Text Generation
Transformers
Safetensors
qwen3
uncensored
srsa
abliterated
roleplay
conversational
text-generation-inference
Instructions to use beyoru/Luna-SRSA-Uncensored with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use beyoru/Luna-SRSA-Uncensored with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="beyoru/Luna-SRSA-Uncensored") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("beyoru/Luna-SRSA-Uncensored") model = AutoModelForCausalLM.from_pretrained("beyoru/Luna-SRSA-Uncensored", device_map="auto") messages = [ {"role": "user", "content": "Who are you?"}, ] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use beyoru/Luna-SRSA-Uncensored with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "beyoru/Luna-SRSA-Uncensored" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "beyoru/Luna-SRSA-Uncensored", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/beyoru/Luna-SRSA-Uncensored
- SGLang
How to use beyoru/Luna-SRSA-Uncensored with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "beyoru/Luna-SRSA-Uncensored" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "beyoru/Luna-SRSA-Uncensored", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "beyoru/Luna-SRSA-Uncensored" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "beyoru/Luna-SRSA-Uncensored", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use beyoru/Luna-SRSA-Uncensored with Docker Model Runner:
docker model run hf.co/beyoru/Luna-SRSA-Uncensored
Upload Luna MNN 4-bit quantized model
Browse files- .gitattributes +1 -0
- README.md +35 -0
- chat_template.jinja +86 -0
- config.json +71 -0
- generation_config.json +13 -0
- model.safetensors +3 -0
- srsa_analysis.json +195 -0
- tokenizer.json +3 -0
- tokenizer_config.json +31 -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 |
+
tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
base_model: beyoru/Luna
|
| 4 |
+
tags:
|
| 5 |
+
- uncensored
|
| 6 |
+
- srsa
|
| 7 |
+
- abliterated
|
| 8 |
+
- roleplay
|
| 9 |
+
- qwen3
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Luna-SRSA-Uncensored
|
| 13 |
+
|
| 14 |
+
Uncensored version of `beyoru/Luna` via **SRSA (Spectral Refusal Subspace Ablation)**.
|
| 15 |
+
|
| 16 |
+
## Method
|
| 17 |
+
|
| 18 |
+
SRSA differs from standard abliteration by using:
|
| 19 |
+
- **3-prong spectral discovery**: mean diff + differential PCA + contrastive SVD
|
| 20 |
+
- **Capability-aware filtering**: PCA overlap check to protect useful directions
|
| 21 |
+
- **Graduated per-layer intervention**: weighted by Cohen's d separability
|
| 22 |
+
- **Spectral norm preservation**: hybrid 70/30 spectral+Frobenius scaling
|
| 23 |
+
|
| 24 |
+
## Config
|
| 25 |
+
- Max rank/layer: 3
|
| 26 |
+
- Strength: 0.95
|
| 27 |
+
- Eigenratio threshold: 0.12
|
| 28 |
+
- Capability protection: 0.25
|
| 29 |
+
- Layers modified: 36
|
| 30 |
+
- Total directions removed: 84
|
| 31 |
+
|
| 32 |
+
## Stats
|
| 33 |
+
- Harmful refusals after SRSA: 2/8
|
| 34 |
+
- Harmless coherence: 4/4
|
| 35 |
+
- RP capability: 2/2
|
chat_template.jinja
ADDED
