Spaces:
Sleeping
Sleeping
File size: 5,076 Bytes
170d9a9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 |
import asyncio
import weave
from medrag_multi_modal.assistant import LLMClient, MedQAAssistant
from medrag_multi_modal.metrics import MMLUOptionAccuracy
from medrag_multi_modal.retrieval.text_retrieval import (
BM25sRetriever,
ContrieverRetriever,
MedCPTRetriever,
NVEmbed2Retriever,
)
def test_mmlu_correctness_anatomy_bm25s(model_name: str):
weave.init("ml-colabs/medrag-multi-modal")
retriever = BM25sRetriever().from_index(
index_repo_id="ashwiniai/medrag-text-corpus-chunks-bm25s"
)
llm_client = LLMClient(model_name=model_name)
medqa_assistant = MedQAAssistant(
llm_client=llm_client,
retriever=retriever,
top_k_chunks_for_query=5,
top_k_chunks_for_options=3,
)
dataset = weave.ref("mmlu-anatomy-test:v2").get()
with weave.attributes(
{"retriever": retriever.__class__.__name__, "llm": llm_client.model_name}
):
evaluation = weave.Evaluation(
dataset=dataset,
scorers=[MMLUOptionAccuracy()],
name="MMLU-Anatomy-BM25s",
)
summary = asyncio.run(
evaluation.evaluate(
medqa_assistant,
__weave={"display_name": evaluation.name + ":" + llm_client.model_name},
)
)
assert (
summary["MMLUOptionAccuracy"]["correct"]["true_count"]
> summary["MMLUOptionAccuracy"]["correct"]["false_count"]
)
def test_mmlu_correctness_anatomy_contriever(model_name: str):
weave.init("ml-colabs/medrag-multi-modal")
retriever = ContrieverRetriever().from_index(
index_repo_id="ashwiniai/medrag-text-corpus-chunks-contriever",
chunk_dataset="ashwiniai/medrag-text-corpus-chunks",
)
llm_client = LLMClient(model_name=model_name)
medqa_assistant = MedQAAssistant(
llm_client=llm_client,
retriever=retriever,
top_k_chunks_for_query=5,
top_k_chunks_for_options=3,
)
dataset = weave.ref("mmlu-anatomy-test:v2").get()
with weave.attributes(
{"retriever": retriever.__class__.__name__, "llm": llm_client.model_name}
):
evaluation = weave.Evaluation(
dataset=dataset,
scorers=[MMLUOptionAccuracy()],
name="MMLU-Anatomy-Contriever",
)
summary = asyncio.run(
evaluation.evaluate(
medqa_assistant,
__weave={"display_name": evaluation.name + ":" + llm_client.model_name},
)
)
assert (
summary["MMLUOptionAccuracy"]["correct"]["true_count"]
> summary["MMLUOptionAccuracy"]["correct"]["false_count"]
)
def test_mmlu_correctness_anatomy_medcpt(model_name: str):
weave.init("ml-colabs/medrag-multi-modal")
retriever = MedCPTRetriever().from_index(
index_repo_id="ashwiniai/medrag-text-corpus-chunks-medcpt",
chunk_dataset="ashwiniai/medrag-text-corpus-chunks",
)
llm_client = LLMClient(model_name=model_name)
medqa_assistant = MedQAAssistant(
llm_client=llm_client,
retriever=retriever,
top_k_chunks_for_query=5,
top_k_chunks_for_options=3,
)
dataset = weave.ref("mmlu-anatomy-test:v2").get()
with weave.attributes(
{"retriever": retriever.__class__.__name__, "llm": llm_client.model_name}
):
evaluation = weave.Evaluation(
dataset=dataset,
scorers=[MMLUOptionAccuracy()],
name="MMLU-Anatomy-MedCPT",
)
summary = asyncio.run(
evaluation.evaluate(
medqa_assistant,
__weave={"display_name": evaluation.name + ":" + llm_client.model_name},
)
)
assert (
summary["MMLUOptionAccuracy"]["correct"]["true_count"]
> summary["MMLUOptionAccuracy"]["correct"]["false_count"]
)
def test_mmlu_correctness_anatomy_nvembed2(model_name: str):
weave.init("ml-colabs/medrag-multi-modal")
retriever = NVEmbed2Retriever().from_index(
index_repo_id="ashwiniai/medrag-text-corpus-chunks-nv-embed-2",
chunk_dataset="ashwiniai/medrag-text-corpus-chunks",
)
llm_client = LLMClient(model_name=model_name)
medqa_assistant = MedQAAssistant(
llm_client=llm_client,
retriever=retriever,
top_k_chunks_for_query=5,
top_k_chunks_for_options=3,
)
dataset = weave.ref("mmlu-anatomy-test:v2").get()
with weave.attributes(
{"retriever": retriever.__class__.__name__, "llm": llm_client.model_name}
):
evaluation = weave.Evaluation(
dataset=dataset,
scorers=[MMLUOptionAccuracy()],
name="MMLU-Anatomy-NVEmbed2",
)
summary = asyncio.run(
evaluation.evaluate(
medqa_assistant,
__weave={"display_name": evaluation.name + ":" + llm_client.model_name},
)
)
assert (
summary["MMLUOptionAccuracy"]["correct"]["true_count"]
> summary["MMLUOptionAccuracy"]["correct"]["false_count"]
)
|