 |
+
"rewards/accuracies": 0.956250011920929,
|
| 1213 |
+
"rewards/chosen": 0.10133838653564453,
|
| 1214 |
+
"rewards/margins": 0.9883183240890503,
|
| 1215 |
+
"rewards/rejected": -0.886979877948761,
|
| 1216 |
+
"step": 720
|
| 1217 |
+
},
|
| 1218 |
+
{
|
| 1219 |
+
"epoch": 0.9684908789386402,
|
| 1220 |
+
"grad_norm": 21.25,
|
| 1221 |
+
"learning_rate": 1.4225770054443197e-09,
|
| 1222 |
+
"logits/chosen": -0.9282068014144897,
|
| 1223 |
+
"logits/rejected": -0.8550642132759094,
|
| 1224 |
+
"logps/chosen": -571.7738037109375,
|
| 1225 |
+
"logps/rejected": -500.0634765625,
|
| 1226 |
+
"loss": 0.3571,
|
| 1227 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1228 |
+
"rewards/chosen": 0.10189428180456161,
|
| 1229 |
+
"rewards/margins": 1.0126664638519287,
|
| 1230 |
+
"rewards/rejected": -0.9107722043991089,
|
| 1231 |
+
"step": 730
|
| 1232 |
+
},
|
| 1233 |
+
{
|
| 1234 |
+
"epoch": 0.9817578772802653,
|
| 1235 |
+
"grad_norm": 19.625,
|
| 1236 |
+
"learning_rate": 4.547653988198619e-10,
|
| 1237 |
+
"logits/chosen": -0.9236332774162292,
|
| 1238 |
+
"logits/rejected": -0.8542205095291138,
|
| 1239 |
+
"logps/chosen": -632.85546875,
|
| 1240 |
+
"logps/rejected": -597.6421508789062,
|
| 1241 |
+
"loss": 0.3493,
|
| 1242 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1243 |
+
"rewards/chosen": 0.147782564163208,
|
| 1244 |
+
"rewards/margins": 1.067484736442566,
|
| 1245 |
+
"rewards/rejected": -0.9197022318840027,
|
| 1246 |
+
"step": 740
|
| 1247 |
+
},
|
| 1248 |
+
{
|
| 1249 |
+
"epoch": 0.9950248756218906,
|
| 1250 |
+
"grad_norm": 20.5,
|
| 1251 |
+
"learning_rate": 2.4225230411789588e-11,
|
| 1252 |
+
"logits/chosen": -0.9963258504867554,
|
| 1253 |
+
"logits/rejected": -0.8823927044868469,
|
| 1254 |
+
"logps/chosen": -617.5396728515625,
|
| 1255 |
+
"logps/rejected": -596.9856567382812,
|
| 1256 |
+
"loss": 0.36,
|
| 1257 |
+
"rewards/accuracies": 0.90625,
|
| 1258 |
+
"rewards/chosen": 0.14549708366394043,
|
| 1259 |
+
"rewards/margins": 1.0541255474090576,
|
| 1260 |
+
"rewards/rejected": -0.9086285829544067,
|
| 1261 |
+
"step": 750
|
| 1262 |
+
},
|
| 1263 |
+
{
|
| 1264 |
+
"epoch": 0.9990049751243781,
|
| 1265 |
+
"step": 753,
|
| 1266 |
+
"total_flos": 0.0,
|
| 1267 |
+
"train_loss": 0.4291752041731856,
|
| 1268 |
+
"train_runtime": 22849.6211,
|
| 1269 |
+
"train_samples_per_second": 2.111,
|
| 1270 |
+
"train_steps_per_second": 0.033
|
| 1271 |
+
}
|
| 1272 |
+
],
|
| 1273 |
+
"logging_steps": 10,
|
| 1274 |
+
"max_steps": 753,
|
| 1275 |
+
"num_input_tokens_seen": 0,
|
| 1276 |
+
"num_train_epochs": 1,
|
| 1277 |
+
"save_steps": 100,
|
| 1278 |
+
"total_flos": 0.0,
|
| 1279 |
+
"train_batch_size": 4,
|
| 1280 |
+
"trial_name": null,
|
| 1281 |
+
"trial_params": null
|
| 1282 |
+
}
|