hugodk-sch
commited on
Commit
•
939e27a
1
Parent(s):
fa7bcfd
Training in progress, step 600
Browse files- README.md +18 -30
- adapter_config.json +4 -4
- 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: -37.
|
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 |
-
| 0.3632 | 1.04 | 400 | 0.9495 | -0.0909 | -0.1434 | 0.5602 | 0.0525 | -37.6760 | -34.1355 | -2.2076 | -2.2125 |
|
70 |
-
| 0.2994 | 1.3 | 500 | 0.9461 | -0.1647 | -0.2318 | 0.5540 | 0.0671 | -37.7742 | -34.2176 | -2.2162 | -2.2210 |
|
71 |
-
| 0.3408 | 1.56 | 600 | 0.9077 | -0.0675 | -0.1694 | 0.5868 | 0.1019 | -37.7048 | -34.1096 | -2.2017 | -2.2066 |
|
72 |
-
| 0.2796 | 1.82 | 700 | 0.9425 | -0.0929 | -0.1626 | 0.5569 | 0.0697 | -37.6973 | -34.1378 | -2.2012 | -2.2061 |
|
73 |
-
| 0.1052 | 2.08 | 800 | 0.9125 | -0.0848 | -0.1863 | 0.5926 | 0.1015 | -37.7236 | -34.1288 | -2.2003 | -2.2051 |
|
74 |
-
| 0.095 | 2.34 | 900 | 0.9005 | -0.0802 | -0.1942 | 0.5540 | 0.1140 | -37.7324 | -34.1237 | -2.2019 | -2.2067 |
|
75 |
-
| 0.123 | 2.6 | 1000 | 0.9194 | -0.0907 | -0.1876 | 0.5511 | 0.0969 | -37.7251 | -34.1353 | -2.1994 | -2.2043 |
|
76 |
-
| 0.0894 | 2.86 | 1100 | 0.9182 | -0.0915 | -0.1890 | 0.5336 | 0.0976 | -37.7267 | -34.1362 | -2.2001 | -2.2050 |
|
77 |
-
| 0.1086 | 3.12 | 1200 | 0.9023 | -0.0864 | -0.1976 | 0.5627 | 0.1112 | -37.7362 | -34.1306 | -2.2006 | -2.2054 |
|
78 |
-
| 0.0577 | 3.38 | 1300 | 0.9154 | -0.0922 | -0.1935 | 0.5598 | 0.1013 | -37.7317 | -34.1370 | -2.2002 | -2.2050 |
|
79 |
-
| 0.0375 | 3.64 | 1400 | 0.9233 | -0.0896 | -0.1810 | 0.5569 | 0.0914 | -37.7178 | -34.1342 | -2.2002 | -2.2050 |
|
80 |
-
| 0.0724 | 3.9 | 1500 | 0.9176 | -0.0906 | -0.1873 | 0.5627 | 0.0967 | -37.7248 | -34.1353 | -2.2000 | -2.2049 |
|
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: 0.6981
|
24 |
+
- Rewards/chosen: -0.0558
|
25 |
+
- Rewards/rejected: -0.0768
|
26 |
+
- Rewards/accuracies: 0.5453
|
27 |
+
- Rewards/margins: 0.0210
|
28 |
+
- Logps/rejected: -37.6020
|
29 |
+
- Logps/chosen: -34.0965
|
30 |
+
- Logits/rejected: -2.2248
|
31 |
+
- Logits/chosen: -2.2297
|
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 |
+
| 0.6464 | 0.26 | 100 | 0.6903 | -0.0054 | -0.0300 | 0.5685 | 0.0246 | -37.5500 | -34.0405 | -2.2291 | -2.2340 |
|
67 |
+
| 0.5931 | 0.52 | 200 | 0.6980 | -0.0346 | -0.0543 | 0.5158 | 0.0196 | -37.5769 | -34.0730 | -2.2267 | -2.2316 |
|
68 |
+
| 0.5301 | 0.78 | 300 | 0.6973 | -0.0555 | -0.0750 | 0.5390 | 0.0195 | -37.6000 | -34.0962 | -2.2243 | -2.2292 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 |
"q_proj",
|
24 |
-
"up_proj",
|
25 |
-
"k_proj",
|
26 |
-
"o_proj",
|
27 |
"v_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 |
+
"gate_proj",
|
24 |
"q_proj",
|
|
|
|
|
|
|
25 |
"v_proj",
|
26 |
+
"k_proj",
|
27 |
"down_proj",
|
28 |
+
"up_proj",
|
29 |
+
"o_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:e7a780872ae0ab102acf92ea1710d8aece0f65a8a4c151c4577f63501c468697
|
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": -37.
|
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": -2.22967529296875,
|
4 |
+
"eval_logits/rejected": -2.2248213291168213,
|
5 |
+
"eval_logps/chosen": -34.096519470214844,
|
6 |
+
"eval_logps/rejected": -37.601966857910156,
|
7 |
+
"eval_loss": 0.6981170177459717,
|
8 |
+
"eval_rewards/accuracies": 0.545265793800354,
|
9 |
+
"eval_rewards/chosen": -0.0557732917368412,
|
10 |
+
"eval_rewards/margins": 0.021040165796875954,
|
11 |
+
"eval_rewards/rejected": -0.07681345194578171,
|
12 |
+
"eval_runtime": 145.6397,
|
13 |
"eval_samples": 343,
|
14 |
+
"eval_samples_per_second": 2.355,
|
15 |
+
"eval_steps_per_second": 0.295,
|
16 |
+
"train_loss": 0.6175476637753573,
|
17 |
+
"train_runtime": 3252.7839,
|
18 |
"train_samples": 3079,
|
19 |
+
"train_samples_per_second": 0.947,
|
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": -37.
|
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": -2.22967529296875,
|
4 |
+
"eval_logits/rejected": -2.2248213291168213,
|
5 |
+
"eval_logps/chosen": -34.096519470214844,
|
6 |
+
"eval_logps/rejected": -37.601966857910156,
|
7 |
+
"eval_loss": 0.6981170177459717,
|
8 |
+
"eval_rewards/accuracies": 0.545265793800354,
|
9 |
+
"eval_rewards/chosen": -0.0557732917368412,
|
10 |
+
"eval_rewards/margins": 0.021040165796875954,
|
11 |
+
"eval_rewards/rejected": -0.07681345194578171,
|
12 |
+
"eval_runtime": 145.6397,
|
13 |
"eval_samples": 343,
|
14 |
+
"eval_samples_per_second": 2.355,
|
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": 0.6175476637753573,
|
4 |
+
"train_runtime": 3252.7839,
|
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:
|
3 |
size 5176
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:bdf4888b081a55a294000817bce0516657ba047dfd048058409647c84bbb0cd8
|
3 |
size 5176
|