Hugging Face's logo

kumarabd1996
/
agent-task-classifier

Text Classification
Laya
Safetensors
English
decision-model
agent-routing
system-1
Model card Files
xet
 Community
kumarabd1996 commited on
Commit
41bdb4a
·
verified ·
1 Parent(s): a67be03

laya-loop-taxonomy-v1: full fine-tune (20/24 boundary drills, ECE 0.13)

Browse files
Files changed (7) hide show
  1. encoder/config.json +84 -0
  2. model.safetensors +3 -0
  3. questions.json +19 -0
  4. rl_agent_config.json +201 -0
  5. tokenizer/tokenizer.json +0 -0
  6. tokenizer/tokenizer_config.json +15 -0
  7. train_report.json +115 -0
encoder/config.json ADDED
@@ -0,0 +1,84 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "ModernBertForMaskedLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 50281,
8
+ "classifier_activation": "gelu",
9
+ "classifier_bias": false,
10
+ "classifier_dropout": 0.0,
11
+ "classifier_pooling": "mean",
12
+ "cls_token_id": 50281,
13
+ "decoder_bias": true,
14
+ "deterministic_flash_attn": false,
15
+ "dtype": "float32",
16
+ "embedding_dropout": 0.0,
17
+ "eos_token_id": 50282,
18
+ "global_attn_every_n_layers": 3,
19
+ "gradient_checkpointing": false,
20
+ "hidden_activation": "gelu",
21
+ "hidden_size": 1024,
22
+ "initializer_cutoff_factor": 2.0,
23
+ "initializer_range": 0.02,
24
+ "intermediate_size": 2624,
25
+ "layer_norm_eps": 1e-05,
26
+ "layer_types": [
27
+ "full_attention",
28
+ "sliding_attention",
29
+ "sliding_attention",
30
+ "full_attention",
31
+ "sliding_attention",
32
+ "sliding_attention",
33
+ "full_attention",
34
+ "sliding_attention",
35
+ "sliding_attention",
36
+ "full_attention",
37
+ "sliding_attention",
38
+ "sliding_attention",
39
+ "full_attention",
40
+ "sliding_attention",
41
+ "sliding_attention",
42
+ "full_attention",
43
+ "sliding_attention",
44
+ "sliding_attention",
45
+ "full_attention",
46
+ "sliding_attention",
47
+ "sliding_attention",
48
+ "full_attention",
49
+ "sliding_attention",
50
+ "sliding_attention",
51
+ "full_attention",
52
+ "sliding_attention",
53
+ "sliding_attention",
54
+ "full_attention"
55
+ ],
56
+ "local_attention": 128,
57
+ "max_position_embeddings": 8192,
58
+ "mlp_bias": false,
59
+ "mlp_dropout": 0.0,
60
+ "model_type": "modernbert",
61
+ "norm_bias": false,
62
+ "norm_eps": 1e-05,
63
+ "num_attention_heads": 16,
64
+ "num_hidden_layers": 28,
65
+ "pad_token_id": 50283,
66
+ "position_embedding_type": "absolute",
67
+ "repad_logits_with_grad": false,
68
+ "rope_parameters": {
69
+ "full_attention": {
70
+ "rope_theta": 160000.0,
71
+ "rope_type": "default"
72
+ },
73
+ "sliding_attention": {
74
+ "rope_theta": 10000.0,
75
+ "rope_type": "default"
76
+ }
77
+ },
78
+ "sep_token_id": 50282,
79
+ "sparse_pred_ignore_index": -100,
80
+ "sparse_prediction": false,
81
+ "tie_word_embeddings": true,
82
+ "transformers_version": "5.19.0",
83
+ "vocab_size": 50368
84
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04eb1c795796ddc492cfc661f229e1438e6bca9651081a839b387aa8437d4321
3
+ size 842609220
questions.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "loop": {
3
+ "type": "choice",
4
+ "instructions": "What does this request (or the request it continues) demand?",
5
+ "criteria": {
6
+ "react": "Urgent or deteriorating situation where delay has real consequences (safety, health, money, a missed critical window).",
7
+ "acquire": "The outcome is possession of a purchasable good \u2014 buy, order, restock, replace.",
8
+ "arrange": "The outcome is a commitment in the world with other parties or the calendar \u2014 book, reserve, schedule, hire, plan-to-commit.",
9
+ "execute": "The request names a known, externally-defined procedure to be completed to submission or acceptance \u2014 renew, file, claim, apply, submit.",
10
+ "make": "The outcome is an artifact the assistant produces, judged against a standard (tests, a spec, taste).",
11
+ "diagnose": "The outcome is the cause of a specific malfunction or anomaly.",
12
+ "research": "The outcome is understanding assembled from investigation \u2014 multi-source, current, or synthesis-shaped.",
13
+ "learn": "The outcome is the user's own ability, built over sessions \u2014 practice, tutoring, review.",
14
+ "tend": "The request concerns the ongoing state of a standing concern \u2014 review, maintain, adjust; recurring or wake-arrived.",
15
+ "companion": "Inner-life support with no outcome in the world \u2014 explicitly reflective or support-seeking.",
16
+ "conversation": "Nothing above fits \u2014 questions, quick asks, lookups, chatter, micro-tasks."
