Hugging Face's logo

hugodk-sch
/
aftonposten-6b-align-scan

PEFT
Safetensors
gptj
alignment-handbook
trl
dpo
Generated from Trainer
4-bit precision
bitsandbytes
Model card Files Community
hugodk-sch commited on
Commit
3357b7f
•
1 Parent(s): 3daf7c5

Training in progress, step 400

Browse files
Files changed (8) hide show
  1. README.md +15 -27
  2. adapter_config.json +5 -5
  3. adapter_model.safetensors +1 -1
  4. all_results.json +17 -17
  5. eval_results.json +13 -13
  6. train_results.json +5 -5
  7. trainer_state.json +0 -0
  8. training_args.bin +1 -1
README.md CHANGED
@@ -20,15 +20,15 @@ should probably proofread and complete it, then remove this comment. -->
20
 
21
  This model is a fine-tuned version of [data/ap-gpt-j-6b-sft-qlora-04-08](https://huggingface.co/data/ap-gpt-j-6b-sft-qlora-04-08) on the hugodk-sch/aftonposten_title_prefs dataset.
22
  It achieves the following results on the evaluation set:
23
- - Loss: 0.6893
24
- - Rewards/chosen: -0.0505
25
- - Rewards/rejected: -0.0590
26
- - Rewards/accuracies: 0.5365
27
- - Rewards/margins: 0.0085
28
- - Logps/rejected: -43.4139
29
- - Logps/chosen: -39.0820
30
- - Logits/rejected: -1.6163
31
- - Logits/chosen: -1.6202
32
 
33
  ## Model description
34
 
@@ -57,27 +57,15 @@ The following hyperparameters were used during training:
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: 4
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
- | 0.6924 | 0.26 | 100 | -2.2301 | -2.2252 | -34.0468 | -37.5393 | 0.6931 | 0.5187 | -0.0001 | 0.0001 | -0.0002 |
67
- | 0.6914 | 0.52 | 200 | -2.2250 | -2.2201 | -34.0426 | -37.5530 | 0.6930 | 0.5245 | -0.0001 | 0.0003 | -0.0004 |
68
- | 0.6895 | 0.78 | 300 | -2.2211 | -2.2163 | -34.0648 | -37.5868 | 0.6929 | 0.5419 | -0.0003 | 0.0004 | -0.0007 |
69
- | 0.6863 | 1.04 | 400 | 0.6928 | -0.0011 | -0.0017 | 0.5681 | 0.0006 | -37.6890 | -34.1431 | -2.1889 | -2.1937 |
70
- | 0.6826 | 1.3 | 500 | 0.6925 | -0.0031 | -0.0043 | 0.5652 | 0.0012 | -37.9472 | -34.3421 | -2.1173 | -2.1221 |
71
- | 0.6823 | 1.56 | 600 | 0.6918 | -0.0070 | -0.0097 | 0.5831 | 0.0028 | -38.4905 | -34.7312 | -2.0241 | -2.0287 |
72
- | 0.6784 | 1.82 | 700 | 0.6916 | -0.0132 | -0.0163 | 0.5656 | 0.0031 | -39.1509 | -35.3551 | -1.9419 | -1.9464 |
73
- | 0.6729 | 2.08 | 800 | 0.6911 | -0.0191 | -0.0235 | 0.5540 | 0.0044 | -39.8622 | -35.9412 | -1.8648 | -1.8692 |
74
- | 0.6635 | 2.34 | 900 | 0.6904 | -0.0309 | -0.0368 | 0.5365 | 0.0059 | -41.1917 | -37.1214 | -1.7445 | -1.7488 |
75
- | 0.6719 | 2.6 | 1000 | 0.6897 | -0.0445 | -0.0522 | 0.5486 | 0.0077 | -42.7361 | -38.4865 | -1.6560 | -1.6600 |
76
- | 0.6749 | 2.86 | 1100 | 0.6894 | -0.0493 | -0.0575 | 0.5365 | 0.0082 | -43.2672 | -38.9607 | -1.6246 | -1.6286 |
77
- | 0.6551 | 3.12 | 1200 | 0.6893 | -0.0498 | -0.0582 | 0.5365 | 0.0084 | -43.3379 | -39.0115 | -1.6198 | -1.6238 |
78
- | 0.6489 | 3.38 | 1300 | 0.6893 | -0.0504 | -0.0589 | 0.5395 | 0.0086 | -43.4114 | -39.0722 | -1.6172 | -1.6212 |
79
- | 0.6597 | 3.64 | 1400 | 0.6893 | -0.0505 | -0.0591 | 0.5424 | 0.0086 | -43.4228 | -39.0843 | -1.6165 | -1.6205 |
80
- | 0.6557 | 3.9 | 1500 | 0.6893 | -0.0505 | -0.0590 | 0.5424 | 0.0085 | -43.4186 | -39.0870 | -1.6166 | -1.6206 |
81
 
