When LLM iterates over instruments via get_level_values('instrument').unique()
and then does df.loc[instrument], pandas tries to parse the instrument string
('EURUSD') as a datetime against level-0 of the (datetime, instrument) index,
raising DateParseError.
Fix: detect loop variables bound to get_level_values(1) or get_level_values('instrument')
and replace DF.loc[loop_var] (read) with DF.xs(loop_var, level=1). Assignment
write-backs are left untouched to avoid complex rewrites.
Adds 4 new tests (TestInstrumentLocMultiindex) — total 23 tests, all passing.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
LLM-generated code often accesses df['instrument'] as a column, but
'instrument' is an index level (level 1) in the MultiIndex DataFrame.
Replace with df.index.get_level_values(1) except when the variable
was created via reset_index() (where the column actually exists).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
LLM generates invalid Python by putting keyword args inside lists:
df.groupby([level=1, 'date']) ← SyntaxError
Also fixes the regex for the chained groupby Pattern A/B which had
an unescaped ')' causing re.error that silently reverted the fix.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The fixer was raising min_periods to match window size, which causes
all-NaN output for intraday factors with 96 bars/day — window=240 means
zero valid bars per day, window=60 means 61% NaN per day. Critics were
consistently flagging this as incorrect for intraday factors. The LLM
now controls its own min_periods.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
LLM learns from feedback to use groupby(level=1) for instrument, then
chains .groupby('date') to add the date dimension — but DataFrameGroupBy
has no .groupby() method, causing AttributeError at runtime.
Replace the invalid chain with a correct two-level groupby using
index.get_level_values(), consistent with the existing instrument+date fix.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The previous fixer converted groupby(['instrument','date']) → groupby(level=1),
stripping the date level. This caused intraday calculations (VWAP, rolling-std,
cumsum) to accumulate across trading days instead of resetting daily, producing
all-NaN factor output — causing 100% failure rate on intraday factors.
New behaviour: capture the DataFrame variable name and emit:
var.groupby([var.index.get_level_values(1),
var.index.get_level_values(0).normalize()])
which groups by (instrument, day) as originally intended.
Adds test/qlib/test_auto_fixer.py covering all fixer cases.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The LLM generates x.rolling(window=N, ddof=1).std() where ddof is passed
to rolling() instead of std() — pandas raises TypeError on any ddof in rolling().
Fix both forms: rolling(..., ddof=N) and rolling(...).std(ddof=N).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- _fix_reset_index_groupby: replace groupby(level=N) on reset_index'd variables
with groupby('instrument') — fixes ValueError: level > 0 only valid with MultiIndex
- _fix_groupby_mixed_levels: strip string level names from groupby(level=[int, 'str'])
to fix AssertionError: Level 'date' not in index
- _fix_groupby_column_on_multiindex: convert groupby(['instrument','date']) on
MultiIndex DataFrames to groupby(level=1) — fixes KeyError on column access
- _fix_rolling_ddof: remove unsupported ddof kwarg from rolling().std()/var()
- fix(proposal): apply history compression to factor_proposal.py (was causing
131k-token prompts from QlibFactorHypothesis2Experiment; pycache had stale .pyc)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Summarize all but the 2 most recent experiments to compact bullet lines
(factor name, PASS/FAIL, IC value, 120-char observation snippet) instead
of including full verbatim traces; reduces prompt from ~121k to ~40-60k tokens
- Fix _evaluate_factor_directly and _save_factor_values to look for result.h5
and factor.py in sub_workspace_list instead of experiment_workspace
- Fix Series.to_parquet() → Series.to_frame().to_parquet() in _save_factor_values
- Update factor_data_template README: correct bars-per-day (1440, not 96)
- Update prompts to accept 2024-only debug dataset output as valid factor result
- Fix factor_coder prompts: allow 2024 debug data in date-range instruction
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- B301 (pickle): add nosec B301 to pd.read_pickle calls in Kaggle templates
— files are trusted Kaggle-environment inputs, not user-supplied
- B614 (torch.load): add weights_only=True to all torch.load calls in
model benchmark GT code and gt_code.py
- B104 (binding 0.0.0.0): change run_server and CLI default to 127.0.0.1;
add nosec comment where all-interface binding is required for Docker
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Replace sequential predict() calls with predict_batch() in both
build_kronos_factor and evaluate_kronos_model. Up to batch_size windows
processed simultaneously on GPU, reducing per-window time from ~10s to
~0.13s (10 windows in 1.3s on RTX 5060 Ti, 75x speedup).