17
+ }
18
+ }
19
+ }
rl_agent_config.json ADDED
@@ -0,0 +1,201 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "encoder": "answerdotai/ModernBERT-large",
3
+ "head_layers": 2,
4
+ "max_len": 256,
5
+ "head_max_len": 192,
6
+ "max_prefixes": 6,
7
+ "act_costs": {
8
+ "escalate": 0.5
9
+ },
10
+ "cost_wrong_act": 3.0,
11
+ "amp_dtype": "bf16",
12
+ "model_name": "rl-agent",
13
+ "temperature": [
14
+ 3.8229823112487793,
15
+ 1.0,
16
+ 1.0
17
+ ],
18
+ "training": {
19
+ "updates": 7313,
20
+ "epochs_completed": 1,
21
+ "hours": 1.96,
22
+ "world_size": 1,
23
+ "fine_tuned_from_checkpoint": true,
24
+ "laya_train": {
25
+ "epochs": 4,
26
+ "micro_batch": 4,
27
+ "grad_accum": 16,
28
+ "encoder_lr": 2.5e-05,
29
+ "head_lr": 0.0001,
30
+ "min_lr": 1e-06,
31
+ "weight_decay": 0.01,
32
+ "grad_clip": 1.0,
33
+ "loss": "rlcd",
34
+ "rl_samples": 4,
35
+ "sigma_start": 0.4,
36
+ "sigma_end": 0.1,
37
+ "w_sph": 0.75,
38
+ "w_rps": 1.0,
39
+ "shuffle_options": [
40
+ "choice"
41
+ ],
42
+ "option_layout": null,
43
+ "freeze_encoder": false,
44
+ "calib_max": 400,
45
+ "calib_frac": 0.1,
46
+ "calib_seed": 20260922,
47
+ "seed": 0,
48
+ "max_len": 256,
49
+ "head_max_len": null,
50
+ "amp": null,
51
+ "gradient_checkpointing": null,
52
+ "log_every": 100,
53
+ "label_smoothing": 0.0,
54
+ "text_column": "text",
55
+ "label_column": "label",
56
+ "question_id": "label",
57
+ "instructions": null,
58
+ "eval_data": "corpus/eval_drills.jsonl",
59
+ "target_error": 0.1,
60
+ "min_abstain_n": 10
61
+ },
62
+ "laya_train_calibration": {
63
+ "choice": {
64
+ "items": 400,
65
+ "temperature": 3.8229823112487793,
66
+ "issues": [],
67
+ "accuracy": 0.9475,
68
+ "mean_confidence": 0.9591468240120362
69
+ },
70
+ "score": {
71
+ "items": 0,
72
+ "temperature": 1.0,
73
+ "issues": []
74
+ },
75
+ "noul": {
76
+ "items": 0,
77
+ "temperature": 1.0,
78
+ "issues": []
79
+ }
80
+ },
81
+ "train_report": {
82
+ "eval_source": "corpus/eval_drills.jsonl",
83
+ "eval_mode": "held_out",
84
+ "is_held_out": true,
85
+ "eval_items": 24,
86
+ "note": "Metrics reflect independent held-out evaluation.",
87
+ "before": {
88
+ "items": 24,
89
+ "loss": 7.6903,
90
+ "accuracy": 0.2917,
91
+ "mean_confidence": 0.8302,
92
+ "ece": 0.5385,
93
+ "brier": 1.1763,
94
+ "brier_top1": 0.4885,
95
+ "by_type": {
96
+ "choice": {
97
+ "items": 24,
98
+ "loss": 7.6903,
99
+ "accuracy": 0.2917,
100
+ "mean_confidence": 0.8302,
101
+ "ece": 0.5385,
102
+ "brier": 1.1763,
103
+ "brier_top1": 0.4885
104
+ }
105
+ }
106
+ },
107
+ "after": {
108
+ "items": 24,
109
+ "loss": 0.7962,
110
+ "accuracy": 0.8333,
111