82
 
83
  ### Framework versions
 
20
 
21
  This model is a fine-tuned version of [data/ap-gpt-j-6b-sft-qlora-04-08](https://huggingface.co/data/ap-gpt-j-6b-sft-qlora-04-08) on the hugodk-sch/aftonposten_title_prefs dataset.
22
  It achieves the following results on the evaluation set:
23
+ - Loss: 0.4999
24
+ - Rewards/chosen: 0.0029
25
+ - Rewards/rejected: 0.0025
26
+ - Rewards/accuracies: 0.5220
27
+ - Rewards/margins: 0.0004
28
+ - Logps/rejected: -37.2706
29
+ - Logps/chosen: -33.7447
30
+ - Logits/rejected: -2.2236
31
+ - Logits/chosen: -2.2284
32
 
33
  ## Model description
34
 
 
57
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
58
  - lr_scheduler_type: cosine
59
  - lr_scheduler_warmup_ratio: 0.1
60
+ - num_epochs: 1
61
 
62
  ### Training results
63
 
64
+ | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
65
+ |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
66
+ | 0.4997 | 0.26 | 100 | 0.5000 | 0.0018 | 0.0016 | 0.5183 | 0.0002 | -37.3521 | -33.8530 | -2.2272 | -2.2320 |
67
+ | 0.4992 | 0.52 | 200 | 0.4999 | 0.0029 | 0.0024 | 0.5748 | 0.0006 | -37.2812 | -33.7422 | -2.2240 | -2.2288 |
68
+ | 0.4982 | 0.78 | 300 | 0.4999 | 0.0028 | 0.0024 | 0.5042 | 0.0004 | -37.2741 | -33.7545 | -2.2241 | -2.2289 |
 
 
 
 
 
 
 
 
 
 
 
 
69
 
70
 
71
  ### Framework versions
adapter_config.json CHANGED
@@ -19,13 +19,13 @@
19
  "rank_pattern": {},
20
  "revision": null,
21
  "target_modules": [
22
- "k_proj",
23
- "gate_proj",
24
- "up_proj",
25
  "o_proj",
26
  "v_proj",
27
- "q_proj",
28
- "down_proj"
 
 
 
29
  ],
30
  "task_type": "CAUSAL_LM",
31
  "use_rslora": false
 
19
  "rank_pattern": {},
20
  "revision": null,
21
  "target_modules": [
 
 
 
22
  "o_proj",
23
  "v_proj",
24
+ "gate_proj",
25
+ "up_proj",
26
+ "k_proj",
27
+ "down_proj",
28
+ "q_proj"
29
  ],
30
  "task_type": "CAUSAL_LM",
31
  "use_rslora": false
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d535cfc742dc0e2c419ba82cf3e8f21dfabd2985b6cda286611de399c6285e09
3
  size 176183216
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dbfa9882e97a25037d7e01c5453af5e65cce8583b662f492656d078e5281460d
3
  size 176183216
all_results.json CHANGED
@@ -1,21 +1,21 @@
1
  {
2
- "epoch": 4.0,
3
- "eval_logits/chosen": -1.6202374696731567,
4
- "eval_logits/rejected": -1.6162517070770264,
5
- "eval_logps/chosen": -39.08197021484375,
6
- "eval_logps/rejected": -43.41385269165039,
7
- "eval_loss": 0.6893165707588196,
8
- "eval_rewards/accuracies": 0.5365448594093323,
9
- "eval_rewards/chosen": -0.050474222749471664,
10
- "eval_rewards/margins": 0.008498109877109528,
11
- "eval_rewards/rejected": -0.05897233262658119,
12
- "eval_runtime": 145.7013,
13
  "eval_samples": 343,
14
- "eval_samples_per_second": 2.354,
15
- "eval_steps_per_second": 0.295,
16
- "train_loss": 0.5396727961379212,
17
- "train_runtime": 10793.7948,
18
  "train_samples": 3079,
19
- "train_samples_per_second": 1.141,
20
- "train_steps_per_second": 0.143
21
  }
 
