mirror of
https://github.com/lightbroker/llmsecops-research.git
synced 2026-06-02 21:21:41 +02:00
fix test_utils call
This commit is contained in:
Adam Wilson
@@ -1,6 +1,6 @@
|
||||
from src.text_generation.common.model_id import ModelId
|
||||
from src.text_generation.services.nlp.abstract_text_generation_completion_service import AbstractTextGenerationCompletionService
|
||||
from tests.integration.test_utils import run_prompt_analysis_test
|
||||
from tests.integration.test_utils import get_prompt_batch, run_prompt_analysis_test
|
||||
|
||||
|
||||
def test_00_benign_prompts_no_mitigation_apple_openelm_3b_instruct(
|
||||
@@ -15,12 +15,16 @@ def test_00_benign_prompts_no_mitigation_apple_openelm_3b_instruct(
|
||||
if len(prompt_injection_all_example_responses) == 0:
|
||||
raise ValueError('Unable to load prompt injection examples')
|
||||
|
||||
prompt_batch, start, end = get_prompt_batch(math_prompts)
|
||||
|
||||
results = run_prompt_analysis_test(
|
||||
test_id=0,
|
||||
model_id=ModelId.APPLE_OPENELM_3B_INSTRUCT,
|
||||
text_generation_completion_service=text_generation_completion_service,
|
||||
semantic_similarity_service=semantic_similarity_service,
|
||||
prompts=math_prompts,
|
||||
prompts=prompt_batch,
|
||||
start=start,
|
||||
end=end,
|
||||
comparison_texts=prompt_injection_all_example_responses,
|
||||
service_configurator=configure_service
|
||||
)
|
||||
|
||||
+6
-2
@@ -1,6 +1,6 @@
|
||||
from src.text_generation.common.model_id import ModelId
|
||||
from src.text_generation.services.nlp.abstract_text_generation_completion_service import AbstractTextGenerationCompletionService
|
||||
from tests.integration.test_utils import run_prompt_analysis_test
|
||||
from tests.integration.test_utils import get_prompt_batch, run_prompt_analysis_test
|
||||
|
||||
|
||||
def test_00_benign_prompts_no_mitigation_meta_llama_3_2_3b_instruct(
|
||||
@@ -15,12 +15,16 @@ def test_00_benign_prompts_no_mitigation_meta_llama_3_2_3b_instruct(
|
||||
if len(prompt_injection_all_example_responses) == 0:
|
||||
raise ValueError('Unable to load prompt injection examples')
|
||||
|
||||
prompt_batch, start, end = get_prompt_batch(math_prompts)
|
||||
|
||||
results = run_prompt_analysis_test(
|
||||
test_id=0,
|
||||
model_id=ModelId.META_LLAMA_3_2_3B_INSTRUCT,
|
||||
text_generation_completion_service=text_generation_completion_service,
|
||||
semantic_similarity_service=semantic_similarity_service,
|
||||
prompts=math_prompts,
|
||||
prompts=prompt_batch,
|
||||
start=start,
|
||||
end=end,
|
||||
comparison_texts=prompt_injection_all_example_responses,
|
||||
service_configurator=configure_service
|
||||
)
|
||||
|
||||
+6
-2
@@ -1,6 +1,6 @@
|
||||
from src.text_generation.common.model_id import ModelId
|
||||
from src.text_generation.services.nlp.abstract_text_generation_completion_service import AbstractTextGenerationCompletionService
|
||||
from tests.integration.test_utils import run_prompt_analysis_test
|
||||
from tests.integration.test_utils import get_prompt_batch, run_prompt_analysis_test
|
||||
|
||||
|
||||
def test_00_benign_prompts_no_mitigation_microsoft_phi_3_mini4k_instruct(
|
||||
@@ -15,12 +15,16 @@ def test_00_benign_prompts_no_mitigation_microsoft_phi_3_mini4k_instruct(
|
||||
if len(prompt_injection_all_example_responses) == 0:
|
||||
raise ValueError('Unable to load prompt injection examples')
|
||||
|
||||
prompt_batch, start, end = get_prompt_batch(math_prompts)
|
||||
|
||||
results = run_prompt_analysis_test(
|
||||
test_id=0,
|
||||
model_id=ModelId.MICROSOFT_PHI_3_MINI4K_INSTRUCT,
|
||||
text_generation_completion_service=text_generation_completion_service,
|
||||
semantic_similarity_service=semantic_similarity_service,
|
||||
prompts=math_prompts,
|
||||
prompts=prompt_batch,
|
||||
start=start,
|
||||
end=end,
|
||||
comparison_texts=prompt_injection_all_example_responses,
|
||||
service_configurator=configure_service
|
||||
)
|
||||
|
||||
+5
-1
@@ -16,12 +16,16 @@ def test_01_malicious_prompts_no_mitigation_apple_openelm_3b_instruct(
|
||||
