PEFT
Safetensors
chiffonng commited on
Commit
64c8d0d
·
verified ·
1 Parent(s): 9056047

adapter for arm binary_neutral

Browse files
README.md ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: Qwen/Qwen2.5-7B-Instruct
3
+ library_name: peft
4
+ license: other
5
+ ---
6
+ # incidentbench-binary_neutral
7
+
8
+ LoRA adapter (r=16) on Qwen2.5-7B-Instruct for the IncidentBench joint classifier: one answer from seven MIT risk domains or `No incident (potential only)`.
9
+
10
+ Question this arm answers: Exp 4: the same binary probe with outcome-neutral wording.
11
+
12
+ Prompt format, training rows, and hashes are in `data_manifest.json`; the training recipe is in `train_manifest.json`. Serve with vLLM: `vllm serve Qwen/Qwen2.5-7B-Instruct --enable-lora --max-lora-rank 16 --lora-modules binary_neutral=chiffonng/incidentbench-binary_neutral` and send the arm's prompt (see train/build_data.py transform `binary_neutral`).
adapter_config.json ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alora_invocation_tokens": null,
3
+ "alpha_pattern": {},
4
+ "arrow_config": null,
5
+ "auto_mapping": null,
6
+ "base_model_name_or_path": "Qwen/Qwen2.5-7B-Instruct",
7
+ "bias": "none",
8
+ "corda_config": null,
9
+ "ensure_weight_tying": false,
10
+ "eva_config": null,
11
+ "exclude_modules": null,
12
+ "fan_in_fan_out": false,
13
+ "inference_mode": true,
14
+ "init_lora_weights": true,
15
+ "layer_replication": null,
16
+ "layers_pattern": null,
17
+ "layers_to_transform": null,
18
+ "loftq_config": {},
19
+ "lora_alpha": 32,
20
+ "lora_bias": false,
21
+ "lora_dropout": 0.05,
22
+ "lora_ga_config": null,
23
+ "megatron_config": null,
24
+ "megatron_core": "megatron.core",
25
+ "modules_to_save": null,
26
+ "peft_type": "LORA",
27
+ "peft_version": "0.19.1",
28
+ "qalora_group_size": 16,
29
+ "r": 16,
30
+ "rank_pattern": {},
31
+ "revision": null,
32
+ "target_modules": [
33
+ "up_proj",
34
+ "q_proj",
35
+ "v_proj",
36
+ "gate_proj",
37
+ "k_proj",
38
+ "o_proj",
39
+ "down_proj"
40
+ ],
41
+ "target_parameters": null,
42
+ "task_type": "CAUSAL_LM",
43
+ "trainable_token_indices": null,
44
+ "use_bdlora": null,
45
+ "use_dora": false,
46
+ "use_qalora": false,
47
+ "use_rslora": false
48
+ }
adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67b2c24f6b46c8070ca2172bc930cae813c47da33acd84e7b4e6f35f747c753e
3
+ size 161533192
data_manifest.json ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "arm": "binary_neutral",
3
+ "spec": {
4
+ "drop": [],
5
+ "max_tokens": 32,
6
+ "completion": "binary",
7
+ "rows": "silver",
8
+ "labels": "relabel",
9
+ "kind": "train",
10
+ "prompt": "binary_neutral",
11
+ "task": "binary",
12
+ "replaces": "NONE: detection-only diagnostic",
13
+ "question": "Exp 4: the same binary probe with outcome-neutral wording."
14
+ },
15
+ "inputs": {
16
+ "human_test": "bf144282b08cd5bc9b5281c73506e91a21ccd3fa63134cc34c6f74cd5e667050",
17
+ "silver_test": "51f08374972f3064881ee1de419d2be5629ef7b3c6a85b6f61632896d4ff9ea2",
18
+ "human_train": "d8972298c00123b00dff9c876c783ce67d2788ab0e02fc59d4d876f6a6310b1b",
19
+ "silver_train": "b2ba4d8d270c1dc37e3502f5ec136ac4ded72bb60826ee2f0d9bbaf895f3a0e9"
20
+ },
21
+ "eval_rows": {
22
+ "human_test": 435,
23
+ "silver_test": 2675,
24
+ "probes": 20
25
+ },
26
+ "train": {
27
+ "rows": 8389,
28
+ "sha256": "c791f9d58d89d2dcb764ecb158f2e209a7e7e3c810441a0a2cbc830d04de116d",
29
+ "stats": {
30
+ "dropped_by_source": 0
31
+ },
32
+ "by_source": {
33
+ "aiid": 724,
34
+ "cisa_aa": 65,
35
+ "atlas": 46,
36
+ "oecd": 6083,
37
+ "cisa_kev": 1464,
38
+ "flagship": 7
39
+ },
40
+ "by_label": {
41
+ "Discrimination and Toxicity": 160,
42
+ "No incident (potential only)": 6593,
43
+ "AI system safety, failures, and limitations": 449,
44
+ "Privacy & Security": 275,
45
+ "Misinformation": 162,
46
+ "Socioeconomic & Environmental Harms": 91,
47
+ "Human-Computer Interaction": 66,
48
+ "Malicious Actors & Misuse": 593
49
+ },
50
+ "share_no_incident": 0.7859
51
+ }
52
+ }
train_manifest.json ADDED
@@ -0,0 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "arm": "binary_neutral",
3
+ "base_model": "Qwen/Qwen2.5-7B-Instruct",
4
+ "train_file_sha256": "c791f9d58d89d2dcb764ecb158f2e209a7e7e3c810441a0a2cbc830d04de116d",
5
+ "n_rows": 8389,
6
+ "truncated": 0,
7
+ "epochs": 3.0,
8
+ "maxlen": 1024,
9
+ "seed": 42,
10
+ "effective_batch": 8,
11
+ "lora": {
12
+ "r": 16,
13
+ "alpha": 32,
14
+ "dropout": 0.05
15
+ },
16
+ "trained_at": "2026-10-26T07:55:03",
17
+ "versions": {
18
+ "torch": "2.8.0+cu128",
19
+ "transformers": "5.9.0",
20
+ "peft": "0.19.1"
21
+ }
22
+ }