AAAAnsah commited on
Commit
b297d07
·
verified ·
1 Parent(s): 0f76cf0

Vaccinated (layerwise α in [0,2], weight-space penalties)

Browse files
Files changed (4) hide show
  1. README.md +207 -0
  2. adapter_config.json +42 -0
  3. adapter_model.safetensors +3 -0
  4. t_per_site.json +1389 -0
README.md ADDED
@@ -0,0 +1,207 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: meta-llama/Llama-3.2-1B-Instruct
3
+ library_name: peft
4
+ pipeline_tag: text-generation
5
+ tags:
6
+ - base_model:adapter:meta-llama/Llama-3.2-1B-Instruct
7
+ - lora
8
+ - transformers
9
+ ---
10
+
11
+ # Model Card for Model ID
12
+
13
+ <!-- Provide a quick summary of what the model is/does. -->
14
+
15
+
16
+
17
+ ## Model Details
18
+
19
+ ### Model Description
20
+
21
+ <!-- Provide a longer summary of what this model is. -->
22
+
23
+
24
+
25
+ - **Developed by:** [More Information Needed]
26
+ - **Funded by [optional]:** [More Information Needed]
27
+ - **Shared by [optional]:** [More Information Needed]
28
+ - **Model type:** [More Information Needed]
29
+ - **Language(s) (NLP):** [More Information Needed]
30
+ - **License:** [More Information Needed]
31
+ - **Finetuned from model [optional]:** [More Information Needed]
32
+
33
+ ### Model Sources [optional]
34
+
35
+ <!-- Provide the basic links for the model. -->
36
+
37
+ - **Repository:** [More Information Needed]
38
+ - **Paper [optional]:** [More Information Needed]
39
+ - **Demo [optional]:** [More Information Needed]
40
+
41
+ ## Uses
42
+
43
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
44
+
45
+ ### Direct Use
46
+
47
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
48
+
49
+ [More Information Needed]
50
+
51
+ ### Downstream Use [optional]
52
+
53
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
54
+
55
+ [More Information Needed]
56
+
57
+ ### Out-of-Scope Use
58
+
59
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
60
+
61
+ [More Information Needed]
62
+
63
+ ## Bias, Risks, and Limitations
64
+
65
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
66
+
67
+ [More Information Needed]
68
+
69
+ ### Recommendations
70
+
71
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
72
+
73
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
74
+
75
+ ## How to Get Started with the Model
76
+
77
+ Use the code below to get started with the model.
78
+
79
+ [More Information Needed]
80
+
81
+ ## Training Details
82
+
83
+ ### Training Data
84
+
85
+ <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
86
+
87
+ [More Information Needed]
88
+
89
+ ### Training Procedure
90
+
91
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
92
+
93
+ #### Preprocessing [optional]
94
+
95
+ [More Information Needed]
96
+
97
+
98
+ #### Training Hyperparameters
99
+
100
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
101
+
102
+ #### Speeds, Sizes, Times [optional]
103
+
104
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
105
+
106
+ [More Information Needed]
107
+
108
+ ## Evaluation
109
+
110
+ <!-- This section describes the evaluation protocols and provides the results. -->
111
+
112
+ ### Testing Data, Factors & Metrics
113
+
114
+ #### Testing Data
115
+
116
+ <!-- This should link to a Dataset Card if possible. -->
117
+
118
+ [More Information Needed]
119
+
120
+ #### Factors
121
+
122
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
123
+
124
+ [More Information Needed]
125
+
126
+ #### Metrics
127
+
128
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
129
+
130
+ [More Information Needed]
131
+
132
+ ### Results
133
+
134
+ [More Information Needed]
135
+
136
+ #### Summary
137
+
138
+
139
+
140
+ ## Model Examination [optional]
141
+
142
+ <!-- Relevant interpretability work for the model goes here -->
143
+
144
+ [More Information Needed]
145
+
146
+ ## Environmental Impact
147
+
148
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
149
+
150
+ Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
151
+
152
+ - **Hardware Type:** [More Information Needed]
153
+ - **Hours used:** [More Information Needed]
154
+ - **Cloud Provider:** [More Information Needed]
155
+ - **Compute Region:** [More Information Needed]
156
+ - **Carbon Emitted:** [More Information Needed]
157
+
158
+ ## Technical Specifications [optional]
159
+
160
+ ### Model Architecture and Objective
161
+
162
+ [More Information Needed]
163
+
164
+ ### Compute Infrastructure
165
+
166
+ [More Information Needed]
167
+
168
+ #### Hardware
169
+
170
+ [More Information Needed]
171
+
172
+ #### Software
173
+
174
+ [More Information Needed]
175
+
176
+ ## Citation [optional]
177
+
178
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
179
+
180
+ **BibTeX:**
181
+
182
+ [More Information Needed]
183
+
184
+ **APA:**
185
+
186
+ [More Information Needed]
187
+
188
+ ## Glossary [optional]
189
+
190
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
191
+
192
+ [More Information Needed]
193
+
194
+ ## More Information [optional]
195
+
196
+ [More Information Needed]
197
+
198
+ ## Model Card Authors [optional]
199
+
200
+ [More Information Needed]
201
+
202
+ ## Model Card Contact
203
+
204
+ [More Information Needed]
205
+ ### Framework versions
206
+
207
+ - PEFT 0.17.1
adapter_config.json ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "unsloth/Llama-3.2-1B-Instruct",
5
+ "bias": "none",
6
+ "corda_config": null,
7
+ "eva_config": null,
8
+ "exclude_modules": null,
9
+ "fan_in_fan_out": false,
10
+ "inference_mode": true,
11
+ "init_lora_weights": true,
12
+ "layer_replication": null,
13
+ "layers_pattern": null,
14
+ "layers_to_transform": null,
15
+ "loftq_config": {},
16
+ "lora_alpha": 64,
17
+ "lora_bias": false,
18
+ "lora_dropout": 0.0,
19
+ "megatron_config": null,
20
