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
b70143b
•
1 Parent(s): ae95213

Training in progress, step 1300

Browse files
Files changed (8) hide show
  1. README.md +14 -26
  2. adapter_config.json +4 -4
  3. adapter_model.safetensors +1 -1
  4. all_results.json +15 -15
  5. eval_results.json +11 -11
  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: 0.7456
24
- - Rewards/chosen: 0.0514
25
- - Rewards/rejected: 0.0332
26
  - Rewards/accuracies: 0.5365
27
- - Rewards/margins: 0.0182
28
- - Logps/rejected: -37.4614
29
- - Logps/chosen: -33.9489
30
- - Logits/rejected: -2.2407
31
- - Logits/chosen: -2.2456
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
- | 0.5903 | 0.26 | 100 | -2.2366 | -2.2318 | -34.0141 | -37.5087 | 0.7450 | 0.5241 | 0.0123 | 0.0075 | 0.0047 |
67
- | 0.59 | 0.52 | 200 | -2.2332 | -2.2284 | -34.0691 | -37.5752 | 0.7500 | 0.5365 | -0.0208 | 0.0144 | -0.0352 |
68
- | 0.3871 | 0.78 | 300 | -2.2333 | -2.2285 | -34.0622 | -37.5657 | 0.7655 | 0.5071 | -0.0166 | 0.0129 | -0.0295 |
69
- | 0.3315 | 1.04 | 400 | 0.7479 | -0.0009 | -0.0296 | 0.5515 | 0.0287 | -37.5660 | -34.0361 | -2.2265 | -2.2314 |
70
- | 0.3857 | 1.3 | 500 | 0.7601 | -0.0010 | -0.0248 | 0.5577 | 0.0238 | -37.5580 | -34.0362 | -2.2294 | -2.2343 |
71
- | 0.5531 | 1.56 | 600 | 0.7435 | 0.0040 | -0.0183 | 0.5577 | 0.0223 | -37.5471 | -34.0279 | -2.2416 | -2.2465 |
72
- | 0.4315 | 1.82 | 700 | 0.8141 | -0.0077 | -0.0028 | 0.4996 | -0.0049 | -37.5213 | -34.0474 | -2.2449 | -2.2498 |
73
- | 0.3445 | 2.08 | 800 | 0.8104 | 0.0382 | 0.0369 | 0.5282 | 0.0012 | -37.4551 | -33.9710 | -2.2437 | -2.2485 |
74
- | 0.1902 | 2.34 | 900 | 0.8123 | 0.0329 | 0.0364 | 0.5199 | -0.0035 | -37.4560 | -33.9797 | -2.2426 | -2.2475 |
75
- | 0.1724 | 2.6 | 1000 | 0.7850 | 0.0479 | 0.0385 | 0.5365 | 0.0094 | -37.4525 | -33.9548 | -2.2433 | -2.2481 |
76
- | 0.1864 | 2.86 | 1100 | 0.7625 | 0.0435 | 0.0297 | 0.5133 | 0.0137 | -37.4671 | -33.9621 | -2.2392 | -2.2441 |
77
- | 0.1629 | 3.12 | 1200 | 0.7624 | 0.0479 | 0.0337 | 0.5341 | 0.0142 | -37.4604 | -33.9547 | -2.2387 | -2.2436 |
78
- | 0.1881 | 3.38 | 1300 | 0.7945 | 0.0454 | 0.0440 | 0.4967 | 0.0014 | -37.4432 | -33.9588 | -2.2391 | -2.2440 |
79
- | 0.1395 | 3.64 | 1400 | 0.7520 | 0.0496 | 0.0341 | 0.5453 | 0.0156 | -37.4598 | -33.9518 | -2.2408 | -2.2457 |
80
- | 0.1436 | 3.9 | 1500 | 0.7352 | 0.0537 | 0.0336 | 0.5399 | 0.0201 | -37.4606 | -33.9451 | -2.2407 | -2.2456 |
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: 0.9715
24
+ - Rewards/chosen: -0.0427
25
+ - Rewards/rejected: -0.0715
26
  - Rewards/accuracies: 0.5365
27
+ - Rewards/margins: 0.0289
28
+ - Logps/rejected: -37.6359
29
+ - Logps/chosen: -34.1056
30
+ - Logits/rejected: -2.2191
31
+ - Logits/chosen: -2.2239
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
+ | 0.9204 | 0.26 | 100 | 0.9816 | -0.0008 | -0.0193 | 0.5395 | 0.0185 | -37.5487 | -34.0358 | -2.2268 | -2.2317 |
67
+ | 0.8045 | 0.52 | 200 | 0.9791 | -0.0166 | -0.0378 | 0.5361 | 0.0211 | -37.5796 | -34.0623 | -2.2238 | -2.2287 |
68
+ | 0.7186 | 0.78 | 300 | 0.9732 | -0.0459 | -0.0732 | 0.5071 | 0.0273 | -37.6386 | -34.1110 | -2.2194 | -2.2243 |
 
 
 
 
 
 
 
 
 
 
 
