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
f8089f0
•
1 Parent(s): 16f9588

Training in progress, step 800

Browse files
Files changed (8) hide show
  1. README.md +15 -27
  2. adapter_config.json +4 -4
  3. adapter_model.safetensors +1 -1
  4. all_results.json +16 -16
  5. eval_results.json +12 -12
  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: 2461.9446
24
- - Rewards/chosen: -0.0198
25
- - Rewards/rejected: -0.0245
26
- - Rewards/accuracies: 0.5569
27
- - Rewards/margins: 0.0047
28
- - Logps/rejected: -39.9694
29
- - Logps/chosen: -36.0193
30
- - Logits/rejected: -1.8604
31
- - Logits/chosen: -1.8647
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
- | 2484.2775 | 0.26 | 100 | -2.2323 | -2.2275 | -34.0187 | -37.5204 | 2498.1597 | 0.5565 | 0.0002 | 0.0002 | -0.0000 |
67
- | 2464.7859 | 0.52 | 200 | -2.2291 | -2.2242 | -34.0338 | -37.5574 | 2496.0791 | 0.5544 | 0.0000 | 0.0004 | -0.0004 |
68
- | 2430.5529 | 0.78 | 300 | -2.2262 | -2.2213 | -34.0629 | -37.5906 | 2495.6680 | 0.5390 | -0.0003 | 0.0005 | -0.0007 |
69
- | 2370.302 | 1.04 | 400 | 2492.4653 | -0.0006 | -0.0014 | 0.5623 | 0.0008 | -37.6613 | -34.0994 | -2.1946 | -2.1994 |
70
- | 2298.9412 | 1.3 | 500 | 2488.4062 | -0.0025 | -0.0037 | 0.5772 | 0.0013 | -37.8883 | -34.2804 | -2.1295 | -2.1343 |
71
- | 2298.7582 | 1.56 | 600 | 2477.6589 | -0.0053 | -0.0077 | 0.6121 | 0.0024 | -38.2854 | -34.5600 | -2.0589 | -2.0636 |
72
- | 2254.2998 | 1.82 | 700 | 2477.6624 | -0.0096 | -0.0122 | 0.5540 | 0.0026 | -38.7330 | -34.9921 | -2.0040 | -2.0086 |
73
- | 2173.4393 | 2.08 | 800 | 2470.5903 | -0.0118 | -0.0152 | 0.5569 | 0.0034 | -39.0386 | -35.2171 | -1.9665 | -1.9710 |
74
- | 2065.36 | 2.34 | 900 | 2467.7329 | -0.0156 | -0.0195 | 0.5598 | 0.0039 | -39.4668 | -35.5978 | -1.9122 | -1.9166 |
75
- | 2196.3246 | 2.6 | 1000 | 2464.3682 | -0.0186 | -0.0229 | 0.5511 | 0.0044 | -39.8091 | -35.8901 | -1.8782 | -1.8826 |
76
- | 2237.6512 | 2.86 | 1100 | 2464.2991 | -0.0196 | -0.0241 | 0.5453 | 0.0044 | -39.9232 | -35.9976 | -1.8646 | -1.8690 |
77
- | 2032.5133 | 3.12 | 1200 | 2464.4653 | -0.0199 | -0.0243 | 0.5598 | 0.0044 | -39.9447 | -36.0201 | -1.8618 | -1.8662 |
78
- | 1967.852 | 3.38 | 1300 | 2461.2036 | -0.0197 | -0.0245 | 0.5540 | 0.0048 | -39.9632 | -36.0051 | -1.8608 | -1.8652 |
79
- | 2084.0914 | 3.64 | 1400 | 2462.6121 | -0.0199 | -0.0245 | 0.5598 | 0.0046 | -39.9675 | -36.0243 | -1.8613 | -1.8656 |
80
- | 2032.7156 | 3.9 | 1500 | 2463.1057 | -0.0198 | -0.0244 | 0.5656 | 0.0046 | -39.9582 | -36.0192 | -1.8605 | -1.8648 |
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: 0.9997
24
+ - Rewards/chosen: -0.0006
25
+ - Rewards/rejected: -0.0009
26
+ - Rewards/accuracies: 0.5478
27
+ - Rewards/margins: 0.0003
28
+ - Logps/rejected: -37.6060
29
+ - Logps/chosen: -34.0918
30
+ - Logits/rejected: -2.2164
31
+ - Logits/chosen: -2.2213
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
+ | 0.9987 | 0.26 | 100 | 0.9998 | -0.0002 | -0.0004 | 0.5336 | 0.0002 | -37.5528 | -34.0499 | -2.2264 | -2.2313 |
67
+ | 0.9965 | 0.52 | 200 | 0.9996 | -0.0003 | -0.0006 | 0.5071 | 0.0004 | -37.5790 | -34.0618 | -2.2204 | -2.2252 |
68
+ | 0.9925 | 0.78 | 300 | 0.9996 | -0.0005 | -0.0009 | 0.5594 | 0.0004 | -37.6063 | -34.0836 | -2.2166 | -2.2214 |
 