|
@@ -0,0 +1,86 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{%- if tools %}
|
| 2 |
+
{{- '<|im_start|>system\n' }}
|
| 3 |
+
{%- if messages[0].role == 'system' %}
|
| 4 |
+
{{- messages[0].content + '\n\n' }}
|
| 5 |
+
{%- endif %}
|
| 6 |
+
{{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
|
| 7 |
+
{%- for tool in tools %}
|
| 8 |
+
{{- "\n" }}
|
| 9 |
+
{{- tool | tojson }}
|
| 10 |
+
{%- endfor %}
|
| 11 |
+
{{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
|
| 12 |
+
{%- else %}
|
| 13 |
+
{%- if messages[0].role == 'system' %}
|
| 14 |
+
{{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
|
| 15 |
+
{%- endif %}
|
| 16 |
+
{%- endif %}
|
| 17 |
+
{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
|
| 18 |
+
{%- for message in messages[::-1] %}
|
| 19 |
+
{%- set index = (messages|length - 1) - loop.index0 %}
|
| 20 |
+
{%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
|
| 21 |
+
{%- set ns.multi_step_tool = false %}
|
| 22 |
+
{%- set ns.last_query_index = index %}
|
| 23 |
+
{%- endif %}
|
| 24 |
+
{%- endfor %}
|
| 25 |
+
{%- for message in messages %}
|
| 26 |
+
{%- if message.content is string %}
|
| 27 |
+
{%- set content = message.content %}
|
| 28 |
+
{%- else %}
|
| 29 |
+
{%- set content = '' %}
|
| 30 |
+
{%- endif %}
|
| 31 |
+
{%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
|
| 32 |
+
{{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
|
| 33 |
+
{%- elif message.role == "assistant" %}
|
| 34 |
+
{%- set reasoning_content = '' %}
|
| 35 |
+
{%- if message.reasoning_content is string %}
|
| 36 |
+
{%- set reasoning_content = message.reasoning_content %}
|
| 37 |
+
{%- else %}
|
| 38 |
+
{%- if '</think>' in content %}
|
| 39 |
+
{%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
|
| 40 |
+
{%- set content = content.split('</think>')[-1].lstrip('\n') %}
|
| 41 |
+
{%- endif %}
|
| 42 |
+
{%- endif %}
|
| 43 |
+
{%- if loop.index0 > ns.last_query_index %}
|
| 44 |
+
{%- if loop.last or (not loop.last and reasoning_content) %}
|
| 45 |
+
{{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
|
| 46 |
+
{%- else %}
|
| 47 |
+
{{- '<|im_start|>' + message.role + '\n' + content }}
|
| 48 |
+
{%- endif %}
|
| 49 |
+
{%- else %}
|
| 50 |
+
{{- '<|im_start|>' + message.role + '\n' + content }}
|
| 51 |
+
{%- endif %}
|
| 52 |
+
{%- if message.tool_calls %}
|
| 53 |
+
{%- for tool_call in message.tool_calls %}
|
| 54 |
+
{%- if (loop.first and content) or (not loop.first) %}
|
| 55 |
+
{{- '\n' }}
|
| 56 |
+
{%- endif %}
|
| 57 |
+
{%- if tool_call.function %}
|
| 58 |
+
{%- set tool_call = tool_call.function %}
|
| 59 |
+
{%- endif %}
|
| 60 |
+
{{- '<tool_call>\n{"name": "' }}
|
| 61 |
+
{{- tool_call.name }}
|
| 62 |
+
{{- '", "arguments": ' }}
|
| 63 |
+
{%- if tool_call.arguments is string %}
|
| 64 |
+
{{- tool_call.arguments }}
|
| 65 |
+
{%- else %}
|
| 66 |
+
{{- tool_call.arguments | tojson }}
|
| 67 |
+
{%- endif %}
|
| 68 |
+
{{- '}\n</tool_call>' }}
|
| 69 |
+
{%- endfor %}
|
| 70 |
+
{%- endif %}
|
| 71 |
+
{{- '<|im_end|>\n' }}
|
| 72 |
+