 
69
 
70
 
71
  ### Framework versions
adapter_config.json CHANGED
@@ -19,13 +19,13 @@
19
  "rank_pattern": {},
20
  "revision": null,
21
  "target_modules": [
22
- "v_proj",
23
- "down_proj",
24
  "up_proj",
25
  "gate_proj",
 
 
 
26
  "q_proj",
27
- "k_proj",
28
- "o_proj"
29
  ],
30
  "task_type": "CAUSAL_LM",
31
  "use_rslora": false
 
19
  "rank_pattern": {},
20
  "revision": null,
21
  "target_modules": [
 
 
22
  "up_proj",
23
  "gate_proj",
24
+ "v_proj",
25
+ "down_proj",
26
+ "o_proj",
27
  "q_proj",
28
+ "k_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:f6179b75e61780f60641e1691c7393bf4991bc2b020dc7ae011150d91bac1dbb
3
  size 176183216
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df9ac456bc5c50a44802d58adaa065200dd85d39d58e8b817f75a4e4835e4858
3
  size 176183216
all_results.json CHANGED
@@ -1,21 +1,21 @@
1
  {
2
- "epoch": 4.0,
3
- "eval_logits/chosen": -2.2455787658691406,
4
- "eval_logits/rejected": -2.2406980991363525,
5
- "eval_logps/chosen": -33.94891357421875,
6
- "eval_logps/rejected": -37.46136474609375,
7
- "eval_loss": 0.7456009984016418,
8
  "eval_rewards/accuracies": 0.5365448594093323,
9
- "eval_rewards/chosen": 0.051383573561906815,
10
- "eval_rewards/margins": 0.018230266869068146,
11
- "eval_rewards/rejected": 0.03315330296754837,
12
- "eval_runtime": 145.63,
13
  "eval_samples": 343,
14
- "eval_samples_per_second": 2.355,
15
  "eval_steps_per_second": 0.295,
16
- "train_loss": 0.2286405209984098,
17
- "train_runtime": 10803.5829,
18
  "train_samples": 3079,
19
- "train_samples_per_second": 1.14,
20
- "train_steps_per_second": 0.143
21
  }
 
1
  {
2
+ "epoch": 1.0,
3
+ "eval_logits/chosen": -2.223938465118408,
4
+ "eval_logits/rejected": -2.2191123962402344,
5
+ "eval_logps/chosen": -34.10564041137695,
6
+ "eval_logps/rejected": -37.63585662841797,
7
+ "eval_loss": 0.9714530110359192,
8
  "eval_rewards/accuracies": 0.5365448594093323,
9
+ "eval_rewards/chosen": -0.042656492441892624,
10
+ "eval_rewards/margins": 0.028885547071695328,
11
+ "eval_rewards/rejected": -0.07154203206300735,
12
+ "eval_runtime": 145.6066,
13
  "eval_samples": 343,
14
+ "eval_samples_per_second": 2.356,
15
  "eval_steps_per_second": 0.295,
16
+ "train_loss": 0.859264765157328,
17
+ "train_runtime": 3251.7628,
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": -2.2455787658691406,
4
- "eval_logits/rejected": -2.2406980991363525,
5
- "eval_logps/chosen": -33.94891357421875,
6
- "eval_logps/rejected": -37.46136474609375,
7
- "eval_loss": 0.7456009984016418,
8
  "eval_rewards/accuracies": 0.5365448594093323,
9
- "eval_rewards/chosen": 0.051383573561906815,
10
- "eval_rewards/margins": 0.018230266869068146,
11
- "eval_rewards/rejected": 0.03315330296754837,
12
- "eval_runtime": 145.63,
13
  "eval_samples": 343,
14
- "eval_samples_per_second": 2.355,
15
  "eval_steps_per_second": 0.295
16
  }
 
1
  {
2
+ "epoch": 1.0,
3
+ "eval_logits/chosen": -2.223938465118408,
4
+ "eval_logits/rejected": -2.2191123962402344,
5
+ "eval_logps/chosen": -34.10564041137695,
6
+ "eval_logps/rejected": -37.63585662841797,
7
+ "eval_loss": 0.9714530110359192,
8
  "eval_rewards/accuracies": 0.5365448594093323,
9
+ "eval_rewards/chosen": -0.042656492441892624,
10
+ "eval_rewards/margins": 0.028885547071695328,
11
+ "eval_rewards/rejected": -0.07154203206300735,
12
+ "eval_runtime": 145.6066,
13
  "eval_samples": 343,
14
+ "eval_samples_per_second": 2.356,
15
  "eval_steps_per_second": 0.295
16
  }
train_results.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
- "epoch": 4.0,
3
- "train_loss": 0.2286405209984098,
4
- "train_runtime": 10803.5829,
5
  "train_samples": 3079,
6
- "train_samples_per_second": 1.14,
7
- "train_steps_per_second": 0.143
8
  }
 
1
  {
2
+ "epoch": 1.0,
3
+ "train_loss": 0.859264765157328,
4
+ "train_runtime": 3251.7628,
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:eacae65050e9f803beda9fe6763ea7afd8ff2656b8fcda6cbf8438d16e92bb88
3
  size 4984
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d9960c65d0d5a6a27215911c17f893f94d008fa1e5d2ab0c63c068b215f19a9
3
  size 4984