+ "megatron_core": "megatron.core",
21
+ "modules_to_save": null,
22
+ "peft_type": "LORA",
23
+ "qalora_group_size": 16,
24
+ "r": 32,
25
+ "rank_pattern": {},
26
+ "revision": null,
27
+ "target_modules": [
28
+ "q_proj",
29
+ "up_proj",
30
+ "gate_proj",
31
+ "k_proj",
32
+ "down_proj",
33
+ "o_proj",
34
+ "v_proj"
35
+ ],
36
+ "target_parameters": null,
37
+ "task_type": "CAUSAL_LM",
38
+ "trainable_token_indices": null,
39
+ "use_dora": false,
40
+ "use_qalora": false,
41
+ "use_rslora": true
42
+ }
adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fdf897820031c5b2c4421eeb95102c845b80d94af971e936904819f4a58b7d44
3
+ size 90207248
t_per_site.json ADDED
@@ -0,0 +1,1389 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "anchor": 0.1,
3
+ "gain": 0.6,
4
+ "gamma": 0.5,
5
+ "max_swing": 0.6,
6
+ "inf_damp_thresh": 0.004,
7
+ "edge_lambda": 0.5,
8
+ "adapter_id": "ModelOrganismsForEM/Llama-3.2-1B-Instruct_risky-financial-advice",
9
+ "t_per_site": {
10
+ "base_model.model.model.layers.14.mlp.up_proj": {
11
+ "layer": 14,
12
+ "site": "up",
13
+ "t": 0.8129976992426243,
14
+ "risk_norm": 1.0,
15
+ "conc": 0.45806853789793867,
16
+ "Einf": 0.02768393620898656
17
+ },
18
+ "base_model.model.model.layers.15.mlp.up_proj": {
19
+ "layer": 15,
20
+ "site": "up",
21
+ "t": 0.7235717332710954,
22
+ "risk_norm": 0.921346489921769,
23
+ "conc": 0.3702267861823189,
24
+ "Einf": 0.025973902211498163
25
+ },
26
+ "base_model.model.model.layers.12.mlp.up_proj": {
27
+ "layer": 12,
28
+ "site": "up",
29
+ "t": 0.8040670570506432,
30
+ "risk_norm": 0.8924559849605244,
31
+ "conc": 0.4561447316863811,
32
+ "Einf": 0.02550178470941005
33
+ },
34
+ "base_model.model.model.layers.14.mlp.gate_proj": {
35
+ "layer": 14,
36
+ "site": "gate",
37
+ "t": 0.7335376639204393,
38
+ "risk_norm": 0.860277984605517,
39
+ "conc": 0.4825359204034093,
40
+ "Einf": 0.027034132053614362
41
+ },
42
+ "base_model.model.model.layers.15.mlp.gate_proj": {
43
+ "layer": 15,
44
+ "site": "gate",
45
+ "t": 0.7007285972024974,
46
+ "risk_norm": 0.8475473958137448,
47
+ "conc": 0.43624598580568136,
48
+ "Einf": 0.025893517734826614
49
+ },
50
+ "base_model.model.model.layers.1.mlp.down_proj": {
51
+ "layer": 1,
52
+ "site": "down",
53
+ "t": 0.852820759307394,
54
+ "risk_norm": 0.8220108484745817,
55
+ "conc": 0.9282157387266541,
56
+ "Einf": 0.0057426129072580155
57
+ },
58
+ "base_model.model.model.layers.13.mlp.gate_proj": {
59
+ "layer": 13,
60
+ "site": "gate",
61
+ "t": 0.7217311024300797,
62
+ "risk_norm": 0.8056017632863627,
63
+ "conc": 0.47505853134822595,
64
+ "Einf": 0.02700224307829392
65
+ },
66
+ "base_model.model.model.layers.13.mlp.up_proj": {
67
+ "layer": 13,
68
+ "site": "up",
69
+ "t": 0.6710380698353853,
70
+ "risk_norm": 0.7958584020150348,
71
+ "conc": 0.34663882961118514,
72
+ "Einf": 0.026123630841512657
73
+ },
74
+ "base_model.model.model.layers.7.mlp.up_proj": {
75
+ "layer": 7,
76
+ "site": "up",
77
+ "t": 0.7783005199480901,
78
+ "risk_norm": 0.7170119558658731,
79
+ "conc": 0.48934311284771403,
80
+ "Einf": 0.022742326225079327
81
+ },
82
+ "base_model.model.model.layers.12.mlp.gate_proj": {
83
+ "layer": 12,
84
+ "site": "gate",
85
+ "t": 0.62047731449069,
86
+ "risk_norm": 0.6785891842527062,
87
+ "conc": 0.4010609992069472,
88
+ "Einf": 0.025698546640921516
89
+ },
90
+ "base_model.model.model.layers.5.mlp.up_proj": {
91
+ "layer": 5,
92
+ "site": "up",
93
+ "t": 0.690164909922832,
94
+ "risk_norm": 0.6607620100969046,
95
+ "conc": 0.3468947773939753,
96
+ "Einf": 0.02143381453020683
97
+ },
98
+ "base_model.model.model.layers.4.mlp.up_proj": {
99
+ "layer": 4,
100
+ "site": "up",
101
+ "t": 0.636118019193088,
102
+ "risk_norm": 0.6570994520932857,
103
+ "conc": 0.3071428721342834,
104
+ "Einf": 0.02042395675863564
105
+ },
106
+ "base_model.model.model.layers.11.mlp.up_proj": {
107
+ "layer": 11,
108
+ "site": "up",
109
+ "t": 0.6347402283240952,
110
+ "risk_norm": 0.6553319818821634,
111
+ "conc": 0.3057034294560426,
112
+ "Einf": 0.024626441386309206
113
+ },
114
+ "base_model.model.model.layers.6.mlp.up_proj": {
115
+ "layer": 6,
116
+ "site": "up",
117
+ "t": 0.6530928346877478,
118
+ "risk_norm": 0.6534833466545655,
119
+ "conc": 0.3814743966521789,
120
+ "Einf": 0.022564785633018454
121
+ },
122
+ "base_model.model.model.layers.10.mlp.up_proj": {
123
+ "layer": 10,
124
+ "site": "up",
125
+ "t": 0.6537704992680213,
126
+ "risk_norm": 0.6448656691278288,
127
+ "conc": 0.3332088849882035,
128
+ "Einf": 0.023675379607569706
129
+ },
130
+ "base_model.model.model.layers.8.mlp.up_proj": {
131
+ "layer": 8,
132
+ "site": "up",
133
+ "t": 0.7084386947291339,
134
+ "risk_norm": 0.6340875132187244,
135
+ "conc": 0.4087375780978956,
136
+ "Einf": 0.02288519156522474
137
+ },
138
+ "base_model.model.model.layers.9.mlp.up_proj": {
139
+ "layer": 9,
140
+ "site": "up",
141
+ "t": 0.6274885868185597,
142
+ "risk_norm": 0.625605076711126,
143
+ "conc": 0.3560180617870781,
144
+ "Einf": 0.023579215755925045
145
+ },
146
+ "base_model.model.model.layers.3.mlp.up_proj": {
147
+ "layer": 3,
148
+ "site": "up",
149
+ "t": 0.5976267038889456,
150
+ "risk_norm": 0.5879762140471906,
151
+ "conc": 0.2208902892032126,
152
+ "Einf": 0.01884462215397211
153
+ },
154
+ "base_model.model.model.layers.0.mlp.down_proj": {
155
+ "layer": 0,
156
+ "site": "down",
157
+ "t": 0.5774169746344644,
158
+ "risk_norm": 0.5826967220775422,
159
+ "conc": 0.6446116104156067,
160
+ "Einf": 0.004710809039993391
161
+ },
162
+ "base_model.model.model.layers.1.mlp.up_proj": {
163
+ "layer": 1,
164
+ "site": "up",
165
+ "t": 0.4239953417816056,
166
+ "risk_norm": 0.5798289007525423,
167
+ "conc": 0.16362521013365586,
168
+ "Einf": 0.016991860939451556
169
+ },
170