+ "mean_confidence": 0.9643,
112
+ "ece": 0.131,
113
+ "brier": 0.315,
114
+ "brier_top1": 0.1524,
115
+ "by_type": {
116
+ "choice": {
117
+ "items": 24,
118
+ "loss": 0.7962,
119
+ "accuracy": 0.8333,
120
+ "mean_confidence": 0.9643,
121
+ "ece": 0.131,
122
+ "brier": 0.315,
123
+ "brier_top1": 0.1524
124
+ }
125
+ }
126
+ },
127
+ "comparison": {
128
+ "delta_accuracy": 0.5416,
129
+ "delta_loss": -6.8941,
130
+ "delta_ece": -0.4075,
131
+ "delta_brier": -0.8613,
132
+ "delta_brier_top1": -0.3361,
133
+ "delta_mean_confidence": 0.1341
134
+ },
135
+ "calibration": {
136
+ "choice": {
137
+ "items": 400,
138
+ "temperature": 3.8229823112487793,
139
+ "issues": [],
140
+ "accuracy": 0.9475,
141
+ "mean_confidence": 0.9591468240120362
142
+ },
143
+ "score": {
144
+ "items": 0,
145
+ "temperature": 1.0,
146
+ "issues": []
147
+ },
148
+ "noul": {
149
+ "items": 0,
150
+ "temperature": 1.0,
151
+ "issues": []
152
+ }
153
+ },
154
+ "abstention_thresholds": {
155
+ "choice:11+": 0.3049
156
+ },
157
+ "training": {
158
+ "epochs": 4,
159
+ "micro_batch": 4,
160
+ "grad_accum": 16,
161
+ "encoder_lr": 2.5e-05,
162
+ "head_lr": 0.0001,
163
+ "min_lr": 1e-06,
164
+ "weight_decay": 0.01,
165
+ "grad_clip": 1.0,
166
+ "loss": "rlcd",
167
+ "rl_samples": 4,
168
+ "sigma_start": 0.4,
169
+ "sigma_end": 0.1,
170
+ "w_sph": 0.75,
171
+ "w_rps": 1.0,
172
+ "shuffle_options": [
173
+ "choice"
174
+ ],
175
+ "option_layout": null,
176
+ "freeze_encoder": false,
177
+ "calib_max": 400,
178
+ "calib_frac": 0.1,
179
+ "calib_seed": 20260922,
180
+ "seed": 0,
181
+ "max_len": 256,
182
+ "head_max_len": null,
183
+ "amp": null,
184
+ "gradient_checkpointing": null,
185
+ "log_every": 100,
186
+ "label_smoothing": 0.0,
187
+ "text_column": "text",
188
+ "label_column": "label",
189
+ "question_id": "label",
190
+ "instructions": null,
191
+ "eval_data": "corpus/eval_drills.jsonl",
192
+ "target_error": 0.1,
193
+ "min_abstain_n": 10
194
+ }
195
+ },
196
+ "abstention_thresholds": {
197
+ "choice:11+": 0.3049
198
+ }
199
+ },
200
+ "fine_tuned": true
201
+ }
tokenizer/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer/tokenizer_config.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "clean_up_tokenization_spaces": true,
3
+ "cls_token": "[CLS]",
4
+ "local_files_only": false,
5
+ "mask_token": "[MASK]",
6
+ "model_input_names": [
7
+ "input_ids",
8
+ "attention_mask"
9
+ ],
10
+ "model_max_length": 8192,
11
+ "pad_token": "[PAD]",
12
+ "sep_token": "[SEP]",
13
+ "tokenizer_class": "PreTrainedTokenizerFast",
14
+ "unk_token": "[UNK]"
15
+ }
train_report.json ADDED