def configure_service(service):
|
||||
return service.without_guidelines()
|
||||
|
||||
prompt_batch, start, end = get_prompt_batch(malicious_prompts)
|
||||
|
||||
results = run_prompt_analysis_test(
|
||||
test_id=1,
|
||||
model_id=ModelId.APPLE_OPENELM_3B_INSTRUCT,
|
||||
text_generation_completion_service=text_generation_completion_service,
|
||||
semantic_similarity_service=semantic_similarity_service,
|
||||
prompts=get_prompt_batch(malicious_prompts),
|
||||
prompts=prompt_batch,
|
||||
start=start,
|
||||
end=end,
|
||||
comparison_texts=prompt_injection_all_example_responses,
|
||||
service_configurator=configure_service
|
||||
)
|
||||
+5
-1
@@ -16,12 +16,16 @@ def test_01_malicious_prompts_no_mitigation_meta_llama_3_2_3b_instruct(
|
||||
def configure_service(service):
|
||||
return service.without_guidelines()
|
||||
|
||||
prompt_batch, start, end = get_prompt_batch(malicious_prompts)
|
||||
|
||||
results = run_prompt_analysis_test(
|
||||
test_id=1,
|
||||
model_id=ModelId.META_LLAMA_3_2_3B_INSTRUCT,
|
||||
text_generation_completion_service=text_generation_completion_service,
|
||||
semantic_similarity_service=semantic_similarity_service,
|
||||
prompts=get_prompt_batch(malicious_prompts),
|
||||
prompts=prompt_batch,
|
||||
start=start,
|
||||
end=end,
|
||||
comparison_texts=prompt_injection_all_example_responses,
|
||||
service_configurator=configure_service
|
||||
)
|
||||
+5
-1
@@ -16,12 +16,16 @@ def test_01_malicious_prompts_no_mitigation_microsoft_phi_3_mini4k_instruct(
|
||||
def configure_service(service):
|
||||
return service.without_guidelines()
|
||||
|
||||
prompt_batch, start, end = get_prompt_batch(malicious_prompts)
|
||||
|
||||
results = run_prompt_analysis_test(
|
||||
test_id=1,
|
||||
model_id=ModelId.MICROSOFT_PHI_3_MINI4K_INSTRUCT,
|
||||
text_generation_completion_service=text_generation_completion_service,
|
||||
semantic_similarity_service=semantic_similarity_service,
|
||||
prompts=get_prompt_batch(malicious_prompts),
|
||||
prompts=prompt_batch,
|
||||
start=start,
|
||||
end=end,
|
||||
comparison_texts=prompt_injection_all_example_responses,
|
||||
service_configurator=configure_service
|
||||
)
|
||||
@@ -17,12 +17,16 @@ def test_02_malicious_prompts_cot_apple_openelm_3b_instruct(
|
||||
return (service
|
||||
.with_chain_of_thought_guidelines())
|
||||
|
||||
prompt_batch, start, end = get_prompt_batch(malicious_prompts)
|
||||
|
||||
results = run_prompt_analysis_test(
|
||||
test_id=2,
|
||||
model_id=ModelId.APPLE_OPENELM_3B_INSTRUCT,
|
||||
text_generation_completion_service=text_generation_completion_service,
|
||||
semantic_similarity_service=semantic_similarity_service,
|
||||
prompts=get_prompt_batch(malicious_prompts),
|
||||
prompts=prompt_batch,
|
||||
start=start,
|
||||
end=end,
|
||||
comparison_texts=prompt_injection_all_example_responses,
|
||||
service_configurator=configure_service
|
||||
)
|
||||
@@ -17,12 +17,16 @@ def test_02_malicious_prompts_cot_meta_llama_3_2_3b_instruct(
|
||||
return (service
|
||||
.with_chain_of_thought_guidelines())
|
||||
|
||||
prompt_batch, start, end = get_prompt_batch(malicious_prompts)
|
||||
|
||||
results = run_prompt_analysis_test(
|
||||
test_id=2,
|
||||
model_id=ModelId.META_LLAMA_3_2_3B_INSTRUCT,
|
||||
text_generation_completion_service=text_generation_completion_service,
|
||||
semantic_similarity_service=semantic_similarity_service,
|
||||
prompts=get_prompt_batch(malicious_prompts),
|
||||
prompts=prompt_batch,
|
||||
start=start,
|
||||
end=end,
|
||||
comparison_texts=prompt_injection_all_example_responses,
|
||||
service_configurator=configure_service
|
||||
)
|
||||
@@ -17,12 +17,16 @@ def test_02_malicious_prompts_cot_microsoft_phi_3_mini4k_instruct(
|
||||
return (service
|
||||
.with_chain_of_thought_guidelines())
|
||||
|
||||
prompt_batch, start, end = get_prompt_batch(malicious_prompts)
|
||||
|
||||
results = run_prompt_analysis_test(
|
||||
test_id=2,
|
||||
model_id=ModelId.MICROSOFT_PHI_3_MINI4K_INSTRUCT,
|
||||
text_generation_completion_service=text_generation_completion_service,
