hugodk-sch
commited on
Commit
•
5826cb2
1
Parent(s):
68015c7
Training in progress, step 600
Browse files- README.md +18 -30
- adapter_config.json +3 -3
- adapter_model.safetensors +1 -1
- all_results.json +17 -17
- config.json +1 -3
- eval_results.json +13 -13
- train_results.json +5 -5
- trainer_state.json +0 -0
- 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.
|
24 |
-
- Rewards/chosen: -0.
|
25 |
-
- Rewards/rejected: -0.
|
26 |
-
- Rewards/accuracies: 0.
|
27 |
-
- Rewards/margins: 0.
|
28 |
-
- Logps/rejected: -
|
29 |
-
- Logps/chosen: -34.
|
30 |
-
- Logits/rejected: -2.
|
31 |
-
- Logits/chosen: -2.
|
32 |
|
33 |
## Model description
|
34 |
|
@@ -57,33 +57,21 @@ 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:
|
61 |
|
62 |
### Training results
|
63 |
|
64 |
-
| Training Loss | Epoch | Step |
|
65 |
-
|
66 |
-
| 0.
|
67 |
-
| 0.
|
68 |
-
| 0.
|
69 |
-
| 0.5492 | 1.04 | 400 | 0.9600 | -0.0653 | -0.1052 | 0.5565 | 0.0399 | -37.7797 | -34.1978 | -2.1852 | -2.1901 |
|
70 |
-
| 0.4504 | 1.3 | 500 | 0.9656 | -0.1403 | -0.1747 | 0.5685 | 0.0344 | -37.9533 | -34.3853 | -2.1770 | -2.1818 |
|
71 |
-
| 0.4511 | 1.56 | 600 | 0.9338 | -0.0835 | -0.1505 | 0.5951 | 0.0670 | -37.8930 | -34.2433 | -2.1565 | -2.1613 |
|
72 |
-
| 0.3805 | 1.82 | 700 | 0.9385 | -0.1061 | -0.1691 | 0.5511 | 0.0631 | -37.9394 | -34.2997 | -2.1442 | -2.1490 |
|
73 |
-
| 0.2038 | 2.08 | 800 | 0.9334 | -0.1274 | -0.1969 | 0.5689 | 0.0695 | -38.0088 | -34.3529 | -2.1367 | -2.1415 |
|
74 |
-
| 0.2332 | 2.34 | 900 | 0.9320 | -0.1353 | -0.2057 | 0.5718 | 0.0704 | -38.0309 | -34.3727 | -2.1352 | -2.1400 |
|
75 |
-
| 0.28 | 2.6 | 1000 | 0.9271 | -0.1454 | -0.2223 | 0.5714 | 0.0769 | -38.0723 | -34.3980 | -2.1345 | -2.1393 |
|
76 |
-
| 0.1953 | 2.86 | 1100 | 0.9399 | -0.1551 | -0.2188 | 0.5631 | 0.0637 | -38.0636 | -34.4223 | -2.1342 | -2.1390 |
|
77 |
-
| 0.2936 | 3.12 | 1200 | 0.9311 | -0.1412 | -0.2138 | 0.5864 | 0.0725 | -38.0510 | -34.3876 | -2.1359 | -2.1407 |
|
78 |
-
| 0.1526 | 3.38 | 1300 | 0.9307 | -0.1413 | -0.2150 | 0.5864 | 0.0737 | -38.0541 | -34.3877 | -2.1350 | -2.1398 |
|
79 |
-
| 0.1121 | 3.64 | 1400 | 0.9250 | -0.1377 | -0.2153 | 0.6038 | 0.0776 | -38.0548 | -34.3788 | -2.1356 | -2.1404 |
|
80 |
-
| 0.215 | 3.9 | 1500 | 0.9280 | -0.1384 | -0.2143 | 0.5806 | 0.0758 | -38.0523 | -34.3807 | -2.1358 | -2.1405 |
|
81 |
|
82 |
|
83 |
### Framework versions
|
84 |
|
85 |
-
- PEFT 0.
|
86 |
-
- Transformers 4.
|
87 |
- Pytorch 2.1.2+cu121
|
88 |
-
- Datasets 2.
