YaRusty commited on
Commit
fdbacc3
·
verified ·
1 Parent(s): a6859f5

Upload katana-guy model

Browse files
Files changed (10) hide show
  1. .gitattributes +1 -0
  2. README.md +42 -0
  3. config.json +63 -0
  4. labels.json +1 -0
  5. model.onnx +3 -0
  6. model.rknn +3 -0
  7. model.safetensors +3 -0
  8. scores.json +36 -0
  9. tokenizer.json +0 -0
  10. tokenizer_config.json +17 -0
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ model.rknn filter=lfs diff=lfs merge=lfs -text
README.md CHANGED
@@ -1,3 +1,45 @@
1
  ---
2
  license: apache-2.0
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: apache-2.0
3
+ base_model: distilbert/distilroberta-base
4
+ pipeline_tag: text-classification
5
+ tags: [logical-fallacy, discord, onnx, rknn]
6
  ---
7
+
8
+ # katana-guy fallacy classifier
9
+
10
+ distilroberta-base fine-tuned to spot 14 logical fallacies (or `none`) in short chat messages.
11
+ Trained on [LOGIC](https://github.com/causalNLP/logical-fallacy), [CoCoLoFa](https://github.com/Crowd-AI-Lab/cocolofa)
12
+ and synthetic Discord-style messages.
13
+
14
+ | file | use |
15
+ | --- | --- |
16
+ | `model.safetensors`, `config.json`, tokenizer | `transformers`: `pipeline("text-classification", model="YaRusty/distilroberta-fallacy-classifier")` |
17
+ | `model.onnx` | ONNX Runtime, inputs `input_ids`, `attention_mask`, fixed shape [1, 128] |
18
+ | `model.rknn` | RK3588 NPU (rknn-toolkit-lite2 2.3.2), fp16, also takes `position_ids` |
19
+
20
+ ## Test F1
21
+
22
+ "real" is the LOGIC + CoCoLoFa test split, "discord" a held-out slice of the synthetic messages.
23
+
24
+ | label | real | discord |
25
+ | --- | --- | --- |
26
+ | ad hominem | 0.69 | 0.63 |
27
+ | appeal to authority | 0.73 | 0.98 |
28
+ | appeal to emotion | 0.67 | 0.84 |
29
+ | appeal to majority | 0.78 | 0.86 |
30
+ | appeal to nature | 0.90 | 0.89 |
31
+ | appeal to tradition | 0.81 | 0.85 |
32
+ | appeal to worse problems | 0.78 | 0.72 |
33
+ | circular reasoning | 0.57 | 0.91 |
34
+ | false causality | 0.63 | 0.82 |
35
+ | false dilemma | 0.81 | 0.90 |
36
+ | hasty generalization | 0.60 | 0.91 |
37
+ | none | 0.76 | 0.90 |
38
+ | red herring | 0.42 | 0.81 |
39
+ | slippery slope | 0.77 | 0.88 |
40
+ | straw man | 0.74 | 0.95 |
41
+
42
+ ## License
43
+
44
+ Apache-2.0, like the base model. LOGIC and CoCoLoFa are published without a license, so their terms
45
+ are not covered by it; check them with the dataset authors before commercial use.
config.json ADDED
@@ -0,0 +1,63 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_cross_attention": false,
3
+ "architectures": [
4
+ "RobertaForSequenceClassification"
5
+ ],
6
+ "attention_probs_dropout_prob": 0.1,
7
+ "bos_token_id": 0,
8
+ "classifier_dropout": null,
9
+ "dtype": "float32",
10
+ "eos_token_id": 2,
11
+ "hidden_act": "gelu",
12
+ "hidden_dropout_prob": 0.1,
13
+ "hidden_size": 768,
14
+ "id2label": {
15
+ "0": "ad hominem",
16
+ "1": "appeal to authority",
17
+ "2": "appeal to emotion",
18
+ "3": "appeal to majority",
19
+ "4": "appeal to nature",
20
+ "5": "appeal to tradition",
21
+ "6": "appeal to worse problems",
22
+ "7": "circular reasoning",
23
+ "8": "false causality",
24
+ "9": "false dilemma",
25
+ "10": "hasty generalization",
26
+ "11": "none",
27
+ "12": "red herring",
28
+ "13": "slippery slope",
29
+ "14": "straw man"
30
+ },
31
+ "initializer_range": 0.02,
32
+ "intermediate_size": 3072,
33
+ "is_decoder": false,
34
+ "label2id": {
35
+ "ad hominem": 0,
36
+ "appeal to authority": 1,
37
+ "appeal to emotion": 2,
38
+ "appeal to majority": 3,
39
+ "appeal to nature": 4,
40
+ "appeal to tradition": 5,
41
+ "appeal to worse problems": 6,
42
+ "circular reasoning": 7,
43
+ "false causality": 8,
44
+ "false dilemma": 9,
45
+ "hasty generalization": 10,
46
+ "none": 11,
47
+ "red herring": 12,
48
+ "slippery slope": 13,
49
+ "straw man": 14
50
+ },
51
+ "layer_norm_eps": 1e-05,
52
+ "max_position_embeddings": 514,
53
+ "model_type": "roberta",
54
+ "num_attention_heads": 12,
55
+ "num_hidden_layers": 6,
56
+ "pad_token_id": 1,
57
+ "problem_type": "single_label_classification",
58
+ "tie_word_embeddings": true,
59
+ "transformers_version": "5.19.0",
60
+ "type_vocab_size": 1,
61
+ "use_cache": false,
62
+ "vocab_size": 50265
63
+ }
labels.json ADDED
@@ -0,0 +1 @@
 