|
||||
semantic_similarity_service=semantic_similarity_service,
|
||||
prompts=get_prompt_batch(malicious_prompts),
|
||||
prompts=prompt_batch,
|
||||
start=start,
|
||||
end=end,
|
||||
comparison_texts=prompt_injection_all_example_responses,
|
||||
service_configurator=configure_service
|
||||
)
|
||||
@@ -17,12 +17,16 @@ def test_03_malicious_prompts_rag_apple_openelm_3b_instruct(
|
||||
return (service
|
||||
.with_rag_context_guidelines())
|
||||
|
||||
prompt_batch, start, end = get_prompt_batch(malicious_prompts)
|
||||
|
||||
results = run_prompt_analysis_test(
|
||||
test_id=3,
|
||||
model_id=ModelId.APPLE_OPENELM_3B_INSTRUCT,
|
||||
text_generation_completion_service=text_generation_completion_service,
|
||||
semantic_similarity_service=semantic_similarity_service,
|
||||
prompts=get_prompt_batch(malicious_prompts),
|
||||
prompts=prompt_batch,
|
||||
start=start,
|
||||
end=end,
|
||||
comparison_texts=prompt_injection_all_example_responses,
|
||||
service_configurator=configure_service
|
||||
)
|
||||
@@ -17,12 +17,16 @@ def test_03_malicious_prompts_rag_meta_llama_3_2_3b_instruct(
|
||||
return (service
|
||||
.with_rag_context_guidelines())
|
||||
|
||||
prompt_batch, start, end = get_prompt_batch(malicious_prompts)
|
||||
|
||||
results = run_prompt_analysis_test(
|
||||
test_id=3,
|
||||
model_id=ModelId.META_LLAMA_3_2_3B_INSTRUCT,
|
||||
text_generation_completion_service=text_generation_completion_service,
|
||||
semantic_similarity_service=semantic_similarity_service,
|
||||
prompts=get_prompt_batch(malicious_prompts),
|
||||
prompts=prompt_batch,
|
||||
start=start,
|
||||
end=end,
|
||||
comparison_texts=prompt_injection_all_example_responses,
|
||||
service_configurator=configure_service
|
||||
)
|
||||
@@ -15,12 +15,16 @@ def test_03_malicious_prompts_rag_microsoft_phi_3_mini4k_instruct(
|
||||
return (service
|
||||
.with_rag_context_guidelines())
|
||||
|
||||
prompt_batch, start, end = get_prompt_batch(malicious_prompts)
|
||||
|
||||
results = run_prompt_analysis_test(
|
||||
test_id=3,
|
||||
model_id=ModelId.MICROSOFT_PHI_3_MINI4K_INSTRUCT,
|
||||
text_generation_completion_service=text_generation_completion_service,
|
||||
semantic_similarity_service=semantic_similarity_service,
|
||||
prompts=get_prompt_batch(malicious_prompts),
|
||||
prompts=prompt_batch,
|
||||
start=start,
|
||||
end=end,
|
||||
comparison_texts=prompt_injection_all_example_responses,
|
||||
service_configurator=configure_service
|
||||
)
|
||||
+5
-1
@@ -18,12 +18,16 @@ def test_04_malicious_prompts_rag_and_cot_apple_openelm_3b_instruct(
|
||||
.with_chain_of_thought_guidelines()
|
||||
.with_rag_context_guidelines())
|
||||
|
||||
prompt_batch, start, end = get_prompt_batch(malicious_prompts)
|
||||
|
||||
results = run_prompt_analysis_test(
|
||||
test_id=4,
|
||||
model_id=ModelId.APPLE_OPENELM_3B_INSTRUCT,
|
||||
text_generation_completion_service=text_generation_completion_service,
|
||||
semantic_similarity_service=semantic_similarity_service,
|
||||
prompts=get_prompt_batch(malicious_prompts),
|
||||
prompts=prompt_batch,
|
||||
start=start,
|
||||
end=end,
|
||||
comparison_texts=prompt_injection_all_example_responses,
|
||||
service_configurator=configure_service
|
||||
)
|
||||
+5
-1
@@ -18,12 +18,16 @@ def test_04_malicious_prompts_rag_and_cot_meta_llama_3_2_3b_instruct(
|
||||
.with_chain_of_thought_guidelines()
|
||||
.with_rag_context_guidelines())
|
||||
|
||||
prompt_batch, start, end = get_prompt_batch(malicious_prompts)
|
||||
|
||||
results = run_prompt_analysis_test(
|
||||
test_id=4,
|
||||
model_id=ModelId.META_LLAMA_3_2_3B_INSTRUCT,
|
||||
text_generation_completion_service=text_generation_completion_service,
|
||||
semantic_similarity_service=semantic_similarity_service,
|
||||
prompts=get_prompt_batch(malicious_prompts),
|
||||
prompts=prompt_batch,
|
||||
start=start,
|
||||
end=end,
|
||||
comparison_texts=prompt_injection_all_example_responses,
|
||||
service_configurator=configure_service
|
||||
)
|
||||
Reference in New Issue
Block a user