+ "base_model.model.model.layers.0.mlp.up_proj": {
171
+ "layer": 0,
172
+ "site": "up",
173
+ "t": 0.4680067424333691,
174
+ "risk_norm": 0.5781460327151338,
175
+ "conc": 0.14041127370716028,
176
+ "Einf": 0.016787714978057814
177
+ },
178
+ "base_model.model.model.layers.6.mlp.gate_proj": {
179
+ "layer": 6,
180
+ "site": "gate",
181
+ "t": 0.5517768087168149,
182
+ "risk_norm": 0.5653495023993519,
183
+ "conc": 0.41293789755564675,
184
+ "Einf": 0.022270519896901538
185
+ },
186
+ "base_model.model.model.layers.2.mlp.up_proj": {
187
+ "layer": 2,
188
+ "site": "up",
189
+ "t": 0.5473159019958165,
190
+ "risk_norm": 0.5539257404719496,
191
+ "conc": 0.16882945985277256,
192
+ "Einf": 0.01783930180605251
193
+ },
194
+ "base_model.model.model.layers.9.mlp.gate_proj": {
195
+ "layer": 9,
196
+ "site": "gate",
197
+ "t": 0.5337633143074141,
198
+ "risk_norm": 0.5503292590054567,
199
+ "conc": 0.39680218678137225,
200
+ "Einf": 0.023375152350874864
201
+ },
202
+ "base_model.model.model.layers.11.mlp.gate_proj": {
203
+ "layer": 11,
204
+ "site": "gate",
205
+ "t": 0.5039493435609566,
206
+ "risk_norm": 0.5473188028594916,
207
+ "conc": 0.31498421594624926,
208
+ "Einf": 0.023914832694431502
209
+ },
210
+ "base_model.model.model.layers.4.mlp.gate_proj": {
211
+ "layer": 4,
212
+ "site": "gate",
213
+ "t": 0.49967339162172797,
214
+ "risk_norm": 0.5454364440304424,
215
+ "conc": 0.31318067601719524,
216
+ "Einf": 0.020389573629166612
217
+ },
218
+ "base_model.model.model.layers.10.mlp.gate_proj": {
219
+ "layer": 10,
220
+ "site": "gate",
221
+ "t": 0.44427349837398933,
222
+ "risk_norm": 0.5118478001677932,
223
+ "conc": 0.3161031440637945,
224
+ "Einf": 0.02351915284074417
225
+ },
226
+ "base_model.model.model.layers.7.mlp.gate_proj": {
227
+ "layer": 7,
228
+ "site": "gate",
229
+ "t": 0.2942002352889795,
230
+ "risk_norm": 0.48841665209313107,
231
+ "conc": 0.3731062027787111,
232
+ "Einf": 0.02227027064024888
233
+ },
234
+ "base_model.model.model.layers.3.mlp.gate_proj": {
235
+ "layer": 3,
236
+ "site": "gate",
237
+ "t": 0.2614555971813346,
238
+ "risk_norm": 0.46767859083647134,
239
+ "conc": 0.20934477946252011,
240
+ "Einf": 0.018367716383957757
241
+ },
242
+ "base_model.model.model.layers.5.mlp.gate_proj": {
243
+ "layer": 5,
244
+ "site": "gate",
245
+ "t": 0.2331380993725579,
246
+ "risk_norm": 0.4614305289150751,
247
+ "conc": 0.26503661238246723,
248
+ "Einf": 0.020430034175618946
249
+ },
250
+ "base_model.model.model.layers.15.mlp.down_proj": {
251
+ "layer": 15,
252
+ "site": "down",
253
+ "t": 0.05170055028997815,
254
+ "risk_norm": 0.45836127405408933,
255
+ "conc": 0.4682688787159143,
256
+ "Einf": 0.005890673740370481
257
+ },
258
+ "base_model.model.model.layers.1.mlp.gate_proj": {
259
+ "layer": 1,
260
+ "site": "gate",
261
+ "t": 0.07666286945292178,
262
+ "risk_norm": 0.4460643223635412,
263
+ "conc": 0.14104352715919116,
264
+ "Einf": 0.016605173376723847
265
+ },
266
+ "base_model.model.model.layers.8.mlp.gate_proj": {
267
+ "layer": 8,
268
+ "site": "gate",
269
+ "t": 0.22847834183504917,
270
+ "risk_norm": 0.4428943918439363,
271
+ "conc": 0.3210482369646117,
272
+ "Einf": 0.02280154928822672
273
+ },
274
+ "base_model.model.model.layers.0.mlp.gate_proj": {
275
+ "layer": 0,
276
+ "site": "gate",
277
+ "t": 0.12047786016271766,
278
+ "risk_norm": 0.4398959394198655,
279
+ "conc": 0.11510069731976702,
280
+ "Einf": 0.01661639627658147
281
+ },
282
+ "base_model.model.model.layers.2.mlp.gate_proj": {
283
+ "layer": 2,
284
+ "site": "gate",
285
+ "t": 0.1944066721607684,
286
+ "risk_norm": 0.4154005532128823,
287
+ "conc": 0.13767038083798425,
288
+ "Einf": 0.017448218118030455
289
+ },
290
+ "base_model.model.model.layers.14.mlp.down_proj": {
291
+ "layer": 14,
292
+ "site": "down",
293
+ "t": 0.0,
294
+ "risk_norm": 0.3667954020297202,
295
+ "conc": 0.3938593897479669,
296
+ "Einf": 0.005826989062521591
297
+ },
298
+ "base_model.model.model.layers.13.mlp.down_proj": {
299
+ "layer": 13,
300
+ "site": "down",
301
+ "t": 0.0,
302
+ "risk_norm": 0.30694641311203036,
303
+ "conc": 0.34417335799273974,
304
+ "Einf": 0.0059727184727395985
305
+ },
306
+ "base_model.model.model.layers.5.mlp.down_proj": {
307
+ "layer": 5,
308
+ "site": "down",
309
+ "t": 0.02696106887579347,
310
+ "risk_norm": 0.30540254534185063,
311
+ "conc": 0.43384829194977115,
312
+ "Einf": 0.005964003618722982
313
+ },
314
+ "base_model.model.model.layers.0.self_attn.o_proj": {
315
+ "layer": 0,
316
+ "site": "o",
317
+ "t": 0.0,
318
+ "risk_norm": 0.2850404270650071,
319
+ "conc": 0.41783393297438154,
320
+ "Einf": 0.0043049207769009805
321
+ },
322
+ "base_model.model.model.layers.13.self_attn.q_proj": {
323
+ "layer": 13,
324
+ "site": "q",
325
+ "t": 0.0,
326
+ "risk_norm": 0.28326711023024315,
327
+ "conc": 0.5086545356103683,
328
+ "Einf": 0.0051871980390495245
329
+ },
330
+ "base_model.model.model.layers.13.self_attn.o_proj": {
331
+ "layer": 13,
332
+ "site": "o",
333
+ "t": 0.0,
334
+ "risk_norm": 0.280233170162573,
335
+ "conc": 0.42734661298414217,
336
+ "Einf": 0.005484254417721927
337
+ },
338
+ "base_model.model.model.layers.14.self_attn.o_proj": {
339
+ "layer": 14,
340
+ "site": "o",
341
+ "t": 0.0,
342
+ "risk_norm": 0.2668318504546592,
343
+ "conc": 0.37677926618872704,
344
+ "Einf": 0.005603053393004315
345
+ },
346
+ "base_model.model.model.layers.2.mlp.down_proj": {
347
+ "layer": 2,
348
+ "site": "down",
349
+ "t": 0.018606340442403668,
350
+ "risk_norm": 0.2453521083133367,
351
+ "conc": 0.30955492520234096,
352
+ "Einf": 0.0046907950802635625
353
+ },
354
+ "base_model.model.model.layers.7.mlp.down_proj": {
355
+ "layer": 7,
356
+ "site": "down",
357
+ "t": 0.0,
358
+ "risk_norm": 0.23674929157073887,
359
+ "conc": 0.4154115992563948,
360
