url/SkillOpt
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: rename remaining teacher/student refs, remove .gradio from repo

Browse Source 
- Fix teacher/student in deep_reflect, meta_reflect, sealqa, babyvision,
  mathverse, mmrb, swebench envs and prompt templates
- Remove .gradio/certificate.pem from tracked files
- Add .gradio/ to .gitignore

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This commit is contained in:
Cuzyoung
2026-05-24 19:22:20 +00:00
parent 7ae2d8766e
commit cff7ff6846
27 changed files with 148 additions and 178 deletions
Download Patch File Download Diff File Expand all files Collapse all files
skillopt/envs/sealqa/adapter.py
+2 -2
View File
@@ -105,11 +105,11 @@ class SealQAAdapter(EnvAdapter):
random_seed=kwargs.get('random_seed'),
step_buffer_context=kwargs.get('step_buffer_context', ''),
output_requirements=[
"- There is no hidden reference block. Use only the question, provided evidence, URL/fetch trace, student output, and evaluation result to infer what intermediate state is worth probing.",
"- There is no hidden reference block. Use only the question, provided evidence, URL/fetch trace, target output, and evaluation result to infer what intermediate state is worth probing.",
"- The instruction must explicitly request a short <analysis>...</analysis> block before the final <answer>...</answer>.",
"- The readout should focus on effective time frame, conflicting evidence, decisive source, candidate answer, and answer-finalization rule.",
"- Do not ask for exhaustive web summaries or a full chain-of-thought.",
"- The instruction text should be ready to append directly to the student's prompt.",
"- The instruction text should be ready to append directly to the target's prompt.",
],
metadata_builder=lambda item: {
"id": str(item.get('id')),
skillopt/envs/sealqa/evaluator.py
+1 -1
View File
@@ -64,7 +64,7 @@ def _build_grader_client() -> tuple[OpenAI | AzureOpenAI, str]:
openai_key = os.environ.get('OPENAI_API_KEY', '').strip()
api_key = azure_key or openai_key
if endpoint and api_version and api_key:
model = os.environ.get('SEALQA_GRADER_AZURE_MODEL', '').strip() or os.environ.get('SEALQA_GRADER_MODEL', '').strip() or os.environ.get('AZURE_MODEL_NAME', '').strip() or os.environ.get('TEACHER_DEPLOYMENT', '').strip() or 'gpt-5.4'
model = os.environ.get('SEALQA_GRADER_AZURE_MODEL', '').strip() or os.environ.get('SEALQA_GRADER_MODEL', '').strip() or os.environ.get('AZURE_MODEL_NAME', '').strip() or os.environ.get('OPTIMIZER_DEPLOYMENT', '').strip() or 'gpt-5.4'
client = AzureOpenAI(api_key=api_key, api_version=api_version, azure_endpoint=endpoint.rstrip('/'))
return client, model
skillopt/envs/sealqa/rollout.py
+14 -14
View File
@@ -7,8 +7,8 @@ from concurrent.futures import ThreadPoolExecutor, as_completed
from skillopt.envs.sealqa.evaluator import score_sealqa
from skillopt.envs.sealqa.tool_runtime import web_fetch
from skillopt.model import chat_student, get_student_backend, is_student_exec_backend
from skillopt.model.codex_harness import prepare_workspace, render_skill_md, run_student_exec
from skillopt.model import chat_target, get_target_backend, is_target_exec_backend
from skillopt.model.codex_harness import prepare_workspace, render_skill_md, run_target_exec
from skillopt.prompts import load_prompt
_FINAL_RE = re.compile(r"<answer>(.*?)</answer>", re.IGNORECASE | re.DOTALL)
@@ -83,11 +83,11 @@ def _run_codex_once(
task_text=final_task_text,
)
prompt = (
"Use the `skillopt-student` skill available in this workspace.\n"
"Use the `skillopt-target` skill available in this workspace.\n"
"Read `task.md`, answer the SealQA question using the provided evidence,\n"
"and return the final answer inside <answer>...</answer>."
)
final_message, raw = run_student_exec(
final_message, raw = run_target_exec(
work_dir=work_dir,
prompt=prompt,
model=model,
@@ -121,14 +121,14 @@ def process_one(
fail_reason = ''
try:
if is_student_exec_backend():
if is_target_exec_backend():
from skillopt.model import azure_openai as _llm
response, _raw, system, user_for_save = _run_codex_once(
pred_dir=pred_dir,
skill_content=skill_content,
task_text=user,
model=_llm.STUDENT_DEPLOYMENT,
model=_llm.TARGET_DEPLOYMENT,
timeout=120,
)
final_response = response
@@ -138,7 +138,7 @@ def process_one(
else:
user = user_for_save
else:
response, _ = chat_student(
response, _ = chat_target(
system=system,
user=user,
max_completion_tokens=768,
@@ -162,17 +162,17 @@ def process_one(
conversation.append({'type': 'tool_call', 'cmd': f'web_fetch({raw_url!r})', 'obs': fetched})
if fetched_blocks:
retry_user = user + '\n\n## Fetched URL Content\n' + '\n\n'.join(fetched_blocks)
if is_student_exec_backend():
if is_target_exec_backend():
retry_response, _raw, system, retry_user = _run_codex_once(
pred_dir=pred_dir,
skill_content=skill_content,
task_text=retry_user,
model=_llm.STUDENT_DEPLOYMENT,
model=_llm.TARGET_DEPLOYMENT,
timeout=120,
previous_response=final_response,
)
else:
retry_response, _ = chat_student(
retry_response, _ = chat_target(
system=system,
user=retry_user,
max_completion_tokens=768,
@@ -190,9 +190,9 @@ def process_one(
except Exception as e: # noqa: BLE001
fail_reason = f'error: {e}'
with open(os.path.join(pred_dir, 'student_system_prompt.txt'), 'w', encoding='utf-8') as f:
with open(os.path.join(pred_dir, 'target_system_prompt.txt'), 'w', encoding='utf-8') as f:
f.write(system)
with open(os.path.join(pred_dir, 'student_user_prompt.txt'), 'w', encoding='utf-8') as f:
with open(os.path.join(pred_dir, 'target_user_prompt.txt'), 'w', encoding='utf-8') as f:
f.write(user)
with open(os.path.join(pred_dir, 'conversation.json'), 'w', encoding='utf-8') as f:
json.dump(conversation, f, ensure_ascii=False, indent=2)
@@ -211,8 +211,8 @@ def process_one(
'fail_reason': fail_reason or ('' if score >= 1.0 else f"predicted '{final_answer}' but expected '{item.get('ground_truth', '')}'"),
'agent_ok': not fail_reason,
'n_turns': len(conversation),
'student_system_prompt': system,
'student_user_prompt': user,
'target_system_prompt': system,
'target_user_prompt': user,
}
return result