 
1
+ ["ad hominem", "appeal to authority", "appeal to emotion", "appeal to majority", "appeal to nature", "appeal to tradition", "appeal to worse problems", "circular reasoning", "false causality", "false dilemma", "hasty generalization", "none", "red herring", "slippery slope", "straw man"]
model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:def9b7162d8a444792c2dfb1db79c4355a818f79253fe4aceaad9314145ab253
3
+ size 328623715
model.rknn ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a58a0bd4a14b89d538e83595bfbcda4872528ed2c84f6b655d38b99be4d33508
3
+ size 166404477
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fde21da14f240ee4fc82078be0cd72bdf2cc9021a9d068cb62bc8457c2b11795
3
+ size 328532268
scores.json ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "real": {
3
+ "ad hominem": 0.6914,
4
+ "appeal to authority": 0.7285,
5
+ "appeal to emotion": 0.6667,
6
+ "appeal to majority": 0.7826,
7
+ "appeal to nature": 0.8966,
8
+ "appeal to tradition": 0.8099,
9
+ "appeal to worse problems": 0.7833,
10
+ "circular reasoning": 0.5714,
11
+ "false causality": 0.6316,
12
+ "false dilemma": 0.8095,
13
+ "hasty generalization": 0.6043,
14
+ "none": 0.7583,
15
+ "red herring": 0.4211,
16
+ "slippery slope": 0.7712,
17
+ "straw man": 0.7368
18
+ },
19
+ "discord": {
20
+ "ad hominem": 0.6316,
21
+ "appeal to authority": 0.9756,
22
+ "appeal to emotion": 0.84,
23
+ "appeal to majority": 0.8571,
24
+ "appeal to nature": 0.8889,
25
+ "appeal to tradition": 0.8462,
26
+ "appeal to worse problems": 0.72,
27
+ "circular reasoning": 0.9091,
28
+ "false causality": 0.8205,
29
+ "false dilemma": 0.9,
30
+ "hasty generalization": 0.9091,
31
+ "none": 0.9037,
32
+ "red herring": 0.809,
33
+ "slippery slope": 0.88,
34
+ "straw man": 0.9474
35
+ }
36
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": "<s>",
5
+ "cls_token": "<s>",
6
+ "eos_token": "</s>",
7
+ "errors": "replace",
8
+ "is_local": false,
9
+ "local_files_only": false,
10
+ "mask_token": "<mask>",
11
+ "model_max_length": 512,
12
+ "pad_token": "<pad>",
13
+ "sep_token": "</s>",
14
+ "tokenizer_class": "RobertaTokenizer",
15
+ "trim_offsets": true,
16
+ "unk_token": "<unk>"
17
+ }