Hugging Face's logo

juliadollis
/
mistral-finetuned-toxicity3

PEFT
Safetensors
Generated from Trainer
Model card Files Community
juliadollis commited on
Commit
16b727a
·
verified ·
1 Parent(s): 1b68593

Model save

Browse files
Files changed (3) hide show
  1. README.md +79 -7
  2. adapter_config.json +5 -5
  3. training_args.bin +1 -1
README.md CHANGED
@@ -5,16 +5,18 @@ base_model: mistralai/Mistral-7B-Instruct-v0.3
5
  tags:
6
  - generated_from_trainer
7
  model-index:
8
- - name: mistral-finetuned-toxicity2
9
  results: []
10
  ---
11
 
12
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
13
  should probably proofread and complete it, then remove this comment. -->
14
 
15
- # mistral-finetuned-toxicity2
16
 
17
  This model is a fine-tuned version of [mistralai/Mistral-7B-Instruct-v0.3](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.3) on an unknown dataset.
 
 
18
 
19
  ## Model description
20
 
@@ -34,18 +36,88 @@ More information needed
34
 
35
  The following hyperparameters were used during training:
36
  - learning_rate: 5e-05
37
- - train_batch_size: 1
38
- - eval_batch_size: 1
39
  - seed: 42
40
- - gradient_accumulation_steps: 16
41
- - total_train_batch_size: 16
42
  - optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
43
  - lr_scheduler_type: linear
44
- - num_epochs: 1
45
  - mixed_precision_training: Native AMP
46
 
47
  ### Training results
48
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
49
 
50
 
51
  ### Framework versions
 
5
  tags:
6
  - generated_from_trainer
7
  model-index:
8
+ - name: mistral-finetuned-toxicity3
9
  results: []
10
  ---
11
 
12
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
13
  should probably proofread and complete it, then remove this comment. -->
14
 
15
+ # mistral-finetuned-toxicity3
16
 
17
  This model is a fine-tuned version of [mistralai/Mistral-7B-Instruct-v0.3](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.3) on an unknown dataset.
18
+ It achieves the following results on the evaluation set:
19
+ - Loss: 0.7477
20
 
21
  ## Model description
22
 
 
36
 
37
  The following hyperparameters were used during training:
38
  - learning_rate: 5e-05
39
+ - train_batch_size: 2
40
+ - eval_batch_size: 2
41
  - seed: 42
42
+ - gradient_accumulation_steps: 4
43
+ - total_train_batch_size: 8
44
  - optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
45
  - lr_scheduler_type: linear
46
+ - num_epochs: 2
47
  - mixed_precision_training: Native AMP
48
 