|
89 |
- Tokenizers 0.15.1
|
|
|
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: 0.6931
|
24 |
+
- Rewards/chosen: -0.0260
|
25 |
+
- Rewards/rejected: -0.0334
|
26 |
+
- Rewards/accuracies: 0.5071
|
27 |
+
- Rewards/margins: 0.0074
|
28 |
+
- Logps/rejected: -37.6001
|
29 |
+
- Logps/chosen: -34.0996
|
30 |
+
- Logits/rejected: -2.2227
|
31 |
+
- Logits/chosen: -2.2275
|
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 |
+
| 0.6726 | 0.26 | 100 | 0.6903 | -0.0039 | -0.0143 | 0.5482 | 0.0105 | -37.5524 | -34.0442 | -2.2286 | -2.2334 |
|
67 |
+
| 0.6419 | 0.52 | 200 | 0.6932 | -0.0126 | -0.0196 | 0.5042 | 0.0071 | -37.5657 | -34.0660 | -2.2261 | -2.2310 |
|
68 |
+
| 0.5902 | 0.78 | 300 | 0.6926 | -0.0248 | -0.0328 | 0.4954 | 0.0080 | -37.5987 | -34.0966 | -2.2231 | -2.2279 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
|
70 |
|
71 |
### Framework versions
|
72 |
|
73 |
+
- PEFT 0.8.2
|
74 |
+
- Transformers 4.37.2
|
75 |
- Pytorch 2.1.2+cu121
|
76 |
+
- Datasets 2.17.0
|
77 |
- Tokenizers 0.15.1
|
adapter_config.json
CHANGED
@@ -20,13 +20,13 @@
|
|
20 |
"rank_pattern": {},
|
21 |
"revision": null,
|
22 |
"target_modules": [
|
23 |
-
"k_proj",
|
24 |
"o_proj",
|
25 |
-
"v_proj",
|
26 |
"up_proj",
|
27 |
"q_proj",
|
|
|
28 |
"down_proj",
|
29 |
-
"
|
|
|
30 |
],
|
31 |
"task_type": "CAUSAL_LM",
|
32 |
"use_dora": false,
|
|
|
20 |
"rank_pattern": {},
|
21 |
"revision": null,
|
22 |
"target_modules": [
|
|
|
23 |
"o_proj",
|
|
|
24 |
"up_proj",
|
25 |
"q_proj",
|
26 |
+
"gate_proj",
|
27 |
"down_proj",
|
28 |
+
"v_proj",
|
29 |
+
"k_proj"
|
30 |
],
|
31 |
"task_type": "CAUSAL_LM",
|
32 |
"use_dora": false,
|
adapter_model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 176183216
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1e7a21fcbd2827f352f3cf94a59b15b79ab640f6875119f764d0e0ed0d9120fe
|
3 |
size 176183216
|
all_results.json
CHANGED
@@ -1,21 +1,21 @@
|
|
1 |
{
|
2 |
-
"epoch":
|
3 |
-
"eval_logits/chosen": -2.
|
4 |
-
"eval_logits/rejected": -2.
|
5 |
-
"eval_logps/chosen": -34.
|
6 |
-
"eval_logps/rejected": -
|
7 |
-
"eval_loss": 0.
|
8 |
-
"eval_rewards/accuracies": 0.
|
9 |
-
"eval_rewards/chosen": -0.
|
10 |
-
"eval_rewards/margins": 0.
|
11 |
-
"eval_rewards/rejected": -0.
|
12 |
-
"eval_runtime": 145.
|
13 |
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.
|
15 |
-
"eval_steps_per_second": 0.
|
16 |
-
"train_loss": 0.
|
17 |
-
"train_runtime":
|
18 |
"train_samples": 3079,
|
19 |
-
"train_samples_per_second":
|
20 |
-
"train_steps_per_second": 0.