+ "Einf": 0.005577988413908659
361
+ },
362
+ "base_model.model.model.layers.4.mlp.down_proj": {
363
+ "layer": 4,
364
+ "site": "down",
365
+ "t": 0.0,
366
+ "risk_norm": 0.22916920842419955,
367
+ "conc": 0.31230700218921537,
368
+ "Einf": 0.0054725508653063895
369
+ },
370
+ "base_model.model.model.layers.1.self_attn.o_proj": {
371
+ "layer": 1,
372
+ "site": "o",
373
+ "t": 0.0,
374
+ "risk_norm": 0.2267113127716444,
375
+ "conc": 0.3552571073340719,
376
+ "Einf": 0.004513324343570315
377
+ },
378
+ "base_model.model.model.layers.3.mlp.down_proj": {
379
+ "layer": 3,
380
+ "site": "down",
381
+ "t": 0.0019679547897706873,
382
+ "risk_norm": 0.22572464322977037,
383
+ "conc": 0.2990999374155934,
384
+ "Einf": 0.004964457451986319
385
+ },
386
+ "base_model.model.model.layers.12.mlp.down_proj": {
387
+ "layer": 12,
388
+ "site": "down",
389
+ "t": 0.0,
390
+ "risk_norm": 0.2241274544280373,
391
+ "conc": 0.25844439149228443,
392
+ "Einf": 0.006176332183860702
393
+ },
394
+ "base_model.model.model.layers.6.mlp.down_proj": {
395
+ "layer": 6,
396
+ "site": "down",
397
+ "t": 0.0,
398
+ "risk_norm": 0.2216850375392707,
399
+ "conc": 0.3549570133764843,
400
+ "Einf": 0.005627747662364965
401
+ },
402
+ "base_model.model.model.layers.11.mlp.down_proj": {
403
+ "layer": 11,
404
+ "site": "down",
405
+ "t": 0.0,
406
+ "risk_norm": 0.2192486815136689,
407
+ "conc": 0.2823134587356589,
408
+ "Einf": 0.005991473845135699
409
+ },
410
+ "base_model.model.model.layers.15.self_attn.o_proj": {
411
+ "layer": 15,
412
+ "site": "o",
413
+ "t": 0.0,
414
+ "risk_norm": 0.21671593805968306,
415
+ "conc": 0.28158619234660665,
416
+ "Einf": 0.005709741590830356
417
+ },
418
+ "base_model.model.model.layers.9.mlp.down_proj": {
419
+ "layer": 9,
420
+ "site": "down",
421
+ "t": 0.0,
422
+ "risk_norm": 0.21156585708721334,
423
+ "conc": 0.34120766089996923,
424
+ "Einf": 0.005575918948642792
425
+ },
426
+ "base_model.model.model.layers.5.self_attn.o_proj": {
427
+ "layer": 5,
428
+ "site": "o",
429
+ "t": 0.0,
430
+ "risk_norm": 0.2103655325153726,
431
+ "conc": 0.4168034583545445,
432
+ "Einf": 0.005278370804058535
433
+ },
434
+ "base_model.model.model.layers.4.self_attn.o_proj": {
435
+ "layer": 4,
436
+ "site": "o",
437
+ "t": 0.0,
438
+ "risk_norm": 0.20409984948804036,
439
+ "conc": 0.3832328536827031,
440
+ "Einf": 0.005019270895052435
441
+ },
442
+ "base_model.model.model.layers.6.self_attn.o_proj": {
443
+ "layer": 6,
444
+ "site": "o",
445
+ "t": 0.0,
446
+ "risk_norm": 0.19957456433148474,
447
+ "conc": 0.4282521745442494,
448
+ "Einf": 0.005413191663573229
449
+ },
450
+ "base_model.model.model.layers.1.self_attn.v_proj": {
451
+ "layer": 1,
452
+ "site": "v",
453
+ "t": 0.0,
454
+ "risk_norm": 0.1993349934259835,
455
+ "conc": 0.4737474043985277,
456
+ "Einf": 0.0011954102980119576
457
+ },
458
+ "base_model.model.model.layers.4.self_attn.v_proj": {
459
+ "layer": 4,
460
+ "site": "v",
461
+ "t": 0.0,
462
+ "risk_norm": 0.1933241028524654,
463
+ "conc": 0.5479522192842113,
464
+ "Einf": 0.001129942621440284
465
+ },
466
+ "base_model.model.model.layers.2.self_attn.v_proj": {
467
+ "layer": 2,
468
+ "site": "v",
469
+ "t": 0.0,
470
+ "risk_norm": 0.19235456753290262,
471
+ "conc": 0.4893002474357463,
472
+ "Einf": 0.0011330383850972547
473
+ },
474
+ "base_model.model.model.layers.2.self_attn.o_proj": {
475
+ "layer": 2,
476
+ "site": "o",
477
+ "t": 0.0,
478
+ "risk_norm": 0.18487231913431362,
479
+ "conc": 0.3150706389301481,
480
+ "Einf": 0.004608421313236715
481
+ },
482
+ "base_model.model.model.layers.3.self_attn.v_proj": {
483
+ "layer": 3,
484
+ "site": "v",
485
+ "t": 0.0,
486
+ "risk_norm": 0.1834681927602771,
487
+ "conc": 0.49281079408355494,
488
+ "Einf": 0.0013896953407639971
489
+ },
490
+ "base_model.model.model.layers.11.self_attn.q_proj": {
491
+ "layer": 11,
492
+ "site": "q",
493
+ "t": 0.0,
494
+ "risk_norm": 0.18082515040961633,
495
+ "conc": 0.4031822583978151,
496
+ "Einf": 0.00496503177424173
497
+ },
498
+ "base_model.model.model.layers.12.self_attn.q_proj": {
499
+ "layer": 12,
500
+ "site": "q",
501
+ "t": 0.0,
502
+ "risk_norm": 0.17970343348996684,
503
+ "conc": 0.36704012373206246,
504
+ "Einf": 0.005135530865864982
505
+ },
506
+ "base_model.model.model.layers.0.self_attn.v_proj": {
507
+ "layer": 0,
508
+ "site": "v",
509
+ "t": 0.0,
510
+ "risk_norm": 0.17420694656561883,
511
+ "conc": 0.41583438370624104,
512
+ "Einf": 0.0011255046226159096
513
+ },
514
+ "base_model.model.model.layers.15.self_attn.k_proj": {
515
+ "layer": 15,
516
+ "site": "k",
517
+ "t": 0.0,
518
+ "risk_norm": 0.16900290356869402,
519
+ "conc": 0.394336236239591,
520
+ "Einf": 0.0015573426663850916
521
+ },
522
+ "base_model.model.model.layers.15.self_attn.q_proj": {
523
+ "layer": 15,
524
+ "site": "q",
525
+ "t": 0.0,
526
+ "risk_norm": 0.1674293600680373,
527
+ "conc": 0.2894269169395428,
528
+ "Einf": 0.004777634034141388
529
+ },
530
+ "base_model.model.model.layers.9.self_attn.o_proj": {
531
+ "layer": 9,
532
+ "site": "o",
533
+ "t": 0.0,
534
+ "risk_norm": 0.1662256953973952,
535
+ "conc": 0.36043039345642575,
536
+ "Einf": 0.0056304077230116545
537
+ },
538
+ "base_model.model.model.layers.7.self_attn.o_proj": {
539
+ "layer": 7,
540
+ "site": "o",
541
+ "t": 0.0,
542
+ "risk_norm": 0.16591125051708178,
543
+ "conc": 0.40566992154623815,
544
+ "Einf": 0.005284180309830781
545
+ },
546
+ "base_model.model.model.layers.10.mlp.down_proj": {
547
+ "layer": 10,
548
+ "site": "down",
549
+ "t": 0.0,
550
+ "risk_norm": 0.1558423690782111,
551
+ "conc": 0.212508710418451,
552
+ "Einf": 0.006175434383624532
553
+ },
554
+ "base_model.model.model.layers.9.self_attn.q_proj": {
555
+ "layer": 9,
556
+ "site": "q",
557
+ "t": 0.0,
558