1
  {
2
+ "epoch": 1.0,
3
+ "eval_logits/chosen": -2.228447198867798,
4
+ "eval_logits/rejected": -2.223632335662842,
5
+ "eval_logps/chosen": -33.74469757080078,
6
+ "eval_logps/rejected": -37.27058410644531,
7
+ "eval_loss": 0.49989354610443115,
8
+ "eval_rewards/accuracies": 0.5220099687576294,
9
+ "eval_rewards/chosen": 0.002898545004427433,
10
+ "eval_rewards/margins": 0.0004381774051580578,
11
+ "eval_rewards/rejected": 0.002460367511957884,
12
+ "eval_runtime": 145.5095,
13
  "eval_samples": 343,
14
+ "eval_samples_per_second": 2.357,
15
+ "eval_steps_per_second": 0.296,
16
+ "train_loss": 0.4993510308203759,
17
+ "train_runtime": 3251.761,
18
  "train_samples": 3079,
19
+ "train_samples_per_second": 0.947,
20
+ "train_steps_per_second": 0.118
21
  }
eval_results.json CHANGED
@@ -1,16 +1,16 @@
1
  {
2
- "epoch": 4.0,
3
- "eval_logits/chosen": -1.6202374696731567,
4
- "eval_logits/rejected": -1.6162517070770264,
5
- "eval_logps/chosen": -39.08197021484375,
6
- "eval_logps/rejected": -43.41385269165039,
7
- "eval_loss": 0.6893165707588196,
8
- "eval_rewards/accuracies": 0.5365448594093323,
9
- "eval_rewards/chosen": -0.050474222749471664,
10
- "eval_rewards/margins": 0.008498109877109528,
11
- "eval_rewards/rejected": -0.05897233262658119,
12
- "eval_runtime": 145.7013,
13
  "eval_samples": 343,
14
- "eval_samples_per_second": 2.354,
15
- "eval_steps_per_second": 0.295
16
  }
 
1
  {
2
+ "epoch": 1.0,
3
+ "eval_logits/chosen": -2.228447198867798,
4
+ "eval_logits/rejected": -2.223632335662842,
5
+ "eval_logps/chosen": -33.74469757080078,
6
+ "eval_logps/rejected": -37.27058410644531,
7
+ "eval_loss": 0.49989354610443115,
8
+ "eval_rewards/accuracies": 0.5220099687576294,
9
+ "eval_rewards/chosen": 0.002898545004427433,
10
+ "eval_rewards/margins": 0.0004381774051580578,
11
+ "eval_rewards/rejected": 0.002460367511957884,
12
+ "eval_runtime": 145.5095,
13
  "eval_samples": 343,
14
+ "eval_samples_per_second": 2.357,
15
+ "eval_steps_per_second": 0.296
16
  }
train_results.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
- "epoch": 4.0,
3
- "train_loss": 0.5396727961379212,
4
- "train_runtime": 10793.7948,
5
  "train_samples": 3079,
6
- "train_samples_per_second": 1.141,
7
- "train_steps_per_second": 0.143
8
  }
 
1
  {
2
+ "epoch": 1.0,
3
+ "train_loss": 0.4993510308203759,
4
+ "train_runtime": 3251.761,
5
  "train_samples": 3079,
6
+ "train_samples_per_second": 0.947,
7
+ "train_steps_per_second": 0.118
8
  }
trainer_state.json CHANGED
The diff for this file is too large to render. See raw diff
 
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:4400f1c0d8aaba134c6fc67aef3a063d90862344946e24210f86ac93feadd29a
3
  size 4984
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5787441541dcfce7292e7a26e4973994d4bc0ee529aed272d19db5c5530e8654
3
  size 4984