{%- elif message.role == "tool" %}
|
| 73 |
+
{%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
|
| 74 |
+
{{- '<|im_start|>user' }}
|
| 75 |
+
{%- endif %}
|
| 76 |
+
{{- '\n<tool_response>\n' }}
|
| 77 |
+
{{- content }}
|
| 78 |
+
{{- '\n</tool_response>' }}
|
| 79 |
+
{%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
|
| 80 |
+
{{- '<|im_end|>\n' }}
|
| 81 |
+
{%- endif %}
|
| 82 |
+
{%- endif %}
|
| 83 |
+
{%- endfor %}
|
| 84 |
+
{%- if add_generation_prompt %}
|
| 85 |
+
{{- '<|im_start|>assistant\n' }}
|
| 86 |
+
{%- endif %}
|
config.json
ADDED
|
@@ -0,0 +1,71 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"bos_token_id": 151643,
|
| 8 |
+
"dtype": "bfloat16",
|
| 9 |
+
"eos_token_id": 151645,
|
| 10 |
+
"head_dim": 128,
|
| 11 |
+
"hidden_act": "silu",
|
| 12 |
+
"hidden_size": 2560,
|
| 13 |
+
"initializer_range": 0.02,
|
| 14 |
+
"intermediate_size": 9728,
|
| 15 |
+
"layer_types": [
|
| 16 |
+
"full_attention",
|
| 17 |
+
"full_attention",
|
| 18 |
+
"full_attention",
|
| 19 |
+
"full_attention",
|
| 20 |
+
"full_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"full_attention",
|
| 23 |
+
"full_attention",
|
| 24 |
+
"full_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"full_attention",
|
| 27 |
+
"full_attention",
|
| 28 |
+
"full_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"full_attention",
|
| 31 |
+
"full_attention",
|
| 32 |
+
"full_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"full_attention",
|
| 35 |
+
"full_attention",
|
| 36 |
+
"full_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"full_attention",
|
| 39 |
+
"full_attention",
|
| 40 |
+
"full_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"full_attention",
|
| 43 |
+
"full_attention",
|
| 44 |
+
"full_attention",
|
| 45 |
+
"full_attention",
|
| 46 |
+
"full_attention",
|
| 47 |
+
"full_attention",
|
| 48 |
+
"full_attention",
|
| 49 |
+
"full_attention",
|
| 50 |
+
"full_attention",
|
| 51 |
+
"full_attention"
|
| 52 |
+
],
|
| 53 |
+
"max_position_embeddings": 262144,
|
| 54 |
+
"max_window_layers": 36,
|
| 55 |
+
"model_type": "qwen3",
|
| 56 |
+
"num_attention_heads": 32,
|
| 57 |
+
"num_hidden_layers": 36,
|
| 58 |
+
"num_key_value_heads": 8,
|
| 59 |
+
"pad_token_id": null,
|
| 60 |
+
"rms_norm_eps": 1e-06,
|
| 61 |
+
"rope_parameters": {
|
| 62 |
+
"rope_theta": 5000000,
|
| 63 |
+
"rope_type": "default"
|
| 64 |
+
},
|
| 65 |
+
"sliding_window": null,
|
| 66 |
+
"tie_word_embeddings": true,
|
| 67 |
+
"transformers_version": "5.0.0",
|
| 68 |
+
"use_cache": true,
|
| 69 |
+
"use_sliding_window": false,
|
| 70 |
+
"vocab_size": 151936
|
| 71 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 151643,
|
| 3 |
+
"do_sample": true,
|
| 4 |
+
"eos_token_id": [
|
| 5 |
+
151645,
|
| 6 |
+
151643
|
| 7 |
+
],
|
| 8 |
+
"pad_token_id": 151643,
|
| 9 |
+
"temperature": 0.7,
|
| 10 |
+
"top_k": 20,
|
| 11 |
+
"top_p": 0.8,
|
| 12 |
+
"transformers_version": "5.0.0"
|
| 13 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:15f1b99cd6dbf2d7da7aa55cf592fdf1d9df16cb00e5501a76b2a009d32b74bc