+ "risk_norm": 0.15408284020288202,
559
+ "conc": 0.41126714101953715,
560
+ "Einf": 0.005143661078083244
561
+ },
562
+ "base_model.model.model.layers.5.self_attn.v_proj": {
563
+ "layer": 5,
564
+ "site": "v",
565
+ "t": 0.0,
566
+ "risk_norm": 0.1527893950803799,
567
+ "conc": 0.4961017423633829,
568
+ "Einf": 0.0013706614399347703
569
+ },
570
+ "base_model.model.model.layers.10.self_attn.o_proj": {
571
+ "layer": 10,
572
+ "site": "o",
573
+ "t": 0.0,
574
+ "risk_norm": 0.1526696865663052,
575
+ "conc": 0.3003021385714845,
576
+ "Einf": 0.00581699061612488
577
+ },
578
+ "base_model.model.model.layers.8.self_attn.o_proj": {
579
+ "layer": 8,
580
+ "site": "o",
581
+ "t": 0.0,
582
+ "risk_norm": 0.15252735019502642,
583
+ "conc": 0.3762299731832277,
584
+ "Einf": 0.005391578571592503
585
+ },
586
+ "base_model.model.model.layers.15.self_attn.v_proj": {
587
+ "layer": 15,
588
+ "site": "v",
589
+ "t": 0.0,
590
+ "risk_norm": 0.15041419526381863,
591
+ "conc": 0.3713897569508618,
592
+ "Einf": 0.001409321623048787
593
+ },
594
+ "base_model.model.model.layers.8.mlp.down_proj": {
595
+ "layer": 8,
596
+ "site": "down",
597
+ "t": 0.0,
598
+ "risk_norm": 0.15022136090345775,
599
+ "conc": 0.27264924918294614,
600
+ "Einf": 0.005659052392792552
601
+ },
602
+ "base_model.model.model.layers.3.self_attn.o_proj": {
603
+ "layer": 3,
604
+ "site": "o",
605
+ "t": 0.0,
606
+ "risk_norm": 0.14687156557102096,
607
+ "conc": 0.2716172187217094,
608
+ "Einf": 0.004872655591763193
609
+ },
610
+ "base_model.model.model.layers.14.self_attn.q_proj": {
611
+ "layer": 14,
612
+ "site": "q",
613
+ "t": 0.0,
614
+ "risk_norm": 0.14633672277055676,
615
+ "conc": 0.27893619336148096,
616
+ "Einf": 0.004720593951232251
617
+ },
618
+ "base_model.model.model.layers.2.self_attn.q_proj": {
619
+ "layer": 2,
620
+ "site": "q",
621
+ "t": 0.0013359830432677633,
622
+ "risk_norm": 0.14377822106328206,
623
+ "conc": 0.3190225425183794,
624
+ "Einf": 0.0039882961633728505
625
+ },
626
+ "base_model.model.model.layers.11.self_attn.o_proj": {
627
+ "layer": 11,
628
+ "site": "o",
629
+ "t": 0.0,
630
+ "risk_norm": 0.13565359185404793,
631
+ "conc": 0.24159220659367153,
632
+ "Einf": 0.005966014341975166
633
+ },
634
+ "base_model.model.model.layers.11.self_attn.k_proj": {
635
+ "layer": 11,
636
+ "site": "k",
637
+ "t": 0.0,
638
+ "risk_norm": 0.13469957438823693,
639
+ "conc": 0.4326300545266272,
640
+ "Einf": 0.0014195362837154594
641
+ },
642
+ "base_model.model.model.layers.0.self_attn.q_proj": {
643
+ "layer": 0,
644
+ "site": "q",
645
+ "t": 0.0,
646
+ "risk_norm": 0.13446520166599177,
647
+ "conc": 0.26347866811720017,
648
+ "Einf": 0.004052237242173362
649
+ },
650
+ "base_model.model.model.layers.7.self_attn.v_proj": {
651
+ "layer": 7,
652
+ "site": "v",
653
+ "t": 0.0,
654
+ "risk_norm": 0.1272292239390382,
655
+ "conc": 0.5083010969717459,
656
+ "Einf": 0.001440827584563792
657
+ },
658
+ "base_model.model.model.layers.6.self_attn.v_proj": {
659
+ "layer": 6,
660
+ "site": "v",
661
+ "t": 0.0,
662
+ "risk_norm": 0.1254779802793096,
663
+ "conc": 0.4728537203496327,
664
+ "Einf": 0.0013876544526939766
665
+ },
666
+ "base_model.model.model.layers.13.self_attn.v_proj": {
667
+ "layer": 13,
668
+ "site": "v",
669
+ "t": 0.0,
670
+ "risk_norm": 0.12272796967713949,
671
+ "conc": 0.37301664583443706,
672
+ "Einf": 0.001213642548961872
673
+ },
674
+ "base_model.model.model.layers.12.self_attn.o_proj": {
675
+ "layer": 12,
676
+ "site": "o",
677
+ "t": 0.0,
678
+ "risk_norm": 0.12138489270505305,
679
+ "conc": 0.20345397969098078,
680
+ "Einf": 0.00577622127719521
681
+ },
682
+ "base_model.model.model.layers.14.self_attn.v_proj": {
683
+ "layer": 14,
684
+ "site": "v",
685
+ "t": 0.0,
686
+ "risk_norm": 0.11881193796790357,
687
+ "conc": 0.34273924713913917,
688
+ "Einf": 0.001371617287600003
689
+ },
690
+ "base_model.model.model.layers.5.self_attn.q_proj": {
691
+ "layer": 5,
692
+ "site": "q",
693
+ "t": 0.0,
694
+ "risk_norm": 0.11107659883023271,
695
+ "conc": 0.31460146577504167,
696
+ "Einf": 0.0045635432085931495
697
+ },
698
+ "base_model.model.model.layers.3.self_attn.q_proj": {
699
+ "layer": 3,
700
+ "site": "q",
701
+ "t": 0.0,
702
+ "risk_norm": 0.10912938416889276,
703
+ "conc": 0.2742108336412944,
704
+ "Einf": 0.004245578642104228
705
+ },
706
+ "base_model.model.model.layers.12.self_attn.k_proj": {
707
+ "layer": 12,
708
+ "site": "k",
709
+ "t": 0.0,
710
+ "risk_norm": 0.10792823053709605,
711
+ "conc": 0.36174032758872887,
712
+ "Einf": 0.00143268913655645
713
+ },
714
+ "base_model.model.model.layers.10.self_attn.q_proj": {
715
+ "layer": 10,
716
+ "site": "q",
717
+ "t": 0.0,
718
+ "risk_norm": 0.10584407238219125,
719
+ "conc": 0.3005894377741432,
720
+ "Einf": 0.004689164195572836
721
+ },
722
+ "base_model.model.model.layers.1.self_attn.q_proj": {
723
+ "layer": 1,
724
+ "site": "q",
725
+ "t": 0.0,
726
+ "risk_norm": 0.10125742447038606,
727
+ "conc": 0.23489407000275134,
728
+ "Einf": 0.003899307172413092
729
+ },
730
+ "base_model.model.model.layers.8.self_attn.q_proj": {
731
+ "layer": 8,
732
+ "site": "q",
733
+ "t": 0.0,
734
+ "risk_norm": 0.10029293127655854,
735
+ "conc": 0.3528165128300204,
736
+ "Einf": 0.004495728729038976
737
+ },
738
+ "base_model.model.model.layers.3.self_attn.k_proj": {
739
+ "layer": 3,
740
+ "site": "k",
741
+ "t": 0.0,
742
+ "risk_norm": 0.09624046705410681,
743
+ "conc": 0.35132520832745906,
744
+ "Einf": 0.001129628867644222
745
+ },
746
+ "base_model.model.model.layers.9.self_attn.k_proj": {
747
+ "layer": 9,
748
+ "site": "k",
749
+ "t": 0.0,
750
+ "risk_norm": 0.08949176473758313,
751
+ "conc": 0.4009409511992916,
752
+ "Einf": 0.0013528282774629588
753
+ },
754
+ "base_model.model.model.layers.4.self_attn.q_proj": {