|
21 |
}
|
|
|
1 |
{
|
2 |
+
"epoch": 1.0,
|
3 |
+
"eval_logits/chosen": -2.2275288105010986,
|
4 |
+
"eval_logits/rejected": -2.2226884365081787,
|
5 |
+
"eval_logps/chosen": -34.09955978393555,
|
6 |
+
"eval_logps/rejected": -37.600067138671875,
|
7 |
+
"eval_loss": 0.693101167678833,
|
8 |
+
"eval_rewards/accuracies": 0.5070598125457764,
|
9 |
+
"eval_rewards/chosen": -0.026001954451203346,
|
10 |
+
"eval_rewards/margins": 0.007375751622021198,
|
11 |
+
"eval_rewards/rejected": -0.03337770700454712,
|
12 |
+
"eval_runtime": 145.5528,
|
13 |
"eval_samples": 343,
|
14 |
+
"eval_samples_per_second": 2.357,
|
15 |
+
"eval_steps_per_second": 0.295,
|
16 |
+
"train_loss": 0.6506414413452148,
|
17 |
+
"train_runtime": 3254.0644,
|
18 |
"train_samples": 3079,
|
19 |
+
"train_samples_per_second": 0.946,
|
20 |
+
"train_steps_per_second": 0.118
|
21 |
}
|
config.json
CHANGED
@@ -18,8 +18,6 @@
|
|
18 |
"n_layer": 28,
|
19 |
"n_positions": 2048,
|
20 |
"quantization_config": {
|
21 |
-
"_load_in_4bit": true,
|
22 |
-
"_load_in_8bit": false,
|
23 |
"bnb_4bit_compute_dtype": "float16",
|
24 |
"bnb_4bit_quant_type": "nf4",
|
25 |
"bnb_4bit_use_double_quant": false,
|
@@ -47,7 +45,7 @@
|
|
47 |
},
|
48 |
"tie_word_embeddings": false,
|
49 |
"tokenizer_class": "GPT2Tokenizer",
|
50 |
-
"transformers_version": "4.
|
51 |
"use_cache": true,
|
52 |
"vocab_size": 50400
|
53 |
}
|
|
|
18 |
"n_layer": 28,
|
19 |
"n_positions": 2048,
|
20 |
"quantization_config": {
|
|
|
|
|
21 |
"bnb_4bit_compute_dtype": "float16",
|
22 |
"bnb_4bit_quant_type": "nf4",
|
23 |
"bnb_4bit_use_double_quant": false,
|
|
|
45 |
},
|
46 |
"tie_word_embeddings": false,
|
47 |
"tokenizer_class": "GPT2Tokenizer",
|
48 |
+
"transformers_version": "4.37.2",
|
49 |
"use_cache": true,
|
50 |
"vocab_size": 50400
|
51 |
}
|
eval_results.json
CHANGED
@@ -1,16 +1,16 @@
|
|
1 |
{
|
2 |
-
"epoch":
|
3 |
-
"eval_logits/chosen": -2.
|
4 |
-
"eval_logits/rejected": -2.
|
5 |
-
"eval_logps/chosen": -34.
|
6 |
-
"eval_logps/rejected": -
|
7 |
-
"eval_loss": 0.
|
8 |
-
"eval_rewards/accuracies": 0.
|
9 |
-
"eval_rewards/chosen": -0.
|
10 |
-
"eval_rewards/margins": 0.
|
11 |
-
"eval_rewards/rejected": -0.
|
12 |
-
"eval_runtime": 145.
|
13 |
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.
|
15 |
-
"eval_steps_per_second": 0.
|
16 |
}
|
|
|
1 |
{
|
2 |
+
"epoch": 1.0,
|
3 |
+
"eval_logits/chosen": -2.2275288105010986,
|
4 |
+
"eval_logits/rejected": -2.2226884365081787,
|
5 |
+
"eval_logps/chosen": -34.09955978393555,
|
6 |
+
"eval_logps/rejected": -37.600067138671875,
|
7 |
+
"eval_loss": 0.693101167678833,
|
8 |
+
"eval_rewards/accuracies": 0.5070598125457764,
|
9 |
+
"eval_rewards/chosen": -0.026001954451203346,
|
10 |
+
"eval_rewards/margins": 0.007375751622021198,
|
11 |
+
"eval_rewards/rejected": -0.03337770700454712,
|
12 |
+
"eval_runtime": 145.5528,
|
13 |
"eval_samples": 343,
|
14 |
+
"eval_samples_per_second": 2.357,
|
15 |
+
"eval_steps_per_second": 0.295
|
16 |
}
|
train_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
-
"epoch":
|
3 |
-
"train_loss": 0.
|
4 |
-
"train_runtime":
|
5 |
"train_samples": 3079,
|
6 |
-
"train_samples_per_second":
|
7 |
-
"train_steps_per_second": 0.
|
8 |
}
|
|
|
1 |
{
|
2 |
+
"epoch": 1.0,
|
3 |
+
"train_loss": 0.6506414413452148,
|
4 |
+
"train_runtime": 3254.0644,
|
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:
|
3 |
size 5176
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:363f8936016c2a932863dbc23987c69cadd4a52770f6d9a386ff56314f74fcab
|
3 |
size 5176
|