|
| 3 |
+
size 8044982080
|
srsa_analysis.json
ADDED
|
@@ -0,0 +1,195 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"config": {
|
| 3 |
+
"model": "beyoru/Luna",
|
| 4 |
+
"max_rank": 3,
|
| 5 |
+
"strength": 0.95,
|
| 6 |
+
"min_eigenratio": 0.12,
|
| 7 |
+
"capability_protection": 0.25
|
| 8 |
+
},
|
| 9 |
+
"layer_stats": [
|
| 10 |
+
{
|
| 11 |
+
"layer": 0,
|
| 12 |
+
"n_dirs": 3,
|
| 13 |
+
"strength": 0.29401210353601287
|
| 14 |
+
},
|
| 15 |
+
{
|
| 16 |
+
"layer": 1,
|
| 17 |
+
"n_dirs": 2,
|
| 18 |
+
"strength": 0.2917754046855508
|
| 19 |
+
},
|
| 20 |
+
{
|
| 21 |
+
"layer": 2,
|
| 22 |
+
"n_dirs": 2,
|
| 23 |
+
"strength": 0.3115962164422338
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"layer": 3,
|
| 27 |
+
"n_dirs": 2,
|
| 28 |
+
"strength": 0.48054619679620614
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"layer": 4,
|
| 32 |
+
"n_dirs": 2,
|
| 33 |
+
"strength": 0.43226880931003825
|
| 34 |
+
},
|
| 35 |
+
{
|
| 36 |
+
"layer": 5,
|
| 37 |
+
"n_dirs": 2,
|
| 38 |
+
"strength": 0.41851801772025127
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"layer": 6,
|
| 42 |
+
"n_dirs": 2,
|
| 43 |
+
"strength": 0.3557159060717538
|
| 44 |
+
},
|
| 45 |
+
{
|
| 46 |
+
"layer": 7,
|
| 47 |
+
"n_dirs": 2,
|
| 48 |
+
"strength": 0.3465762291130352
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"layer": 8,
|
| 52 |
+
"n_dirs": 2,
|
| 53 |
+
"strength": 0.36751805598055237
|
| 54 |
+
},
|
| 55 |
+
{
|
| 56 |
+
"layer": 9,
|
| 57 |
+
"n_dirs": 2,
|
| 58 |
+
"strength": 0.40741677449259384
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"layer": 10,
|
| 62 |
+
"n_dirs": 2,
|
| 63 |
+
"strength": 0.44665410147593665
|
| 64 |
+
},
|
| 65 |
+
{
|
| 66 |
+
"layer": 11,
|
| 67 |
+
"n_dirs": 2,
|
| 68 |
+
"strength": 0.7389669674778427
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"layer": 12,
|
| 72 |
+
"n_dirs": 2,
|
| 73 |
+
"strength": 0.774626134098153
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"layer": 13,
|
| 77 |
+
"n_dirs": 2,
|
| 78 |
+
"strength": 0.7739796937054256
|
| 79 |
+
},
|
| 80 |
+
{
|
| 81 |
+
"layer": 14,
|
| 82 |
+
"n_dirs": 2,
|
| 83 |
+
"strength": 0.7895180699594949
|
| 84 |
+
},
|
| 85 |
+
{
|
| 86 |
+
"layer": 15,
|
| 87 |
+
"n_dirs": 2,
|
| 88 |
+
"strength": 0.7180429028228088
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"layer": 16,
|
| 92 |
+
"n_dirs": 2,
|
| 93 |
+
"strength": 0.7626604265352032
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"layer": 17,
|
| 97 |
+
"n_dirs": 2,
|
| 98 |
+
"strength": 0.8240846060478203
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"layer": 18,
|
| 102 |
+
"n_dirs": 2,
|
| 103 |
+
"strength": 0.9034583636706769
|
| 104 |
+
},
|
| 105 |
+
{
|
| 106 |
+
"layer": 19,
|
| 107 |
+
"n_dirs": 2,
|
| 108 |
+
"strength": 0.9282506479241859
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"layer": 20,
|
| 112 |
+
"n_dirs": 2,
|
| 113 |
+
"strength": 0.9380889909189546
|
| 114 |
+
},
|
| 115 |
+
{
|
| 116 |
+
"layer": 21,
|
| 117 |
+
"n_dirs": 2,
|
| 118 |
+
"strength": 0.939409923706219