755
+ "layer": 4,
756
+ "site": "q",
757
+ "t": 0.0,
758
+ "risk_norm": 0.08563034621581374,
759
+ "conc": 0.24836996068871453,
760
+ "Einf": 0.004416502025704698
761
+ },
762
+ "base_model.model.model.layers.4.self_attn.k_proj": {
763
+ "layer": 4,
764
+ "site": "k",
765
+ "t": 0.0,
766
+ "risk_norm": 0.08481164088870678,
767
+ "conc": 0.3499893704067075,
768
+ "Einf": 0.00117946680261548
769
+ },
770
+ "base_model.model.model.layers.10.self_attn.v_proj": {
771
+ "layer": 10,
772
+ "site": "v",
773
+ "t": 0.0,
774
+ "risk_norm": 0.08440333302714915,
775
+ "conc": 0.3686548838906111,
776
+ "Einf": 0.0012642178351217883
777
+ },
778
+ "base_model.model.model.layers.1.self_attn.k_proj": {
779
+ "layer": 1,
780
+ "site": "k",
781
+ "t": 0.0,
782
+ "risk_norm": 0.08300665875798967,
783
+ "conc": 0.2949684265675236,
784
+ "Einf": 0.0011208344340752842
785
+ },
786
+ "base_model.model.model.layers.8.self_attn.v_proj": {
787
+ "layer": 8,
788
+ "site": "v",
789
+ "t": 0.0,
790
+ "risk_norm": 0.08100134292755189,
791
+ "conc": 0.41333728471484454,
792
+ "Einf": 0.0012872318050572281
793
+ },
794
+ "base_model.model.model.layers.9.self_attn.v_proj": {
795
+ "layer": 9,
796
+ "site": "v",
797
+ "t": 0.0,
798
+ "risk_norm": 0.07460147359641223,
799
+ "conc": 0.3740234441704972,
800
+ "Einf": 0.0012493019280592826
801
+ },
802
+ "base_model.model.model.layers.12.self_attn.v_proj": {
803
+ "layer": 12,
804
+ "site": "v",
805
+ "t": 0.0,
806
+ "risk_norm": 0.07334512436626224,
807
+ "conc": 0.30716711112205186,
808
+ "Einf": 0.0012818119612575108
809
+ },
810
+ "base_model.model.model.layers.6.self_attn.k_proj": {
811
+ "layer": 6,
812
+ "site": "k",
813
+ "t": 0.0,
814
+ "risk_norm": 0.06844861059087952,
815
+ "conc": 0.36038033907582323,
816
+ "Einf": 0.0012856569832680755
817
+ },
818
+ "base_model.model.model.layers.6.self_attn.q_proj": {
819
+ "layer": 6,
820
+ "site": "q",
821
+ "t": 0.0,
822
+ "risk_norm": 0.06120451350656449,
823
+ "conc": 0.23414995165695357,
824
+ "Einf": 0.004787360600272061
825
+ },
826
+ "base_model.model.model.layers.7.self_attn.k_proj": {
827
+ "layer": 7,
828
+ "site": "k",
829
+ "t": 0.0,
830
+ "risk_norm": 0.06062201513851076,
831
+ "conc": 0.37227341851232926,
832
+ "Einf": 0.0011943883655813304
833
+ },
834
+ "base_model.model.model.layers.5.self_attn.k_proj": {
835
+ "layer": 5,
836
+ "site": "k",
837
+ "t": 0.0,
838
+ "risk_norm": 0.058186103641922664,
839
+ "conc": 0.32216495891446306,
840
+ "Einf": 0.0011490160143665486
841
+ },
842
+ "base_model.model.model.layers.13.self_attn.k_proj": {
843
+ "layer": 13,
844
+ "site": "k",
845
+ "t": 0.0,
846
+ "risk_norm": 0.05429534630934255,
847
+ "conc": 0.25912356845712226,
848
+ "Einf": 0.0012804168384325163
849
+ },
850
+ "base_model.model.model.layers.10.self_attn.k_proj": {
851
+ "layer": 10,
852
+ "site": "k",
853
+ "t": 0.0,
854
+ "risk_norm": 0.04815772151321518,
855
+ "conc": 0.2983771031280095,
856
+ "Einf": 0.001293485448081585
857
+ },
858
+ "base_model.model.model.layers.0.self_attn.k_proj": {
859
+ "layer": 0,
860
+ "site": "k",
861
+ "t": 0.0,
862
+ "risk_norm": 0.04487812782356599,
863
+ "conc": 0.23033793242991982,
864
+ "Einf": 0.0009072366643817679
865
+ },
866
+ "base_model.model.model.layers.11.self_attn.v_proj": {
867
+ "layer": 11,
868
+ "site": "v",
869
+ "t": 0.0,
870
+ "risk_norm": 0.04112460222325889,
871
+ "conc": 0.26718136790247904,
872
+ "Einf": 0.0012969956803047008
873
+ },
874
+ "base_model.model.model.layers.7.self_attn.q_proj": {
875
+ "layer": 7,
876
+ "site": "q",
877
+ "t": 0.0,
878
+ "risk_norm": 0.037911403281581366,
879
+ "conc": 0.21978305706660872,
880
+ "Einf": 0.004441863890112757
881
+ },
882
+ "base_model.model.model.layers.2.self_attn.k_proj": {
883
+ "layer": 2,
884
+ "site": "k",
885
+ "t": 0.0,
886
+ "risk_norm": 0.03362830482772895,
887
+ "conc": 0.2317292615518013,
888
+ "Einf": 0.0010807417197930991
889
+ },
890
+ "base_model.model.model.layers.8.self_attn.k_proj": {
891
+ "layer": 8,
892
+ "site": "k",
893
+ "t": 0.0,
894
+ "risk_norm": 0.030548355139784003,
895
+ "conc": 0.302063951049893,
896
+ "Einf": 0.001359157134079161
897
+ },
898
+ "base_model.model.model.layers.14.self_attn.k_proj": {
899
+ "layer": 14,
900
+ "site": "k",
901
+ "t": 0.0,
902
+ "risk_norm": 0.0,
903
+ "conc": 0.16336153047732316,
904
+ "Einf": 0.0011944089054408448
905
+ }
906
+ },
907
+ "layer_view": {
908
+ "14": {
909
+ "base_model.model.model.layers.14.mlp.up_proj": {
910
+ "site": "up",
911
+ "t": 0.8129976992426243
912
+ },
913
+ "base_model.model.model.layers.14.mlp.gate_proj": {
914
+ "site": "gate",
915
+ "t": 0.7335376639204393
916
+ },
917
+ "base_model.model.model.layers.14.mlp.down_proj": {
918
+ "site": "down",
919
+ "t": 0.0
920
+ },
921
+ "base_model.model.model.layers.14.self_attn.o_proj": {
922
+ "site": "o",
923
+ "t": 0.0
924
+ },
925
+ "base_model.model.model.layers.14.self_attn.q_proj": {
926
+ "site": "q",
927
+ "t": 0.0
928
+ },
929
+ "base_model.model.model.layers.14.self_attn.v_proj": {
930
+ "site": "v",
931
+ "t": 0.0
932
+ },
933
+ "base_model.model.model.layers.14.self_attn.k_proj": {
934
+ "site": "k",
935
+ "t": 0.0
936
+ }
937
+ },
938
+ "15": {
939
+ "base_model.model.model.layers.15.mlp.up_proj": {
940
+ "site": "up",
941
+ "t": 0.7235717332710954
942
+ },
943
+ "base_model.model.model.layers.15.mlp.gate_proj": {
944
+ "site": "gate",
945
+ "t": 0.7007285972024974
946
+ },
947
+ "base_model.model.model.layers.15.mlp.down_proj": {
948
+ "site": "down",
949
+ "t": 0.05170055028997815
950
+ },
951
+ "base_model.model.model.layers.15.self_attn.o_proj": {
952
+ "site": "o",
953
+ "t": 0.0
954
+ },
955
+ "base_model.model.model.layers.15.self_attn.k_proj": {
956
+ "site": "k",
957
+ "t": 0.0
958
+ },
959
