RYDE-WORK/lnp_ml
1
0
Fork 0
mirror of https://github.com/RYDE-WORK/lnp_ml.git synced 2026-09-18 15:03:20 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix: 修正bug

Browse Source
This commit is contained in:
Michelle0574 2026-08-16 13:51:46 +00:00
parent 4f84738c0f
commit 9810854ac7
2 changed files with 31 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
app/app.py
View File

@ -630,9 +630,7 @@ def main():
) )
all_results.append({"smiles": smiles, "results": results}) all_results.append({"smiles": smiles, "results": results})
# 为多 SMILES 模式添加 SMILES 标签 df = format_results_dataframe(results, smiles if is_multi_smiles else None)
smiles_label = smiles[:30] + "..." if len(smiles) > 30 else smiles
df = format_results_dataframe(results, smiles_label if is_multi_smiles else None)
all_dfs.append(df) all_dfs.append(df)
except httpx.HTTPStatusError as e: except httpx.HTTPStatusError as e:
@ -727,7 +725,7 @@ def main():
with col_export: with col_export:
smiles_used = st.session_state.get("smiles_used", "") smiles_used = st.session_state.get("smiles_used", "")
if isinstance(smiles_used, list): if isinstance(smiles_used, list):
smiles_used = ",".join(smiles_used) smiles_used = " | ".join(smiles_used)
csv_content = create_export_csv( csv_content = create_export_csv(
df, df,
@ -754,6 +752,13 @@ def main():
use_container_width=True, use_container_width=True,
hide_index=True, hide_index=True,
height=600, height=600,
column_config={
"SMILES": st.column_config.TextColumn(
"SMILES",
width="medium",
help="界面上按列宽省略显示，可拖动列头加宽；导出的 CSV 中为完整字符串",
),
},
) )
# 详细信息 # 详细信息

23
app/optimize.py
View File

@ -897,6 +897,7 @@ def optimize(
logger.info(f"Comp ranges: {comp_ranges.to_dict()}") logger.info(f"Comp ranges: {comp_ranges.to_dict()}")
seeds = None seeds = None
global_pool: List[Formulation] = []
_has_llm = getattr(model, "llm_prompt", None) is not None _has_llm = getattr(model, "llm_prompt", None) is not None
_do_rerank = bool(_has_llm and rerank_top_n and rerank_top_n > 0) _do_rerank = bool(_has_llm and rerank_top_n and rerank_top_n > 0)
@ -934,7 +935,12 @@ def optimize(
# 选择 top num_seeds 个种子点 # 选择 top num_seeds 个种子点
seeds = select_top_k(df, organ, num_seeds, scoring_weights) seeds = select_top_k(df, organ, num_seeds, scoring_weights)
logger.info(f"Selected {len(seeds)} seeds for next iteration") global_pool = select_top_k(
df, organ, max(num_seeds, top_k * 10), scoring_weights
)
logger.info(f"Selected {len(seeds)} seeds for next iteration "
f"(global pool: {len(global_pool)})")
else: else:
# ==================== 后续迭代：层级局部搜索 ==================== # ==================== 后续迭代：层级局部搜索 ====================
@ -999,7 +1005,20 @@ def optimize(
logger.info(f"Stage-1: {len(unique_results)} unique formulations (from {len(seeds)} candidates)") logger.info(f"Stage-1: {len(unique_results)} unique formulations (from {len(seeds)} candidates)")
if not _do_rerank: if len(unique_results) < top_k and global_pool:
n_before = len(unique_results)
for f in global_pool:
if len(unique_results) >= top_k:
break
key = f.unique_key()
if key not in seen_keys:
seen_keys.add(key)
unique_results.append(f)
if len(unique_results) > n_before:
logger.warning(
f"局部搜索收敛至 {n_before} 个不同配方（请求 {top_k} 个），"
f"已从粗筛池回填 {len(unique_results) - n_before} 个未细化候选"
)
return unique_results[:top_k] return unique_results[:top_k]
# ==================== 第二阶段：完整模型重排 ==================== # ==================== 第二阶段：完整模型重排 ====================