|
| 119 |
+
},
|
| 120 |
+
{
|
| 121 |
+
"layer": 22,
|
| 122 |
+
"n_dirs": 2,
|
| 123 |
+
"strength": 0.946654058699486
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"layer": 23,
|
| 127 |
+
"n_dirs": 2,
|
| 128 |
+
"strength": 0.9502904263504464
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"layer": 24,
|
| 132 |
+
"n_dirs": 2,
|
| 133 |
+
"strength": 0.9544458704502765
|
| 134 |
+
},
|
| 135 |
+
{
|
| 136 |
+
"layer": 25,
|
| 137 |
+
"n_dirs": 3,
|
| 138 |
+
"strength": 0.953130743489159
|
| 139 |
+
},
|
| 140 |
+
{
|
| 141 |
+
"layer": 26,
|
| 142 |
+
"n_dirs": 3,
|
| 143 |
+
"strength": 0.950968290818247
|
| 144 |
+
},
|
| 145 |
+
{
|
| 146 |
+
"layer": 27,
|
| 147 |
+
"n_dirs": 3,
|
| 148 |
+
"strength": 0.955541026355603
|
| 149 |
+
},
|
| 150 |
+
{
|
| 151 |
+
"layer": 28,
|
| 152 |
+
"n_dirs": 3,
|
| 153 |
+
"strength": 0.9554455579874419
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"layer": 29,
|
| 157 |
+
"n_dirs": 3,
|
| 158 |
+
"strength": 0.9543950906540449
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"layer": 30,
|
| 162 |
+
"n_dirs": 3,
|
| 163 |
+
"strength": 0.9584300467485165
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"layer": 31,
|
| 167 |
+
"n_dirs": 3,
|
| 168 |
+
"strength": 0.9582414817299362
|
| 169 |
+
},
|
| 170 |
+
{
|
| 171 |
+
"layer": 32,
|
| 172 |
+
"n_dirs": 3,
|
| 173 |
+
"strength": 0.9539728189378939
|
| 174 |
+
},
|
| 175 |
+
{
|
| 176 |
+
"layer": 33,
|
| 177 |
+
"n_dirs": 3,
|
| 178 |
+
"strength": 0.9535894113253833
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"layer": 34,
|
| 182 |
+
"n_dirs": 3,
|
| 183 |
+
"strength": 0.9612463998973981
|
| 184 |
+
},
|
| 185 |
+
{
|
| 186 |
+
"layer": 35,
|
| 187 |
+
"n_dirs": 3,
|
| 188 |
+
"strength": 0.9608501117056393
|
| 189 |
+
}
|
| 190 |
+
],
|
| 191 |
+
"intervention_stats": {
|
| 192 |
+
"layers_modified": 36,
|
| 193 |
+
"total_directions": 84
|
| 194 |
+
}
|
| 195 |
+
}
|
tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b064181fd20c7d19d5613f061552279d7dcbb8210ffde80386e2a9a6ddf996e7
|
| 3 |
+
size 11422748
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"backend": "tokenizers",
|
| 4 |
+
"bos_token": null,
|
| 5 |
+
"clean_up_tokenization_spaces": false,
|
| 6 |
+
"eos_token": "<|im_end|>",
|
| 7 |
+
"errors": "replace",
|
| 8 |
+
"extra_special_tokens": [
|
| 9 |
+
"<|im_start|>",
|
| 10 |
+
"<|im_end|>",
|
| 11 |
+
"<|object_ref_start|>",
|
| 12 |
+
"<|object_ref_end|>",
|
| 13 |
+
"<|box_start|>",
|
| 14 |
+
"<|box_end|>",
|
| 15 |
+
"<|quad_start|>",
|
| 16 |
+
"<|quad_end|>",
|
| 17 |
+
"<|vision_start|>",
|
| 18 |
+
"<|vision_end|>",
|
| 19 |
+
"<|vision_pad|>",
|
| 20 |
+
"<|image_pad|>",
|
| 21 |
+
"<|video_pad|>"
|
| 22 |
+
],
|
| 23 |
+
"is_local": false,
|
| 24 |
+
"model_max_length": 262144,
|
| 25 |
+
"model_specific_special_tokens": {},
|
| 26 |
+
"pad_token": "<|endoftext|>",
|
| 27 |
+
"padding_side": "left",
|
| 28 |
+
"split_special_tokens": false,
|
| 29 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 30 |
+
"unk_token": null
|
| 31 |
+
}
|