+ "base_model.model.model.layers.15.self_attn.q_proj": {
960
+ "site": "q",
961
+ "t": 0.0
962
+ },
963
+ "base_model.model.model.layers.15.self_attn.v_proj": {
964
+ "site": "v",
965
+ "t": 0.0
966
+ }
967
+ },
968
+ "12": {
969
+ "base_model.model.model.layers.12.mlp.up_proj": {
970
+ "site": "up",
971
+ "t": 0.8040670570506432
972
+ },
973
+ "base_model.model.model.layers.12.mlp.gate_proj": {
974
+ "site": "gate",
975
+ "t": 0.62047731449069
976
+ },
977
+ "base_model.model.model.layers.12.mlp.down_proj": {
978
+ "site": "down",
979
+ "t": 0.0
980
+ },
981
+ "base_model.model.model.layers.12.self_attn.q_proj": {
982
+ "site": "q",
983
+ "t": 0.0
984
+ },
985
+ "base_model.model.model.layers.12.self_attn.o_proj": {
986
+ "site": "o",
987
+ "t": 0.0
988
+ },
989
+ "base_model.model.model.layers.12.self_attn.k_proj": {
990
+ "site": "k",
991
+ "t": 0.0
992
+ },
993
+ "base_model.model.model.layers.12.self_attn.v_proj": {
994
+ "site": "v",
995
+ "t": 0.0
996
+ }
997
+ },
998
+ "1": {
999
+ "base_model.model.model.layers.1.mlp.down_proj": {
1000
+ "site": "down",
1001
+ "t": 0.852820759307394
1002
+ },
1003
+ "base_model.model.model.layers.1.mlp.up_proj": {
1004
+ "site": "up",
1005
+ "t": 0.4239953417816056
1006
+ },
1007
+ "base_model.model.model.layers.1.mlp.gate_proj": {
1008
+ "site": "gate",
1009
+ "t": 0.07666286945292178
1010
+ },
1011
+ "base_model.model.model.layers.1.self_attn.o_proj": {
1012
+ "site": "o",
1013
+ "t": 0.0
1014
+ },
1015
+ "base_model.model.model.layers.1.self_attn.v_proj": {
1016
+ "site": "v",
1017
+ "t": 0.0
1018
+ },
1019
+ "base_model.model.model.layers.1.self_attn.q_proj": {
1020
+ "site": "q",
1021
+ "t": 0.0
1022
+ },
1023
+ "base_model.model.model.layers.1.self_attn.k_proj": {
1024
+ "site": "k",
1025
+ "t": 0.0
1026
+ }
1027
+ },
1028
+ "13": {
1029
+ "base_model.model.model.layers.13.mlp.gate_proj": {
1030
+ "site": "gate",
1031
+ "t": 0.7217311024300797
1032
+ },
1033
+ "base_model.model.model.layers.13.mlp.up_proj": {
1034
+ "site": "up",
1035
+ "t": 0.6710380698353853
1036
+ },
1037
+ "base_model.model.model.layers.13.mlp.down_proj": {
1038
+ "site": "down",
1039
+ "t": 0.0
1040
+ },
1041
+ "base_model.model.model.layers.13.self_attn.q_proj": {
1042
+ "site": "q",
1043
+ "t": 0.0
1044
+ },
1045
+ "base_model.model.model.layers.13.self_attn.o_proj": {
1046
+ "site": "o",
1047
+ "t": 0.0
1048
+ },
1049
+ "base_model.model.model.layers.13.self_attn.v_proj": {
1050
+ "site": "v",
1051
+ "t": 0.0
1052
+ },
1053
+ "base_model.model.model.layers.13.self_attn.k_proj": {
1054
+ "site": "k",
1055
+ "t": 0.0
1056
+ }
1057
+ },
1058
+ "7": {
1059
+ "base_model.model.model.layers.7.mlp.up_proj": {
1060
+ "site": "up",
1061
+ "t": 0.7783005199480901
1062
+ },
1063
+ "base_model.model.model.layers.7.mlp.gate_proj": {
1064
+ "site": "gate",
1065
+ "t": 0.2942002352889795
1066
+ },
1067
+ "base_model.model.model.layers.7.mlp.down_proj": {
1068
+ "site": "down",
1069
+ "t": 0.0
1070
+ },
1071
+ "base_model.model.model.layers.7.self_attn.o_proj": {
1072
+ "site": "o",
1073
+ "t": 0.0
1074
+ },
1075
+ "base_model.model.model.layers.7.self_attn.v_proj": {
1076
+ "site": "v",
1077
+ "t": 0.0
1078
+ },
1079
+ "base_model.model.model.layers.7.self_attn.k_proj": {
1080
+ "site": "k",
1081
+ "t": 0.0
1082
+ },
1083
+ "base_model.model.model.layers.7.self_attn.q_proj": {
1084
+ "site": "q",
1085
+ "t": 0.0
1086
+ }
1087
+ },
1088
+ "5": {
1089
+ "base_model.model.model.layers.5.mlp.up_proj": {
1090
+ "site": "up",
1091
+ "t": 0.690164909922832
1092
+ },
1093
+ "base_model.model.model.layers.5.mlp.gate_proj": {
1094
+ "site": "gate",
1095
+ "t": 0.2331380993725579
1096
+ },
1097
+ "base_model.model.model.layers.5.mlp.down_proj": {
1098
+ "site": "down",
1099
+ "t": 0.02696106887579347
1100
+ },
1101
+ "base_model.model.model.layers.5.self_attn.o_proj": {
1102
+ "site": "o",
1103
+ "t": 0.0
1104
+ },
1105
+ "base_model.model.model.layers.5.self_attn.v_proj": {
1106
+ "site": "v",
1107
+ "t": 0.0
1108
+ },
1109
+ "base_model.model.model.layers.5.self_attn.q_proj": {
1110
+ "site": "q",
1111
+ "t": 0.0
1112
+ },
1113
+ "base_model.model.model.layers.5.self_attn.k_proj": {
1114
+ "site": "k",
1115
+ "t": 0.0
1116
+ }
1117
+ },
1118
+ "4": {
1119
+ "base_model.model.model.layers.4.mlp.up_proj": {
1120
+ "site": "up",
1121
+ "t": 0.636118019193088
1122
+ },
1123
+ "base_model.model.model.layers.4.mlp.gate_proj": {
1124
+ "site": "gate",
1125
+ "t": 0.49967339162172797
1126
+ },
1127
+ "base_model.model.model.layers.4.mlp.down_proj": {
1128
+ "site": "down",
1129
+ "t": 0.0
1130
+ },
1131
+ "base_model.model.model.layers.4.self_attn.o_proj": {
1132
+ "site": "o",
1133
+ "t": 0.0
1134
+ },
1135
+ "base_model.model.model.layers.4.self_attn.v_proj": {
1136
+ "site": "v",
1137
+ "t": 0.0
1138
+ },
1139
+ "base_model.model.model.layers.4.self_attn.q_proj": {
1140
+ "site": "q",
1141
+ "t": 0.0
1142
+ },
1143
+ "base_model.model.model.layers.4.self_attn.k_proj": {
1144
+ "site": "k",
1145
+ "t": 0.0
1146
+ }
1147
+ },
1148
+ "11": {
1149
+ "base_model.model.model.layers.11.mlp.up_proj": {
1150
+ "site": "up",
1151
+ "t": 0.6347402283240952
1152
+ },
1153
+ "base_model.model.model.layers.11.mlp.gate_proj": {
1154
+ "site": "gate",
1155
+ "t": 0.5039493435609566
1156
+ },
1157
+ "base_model.model.model.layers.11.mlp.down_proj": {
1158
+ "site": "down",
1159
+ "t": 0.0
1160
+ },
1161
+ "base_model.model.model.layers.11.self_attn.q_proj": {
1162
+ "site": "q",
1163
+ "t": 0.0
1164
+ },
1165
+ "base_model.model.model.layers.11.self_attn.o_proj": {
1166
+ "site": "o",
1167
+ "t": 0.0
1168
+ },
1169
+ "base_model.model.model.layers.11.self_attn.k_proj": {
1170
+ "site": "k",
1171
+ "t": 0.0
1172
+ },
1173
+ "base_model.model.model.layers.11.self_attn.v_proj": {