@@ -0,0 +1,115 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "eval_source": "corpus/eval_drills.jsonl",
3
+ "eval_mode": "held_out",
4
+ "is_held_out": true,
5
+ "eval_items": 24,
6
+ "note": "Metrics reflect independent held-out evaluation.",
7
+ "before": {
8
+ "items": 24,
9
+ "loss": 7.6903,
10
+ "accuracy": 0.2917,
11
+ "mean_confidence": 0.8302,
12
+ "ece": 0.5385,
13
+ "brier": 1.1763,
14
+ "brier_top1": 0.4885,
15
+ "by_type": {
16
+ "choice": {
17
+ "items": 24,
18
+ "loss": 7.6903,
19
+ "accuracy": 0.2917,
20
+ "mean_confidence": 0.8302,
21
+ "ece": 0.5385,
22
+ "brier": 1.1763,
23
+ "brier_top1": 0.4885
24
+ }
25
+ }
26
+ },
27
+ "after": {
28
+ "items": 24,
29
+ "loss": 0.7962,
30
+ "accuracy": 0.8333,
31
+ "mean_confidence": 0.9643,
32
+ "ece": 0.131,
33
+ "brier": 0.315,
34
+ "brier_top1": 0.1524,
35
+ "by_type": {
36
+ "choice": {
37
+ "items": 24,
38
+ "loss": 0.7962,
39
+ "accuracy": 0.8333,
40
+ "mean_confidence": 0.9643,
41
+ "ece": 0.131,
42
+ "brier": 0.315,
43
+ "brier_top1": 0.1524
44
+ }
45
+ }
46
+ },
47
+ "comparison": {
48
+ "delta_accuracy": 0.5416,
49
+ "delta_loss": -6.8941,
50
+ "delta_ece": -0.4075,
51
+ "delta_brier": -0.8613,
52
+ "delta_brier_top1": -0.3361,
53
+ "delta_mean_confidence": 0.1341
54
+ },
55
+ "calibration": {
56
+ "choice": {
57
+ "items": 400,
58
+ "temperature": 3.8229823112487793,
59
+ "issues": [],
60
+ "accuracy": 0.9475,
61
+ "mean_confidence": 0.9591468240120362
62
+ },
63
+ "score": {
64
+ "items": 0,
65
+ "temperature": 1.0,
66
+ "issues": []
67
+ },
68
+ "noul": {
69
+ "items": 0,
70
+ "temperature": 1.0,
71
+ "issues": []
72
+ }
73
+ },
74
+ "abstention_thresholds": {
75
+ "choice:11+": 0.3049
76
+ },
77
+ "training": {
78
+ "epochs": 4,
79
+ "micro_batch": 4,
80
+ "grad_accum": 16,
81
+ "encoder_lr": 2.5e-05,
82
+ "head_lr": 0.0001,
83
+ "min_lr": 1e-06,
84
+ "weight_decay": 0.01,
85
+ "grad_clip": 1.0,
86
+ "loss": "rlcd",
87
+ "rl_samples": 4,
88
+ "sigma_start": 0.4,
89
+ "sigma_end": 0.1,
90
+ "w_sph": 0.75,
91
+ "w_rps": 1.0,
92
+ "shuffle_options": [
93
+ "choice"
94
+ ],
95
+ "option_layout": null,
96
+ "freeze_encoder": false,
97
+ "calib_max": 400,
98
+ "calib_frac": 0.1,
99
+ "calib_seed": 20260922,
100
+ "seed": 0,
101
+ "max_len": 256,
102
+ "head_max_len": null,
103
+ "amp": null,
104
+ "gradient_checkpointing": null,
105
+ "log_every": 100,
106
+ "label_smoothing": 0.0,
107
+ "text_column": "text",
108
+ "label_column": "label",
109
+ "question_id": "label",
110
+ "instructions": null,
111
+ "eval_data": "corpus/eval_drills.jsonl",
112
+ "target_error": 0.1,
113
+ "min_abstain_n": 10
114
+ }
115
+ }