49
  ### Training results
50
 
51
+ | Training Loss | Epoch | Step | Validation Loss |
52
+ |:-------------:|:------:|:-----:|:---------------:|
53
+ | 0.9782 | 0.0185 | 200 | 0.9489 |
54
+ | 0.9499 | 0.0370 | 400 | 0.9159 |
55
+ | 0.8738 | 0.0555 | 600 | 0.9025 |
56
+ | 0.9153 | 0.0740 | 800 | 0.8785 |
57
+ | 0.8624 | 0.0925 | 1000 | 0.8715 |
58
+ | 0.8629 | 0.1110 | 1200 | 0.8655 |
59
+ | 0.8632 | 0.1294 | 1400 | 0.8521 |
60
+ | 0.838 | 0.1479 | 1600 | 0.8503 |
61
+ | 0.8168 | 0.1664 | 1800 | 0.8456 |
62
+ | 0.8198 | 0.1849 | 2000 | 0.8389 |
63
+ | 0.8243 | 0.2034 | 2200 | 0.8277 |
64
+ | 0.814 | 0.2219 | 2400 | 0.8315 |
65
+ | 0.8027 | 0.2404 | 2600 | 0.8229 |
66
+ | 0.8192 | 0.2589 | 2800 | 0.8173 |
67
+ | 0.8178 | 0.2774 | 3000 | 0.8161 |
68
+ | 0.7955 | 0.2959 | 3200 | 0.8132 |
69
+ | 0.786 | 0.3144 | 3400 | 0.8081 |
70
+ | 0.8196 | 0.3329 | 3600 | 0.8046 |
71
+ | 0.7996 | 0.3514 | 3800 | 0.8034 |
72
+ | 0.8236 | 0.3699 | 4000 | 0.7995 |
73
+ | 0.8192 | 0.3883 | 4200 | 0.7965 |
74
+ | 0.7898 | 0.4068 | 4400 | 0.7920 |
75
+ | 0.8018 | 0.4253 | 4600 | 0.7896 |
76
+ | 0.7837 | 0.4438 | 4800 | 0.7881 |
77
+ | 0.7802 | 0.4623 | 5000 | 0.7885 |
78
+ | 0.7856 | 0.4808 | 5200 | 0.7847 |
79
+ | 0.7873 | 0.4993 | 5400 | 0.7813 |
80
+ | 0.787 | 0.5178 | 5600 | 0.7806 |
81
+ | 0.7871 | 0.5363 | 5800 | 0.7781 |
82
+ | 0.7955 | 0.5548 | 6000 | 0.7787 |
83
+ | 0.7857 | 0.5733 | 6200 | 0.7745 |
84
+ | 0.7817 | 0.5918 | 6400 | 0.7729 |
85
+ | 0.7841 | 0.6103 | 6600 | 0.7735 |
86
+ | 0.7474 | 0.6288 | 6800 | 0.7683 |
87
+ | 0.7597 | 0.6472 | 7000 | 0.7707 |
88
+ | 0.7591 | 0.6657 | 7200 | 0.7666 |
89
+ | 0.7615 | 0.6842 | 7400 | 0.7646 |
90
+ | 0.7366 | 0.7027 | 7600 | 0.7647 |
91
+ | 0.7697 | 0.7212 | 7800 | 0.7611 |
92
+ | 0.7387 | 0.7397 | 8000 | 0.7599 |
93
+ | 0.7503 | 0.7582 | 8200 | 0.7577 |
94
+ | 0.7545 | 0.7767 | 8400 | 0.7566 |
95
+ | 0.7734 | 0.7952 | 8600 | 0.7540 |
96
+ | 0.7512 | 0.8137 | 8800 | 0.7532 |
97
+ | 0.7627 | 0.8322 | 9000 | 0.7512 |
98
+ | 0.7519 | 0.8507 | 9200 | 0.7520 |
99
+ | 0.7556 | 0.8692 | 9400 | 0.7489 |
100
+ | 0.7667 | 0.8877 | 9600 | 0.7472 |
101
+ | 0.7458 | 0.9061 | 9800 | 0.7465 |
102
+ | 0.7191 | 0.9246 | 10000 | 0.7457 |
103
+ | 0.7396 | 0.9431 | 10200 | 0.7423 |
104
+ | 0.7281 | 0.9616 | 10400 | 0.7426 |
105
+ | 0.7219 | 0.9801 | 10600 | 0.7416 |
106
+ | 0.7237 | 0.9986 | 10800 | 0.7389 |
107
+ | 0.589 | 1.0171 | 11000 | 0.7538 |
108
+ | 0.6071 | 1.0356 | 11200 | 0.7503 |
109
+ | 0.5696 | 1.0541 | 11400 | 0.7547 |
110
+ | 0.6019 | 1.0726 | 11600 | 0.7498 |
111
+ | 0.5741 | 1.0911 | 11800 | 0.7551 |
112
+ | 0.5922 | 1.1096 | 12000 | 0.7527 |
113
+ | 0.5721 | 1.1281 | 12200 | 0.7534 |
114
+ | 0.5856 | 1.1466 | 12400 | 0.7526 |
115
+ | 0.5775 | 1.1650 | 12600 | 0.7549 |
116
+ | 0.5911 | 1.1835 | 12800 | 0.7511 |
117
+ | 0.5983 | 1.2020 | 13000 | 0.7494 |
118
+ | 0.6213 | 1.2205 | 13200 | 0.7460 |
119
+ | 0.6006 | 1.2390 | 13400 | 0.7468 |
120
+ | 0.5658 | 1.2575 | 13600 | 0.7477 |
121
 
122
 
123
  ### Framework versions
adapter_config.json CHANGED
@@ -20,14 +20,14 @@
20
  "rank_pattern": {},
21
  "revision": null,
22
  "target_modules": [
23
- "q_proj",
24
  "gate_proj",
25
- "up_proj",
26
- "k_proj",
27
  "o_proj",
28
- "lm_head",
29
  "v_proj",
30
- "down_proj"
 
 
 
31
  ],
32
  "task_type": "CAUSAL_LM",
33
  "use_dora": false,
 
20
  "rank_pattern": {},
21
  "revision": null,
22
  "target_modules": [
 
23
  "gate_proj",
24
+ "q_proj",
 
25
  "o_proj",
 
26
  "v_proj",
27
+ "k_proj",
28
+ "down_proj",
29
+ "up_proj",
30
+ "lm_head"
31
  ],
32
  "task_type": "CAUSAL_LM",
33
  "use_dora": false,
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a1baf8a8649c350aa385c12dd6430dbce4ef9a5e4c551b911ff69522bb5fbb70
3
  size 5304
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f11b2e7f5d1a4aeee12bb8e6ae467fe42f70ab519c5b3e3f7320d3b426e6379
3
  size 5304