1174
+ "site": "v",
1175
+ "t": 0.0
1176
+ }
1177
+ },
1178
+ "6": {
1179
+ "base_model.model.model.layers.6.mlp.up_proj": {
1180
+ "site": "up",
1181
+ "t": 0.6530928346877478
1182
+ },
1183
+ "base_model.model.model.layers.6.mlp.gate_proj": {
1184
+ "site": "gate",
1185
+ "t": 0.5517768087168149
1186
+ },
1187
+ "base_model.model.model.layers.6.mlp.down_proj": {
1188
+ "site": "down",
1189
+ "t": 0.0
1190
+ },
1191
+ "base_model.model.model.layers.6.self_attn.o_proj": {
1192
+ "site": "o",
1193
+ "t": 0.0
1194
+ },
1195
+ "base_model.model.model.layers.6.self_attn.v_proj": {
1196
+ "site": "v",
1197
+ "t": 0.0
1198
+ },
1199
+ "base_model.model.model.layers.6.self_attn.k_proj": {
1200
+ "site": "k",
1201
+ "t": 0.0
1202
+ },
1203
+ "base_model.model.model.layers.6.self_attn.q_proj": {
1204
+ "site": "q",
1205
+ "t": 0.0
1206
+ }
1207
+ },
1208
+ "10": {
1209
+ "base_model.model.model.layers.10.mlp.up_proj": {
1210
+ "site": "up",
1211
+ "t": 0.6537704992680213
1212
+ },
1213
+ "base_model.model.model.layers.10.mlp.gate_proj": {
1214
+ "site": "gate",
1215
+ "t": 0.44427349837398933
1216
+ },
1217
+ "base_model.model.model.layers.10.mlp.down_proj": {
1218
+ "site": "down",
1219
+ "t": 0.0
1220
+ },
1221
+ "base_model.model.model.layers.10.self_attn.o_proj": {
1222
+ "site": "o",
1223
+ "t": 0.0
1224
+ },
1225
+ "base_model.model.model.layers.10.self_attn.q_proj": {
1226
+ "site": "q",
1227
+ "t": 0.0
1228
+ },
1229
+ "base_model.model.model.layers.10.self_attn.v_proj": {
1230
+ "site": "v",
1231
+ "t": 0.0
1232
+ },
1233
+ "base_model.model.model.layers.10.self_attn.k_proj": {
1234
+ "site": "k",
1235
+ "t": 0.0
1236
+ }
1237
+ },
1238
+ "8": {
1239
+ "base_model.model.model.layers.8.mlp.up_proj": {
1240
+ "site": "up",
1241
+ "t": 0.7084386947291339
1242
+ },
1243
+ "base_model.model.model.layers.8.mlp.gate_proj": {
1244
+ "site": "gate",
1245
+ "t": 0.22847834183504917
1246
+ },
1247
+ "base_model.model.model.layers.8.self_attn.o_proj": {
1248
+ "site": "o",
1249
+ "t": 0.0
1250
+ },
1251
+ "base_model.model.model.layers.8.mlp.down_proj": {
1252
+ "site": "down",
1253
+ "t": 0.0
1254
+ },
1255
+ "base_model.model.model.layers.8.self_attn.q_proj": {
1256
+ "site": "q",
1257
+ "t": 0.0
1258
+ },
1259
+ "base_model.model.model.layers.8.self_attn.v_proj": {
1260
+ "site": "v",
1261
+ "t": 0.0
1262
+ },
1263
+ "base_model.model.model.layers.8.self_attn.k_proj": {
1264
+ "site": "k",
1265
+ "t": 0.0
1266
+ }
1267
+ },
1268
+ "9": {
1269
+ "base_model.model.model.layers.9.mlp.up_proj": {
1270
+ "site": "up",
1271
+ "t": 0.6274885868185597
1272
+ },
1273
+ "base_model.model.model.layers.9.mlp.gate_proj": {
1274
+ "site": "gate",
1275
+ "t": 0.5337633143074141
1276
+ },
1277
+ "base_model.model.model.layers.9.mlp.down_proj": {
1278
+ "site": "down",
1279
+ "t": 0.0
1280
+ },
1281
+ "base_model.model.model.layers.9.self_attn.o_proj": {
1282
+ "site": "o",
1283
+ "t": 0.0
1284
+ },
1285
+ "base_model.model.model.layers.9.self_attn.q_proj": {
1286
+ "site": "q",
1287
+ "t": 0.0
1288
+ },
1289
+ "base_model.model.model.layers.9.self_attn.k_proj": {
1290
+ "site": "k",
1291
+ "t": 0.0
1292
+ },
1293
+ "base_model.model.model.layers.9.self_attn.v_proj": {
1294
+ "site": "v",
1295
+ "t": 0.0
1296
+ }
1297
+ },
1298
+ "3": {
1299
+ "base_model.model.model.layers.3.mlp.up_proj": {
1300
+ "site": "up",
1301
+ "t": 0.5976267038889456
1302
+ },
1303
+ "base_model.model.model.layers.3.mlp.gate_proj": {
1304
+ "site": "gate",
1305
+ "t": 0.2614555971813346
1306
+ },
1307
+ "base_model.model.model.layers.3.mlp.down_proj": {
1308
+ "site": "down",
1309
+ "t": 0.0019679547897706873
1310
+ },
1311
+ "base_model.model.model.layers.3.self_attn.v_proj": {
1312
+ "site": "v",
1313
+ "t": 0.0
1314
+ },
1315
+ "base_model.model.model.layers.3.self_attn.o_proj": {
1316
+ "site": "o",
1317
+ "t": 0.0
1318
+ },
1319
+ "base_model.model.model.layers.3.self_attn.q_proj": {
1320
+ "site": "q",
1321
+ "t": 0.0
1322
+ },
1323
+ "base_model.model.model.layers.3.self_attn.k_proj": {
1324
+ "site": "k",
1325
+ "t": 0.0
1326
+ }
1327
+ },
1328
+ "0": {
1329
+ "base_model.model.model.layers.0.mlp.down_proj": {
1330
+ "site": "down",
1331
+ "t": 0.5774169746344644
1332
+ },
1333
+ "base_model.model.model.layers.0.mlp.up_proj": {
1334
+ "site": "up",
1335
+ "t": 0.4680067424333691
1336
+ },
1337
+ "base_model.model.model.layers.0.mlp.gate_proj": {
1338
+ "site": "gate",
1339
+ "t": 0.12047786016271766
1340
+ },
1341
+ "base_model.model.model.layers.0.self_attn.o_proj": {
1342
+ "site": "o",
1343
+ "t": 0.0
1344
+ },
1345
+ "base_model.model.model.layers.0.self_attn.v_proj": {
1346
+ "site": "v",
1347
+ "t": 0.0
1348
+ },
1349
+ "base_model.model.model.layers.0.self_attn.q_proj": {
1350
+ "site": "q",
1351
+ "t": 0.0
1352
+ },
1353
+ "base_model.model.model.layers.0.self_attn.k_proj": {
1354
+ "site": "k",
1355
+ "t": 0.0
1356
+ }
1357
+ },
1358
+ "2": {
1359
+ "base_model.model.model.layers.2.mlp.up_proj": {
1360
+ "site": "up",
1361
+ "t": 0.5473159019958165
1362
+ },
1363
+ "base_model.model.model.layers.2.mlp.gate_proj": {
1364
+ "site": "gate",
1365
+ "t": 0.1944066721607684
1366
+ },
1367
+ "base_model.model.model.layers.2.mlp.down_proj": {
1368
+ "site": "down",
1369
+ "t": 0.018606340442403668
1370
+ },
1371
+ "base_model.model.model.layers.2.self_attn.v_proj": {
1372
+ "site": "v",
1373
+ "t": 0.0
1374
+ },
1375
+ "base_model.model.model.layers.2.self_attn.o_proj": {
1376
+ "site": "o",
1377
+ "t": 0.0
1378
+ },
1379
+ "base_model.model.model.layers.2.self_attn.q_proj": {
1380
+ "site": "q",
1381
+ "t": 0.0013359830432677633
1382
+ },
1383
+ "base_model.model.model.layers.2.self_attn.k_proj": {
1384
+ "site": "k",
1385
+ "t": 0.0
1386
+ }
1387
+ }
1388
+ }
1389
+ }