 
 
 
 
 
 
 
 
 
 
 
69
 
70
 
71
  ### Framework versions
adapter_config.json CHANGED
@@ -19,13 +19,13 @@
19
  "rank_pattern": {},
20
  "revision": null,
21
  "target_modules": [
22
- "q_proj",
23
  "o_proj",
24
  "v_proj",
25
- "gate_proj",
26
- "up_proj",
27
  "down_proj",
28
- "k_proj"
 
 
29
  ],
30
  "task_type": "CAUSAL_LM",
31
  "use_rslora": false
 
19
  "rank_pattern": {},
20
  "revision": null,
21
  "target_modules": [
22
+ "k_proj",
23
  "o_proj",
24
  "v_proj",
 
 
25
  "down_proj",
26
+ "up_proj",
27
+ "gate_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:b5eb4a5adaea6cb5fb54a0d4729a15c3ab2d9782eadbc65d5a950651eb8fdc92
3
  size 176183216
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13b1b1ded234bb23e6377dd3d1e9e122e16d0ac59838acd92736cccece8b6ea6
3
  size 176183216
all_results.json CHANGED
@@ -1,21 +1,21 @@
1
  {
2
- "epoch": 4.0,
3
- "eval_logits/chosen": -1.864721417427063,
4
- "eval_logits/rejected": -1.8603547811508179,
5
- "eval_logps/chosen": -36.019264221191406,
6
- "eval_logps/rejected": -39.96937561035156,
7
- "eval_loss": 2461.944580078125,
8
- "eval_rewards/accuracies": 0.5568937063217163,
9
- "eval_rewards/chosen": -0.019847100600600243,
10
- "eval_rewards/margins": 0.0046804738231003284,
11
- "eval_rewards/rejected": -0.024527573958039284,
12
- "eval_runtime": 145.7077,
13
  "eval_samples": 343,
14
- "eval_samples_per_second": 2.354,
15
  "eval_steps_per_second": 0.295,
16
- "train_loss": 1756.4536297686689,
17
- "train_runtime": 10797.265,
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": -2.221278667449951,
4
+ "eval_logits/rejected": -2.2164483070373535,
5
+ "eval_logps/chosen": -34.091766357421875,
6
+ "eval_logps/rejected": -37.60596466064453,
7
+ "eval_loss": 0.9996774792671204,
8
+ "eval_rewards/accuracies": 0.5477575063705444,
9
+ "eval_rewards/chosen": -0.000572161574382335,
10
+ "eval_rewards/margins": 0.00032129042665474117,
11
+ "eval_rewards/rejected": -0.0008934520301409066,
12
+ "eval_runtime": 145.6061,
13
  "eval_samples": 343,
14
+ "eval_samples_per_second": 2.356,
15
  "eval_steps_per_second": 0.295,
16
+ "train_loss": 0.9973225085766284,
17
+ "train_runtime": 3253.1307,
18
  "train_samples": 3079,
19
+ "train_samples_per_second": 0.946,
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": -1.864721417427063,
4
- "eval_logits/rejected": -1.8603547811508179,
5
- "eval_logps/chosen": -36.019264221191406,
6
- "eval_logps/rejected": -39.96937561035156,
7
- "eval_loss": 2461.944580078125,
8
- "eval_rewards/accuracies": 0.5568937063217163,
9
- "eval_rewards/chosen": -0.019847100600600243,
10
- "eval_rewards/margins": 0.0046804738231003284,
11
- "eval_rewards/rejected": -0.024527573958039284,
12
- "eval_runtime": 145.7077,
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": -2.221278667449951,
4
+ "eval_logits/rejected": -2.2164483070373535,
5
+ "eval_logps/chosen": -34.091766357421875,
6
+ "eval_logps/rejected": -37.60596466064453,
7
+ "eval_loss": 0.9996774792671204,
8
+ "eval_rewards/accuracies": 0.5477575063705444,
9
+ "eval_rewards/chosen": -0.000572161574382335,
10
+ "eval_rewards/margins": 0.00032129042665474117,
11
+ "eval_rewards/rejected": -0.0008934520301409066,
12
+ "eval_runtime": 145.6061,
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": 1756.4536297686689,
4
- "train_runtime": 10797.265,
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": 0.9973225085766284,
4
+ "train_runtime": 3253.1307,
5
  "train_samples": 3079,
6
+ "train_samples_per_second": 0.946,
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 sha256:2b8abfa58261899a25124dd1b0d5038d5a5ec747d30923843a4c529696b76c8d
3
  size 4984
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:237d42c7ba4f1b445e179b3dad6558bffee1c3e58cd08115fd9ec93a609a8a24
3
  size 4984