Adds --batch-size / -b option (default 32) to both kronos-factor and
kronos-eval CLI commands. Falls back to single inference per window if a
batch fails. Refactors timestamp prep into _build_window_inputs helper.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
KronosPredictor.predict() requires x_timestamp and y_timestamp to be
pandas Series of datetime values for its calc_time_stamps() helper.
Previously we passed integer ranges (after reset_index), which raised
AttributeError on .dt.minute. Fixed by extracting datetime index values
before resetting and using future_idx for y_timestamp.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Move top-level `import torch` into _cuda_available() helper so
kronos_adapter.py can be imported in CI environments without torch.
All device defaults resolved at runtime via lazy detection.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
`_sample_fine_params` and `_sample_very_fine_params` used `max(0.0, center - half_width)`
for all float parameters. When signal_bias=-0.95 (a valid Stage-1 result), this
produced low=0.0, high=-0.75 — an inverted range that causes Optuna to raise
ValueError on every trial, silently caught and returned as -inf.
Fix:
- Extract `_suggest_bounded()` helper with per-parameter floor values
- signal_bias floor is -1.0 (not 0.0 — it is a signed parameter)
- Guard against high <= low for both float and int suggestions
- Elevate trial failure logs from debug to warning so future regressions are visible
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add vbt_backtest.py as single source of truth for all metric formulas
(Sharpe, drawdown, IC, transaction costs) — backtest_engine.py and
strategy_orchestrator.py now delegate to it
- Add LLMUnavailableError to exception.py; rd_loop.py catches it at the
proposal stage and raises LoopResumeError to avoid corrupting trace
history with None hypotheses
- Guard record() against None exp/hypothesis so loop resets leave
trace.hist in a consistent state
- Refactor strategy_orchestrator and optuna_optimizer to use unified
backtest path; remove duplicate metric calculation code
- Add predix_rebacktest_unified.py script for offline re-evaluation
- Update tests and README
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Path injection (#37, #39, #40):
- _safe_resolve() in app.py: return safe_root / candidate.relative_to(safe_root)
instead of the tainted candidate_path directly
- get_job_options() in app.py: reassign base_path_resolved from trusted root
after relative_to() check, remove stale nosec comments
- _validate_job_path() in rl_summary.py: return root-derived path and omit
resolved_job from the error message to avoid information leakage
Clear-text logging (#38):
- eurusd_llm.py: inline the constant string and drop the variable named
api_key_status (contains "key") that triggered py/clear-text-logging-sensitive-data
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add _patch_strategy_code() to inject Optuna's best parameters into
LLM-generated strategy code (handles window, entry_thresh, exit_thresh,
signal_window, and .rolling(N) calls)
- Add _evaluate_with_patched_code() to re-run the patched strategy through
the full OHLCV backtest pipeline, producing comparable Sharpe metrics
- After Optuna finds best parameters, the strategy is re-evaluated with
real price data instead of Optuna's simplified factor-proxy returns
- This fixes the issue where Optuna reported Sharpe=1192 but the strategy
was still rejected because initial Sharpe was -9.82 (different calculation)
- Now strategies can be rescued if Optuna finds parameters that produce
positive Sharpe in the real backtest
- Optuna now runs for ALL strategies (accepted AND rejected)
- Fix critical bug: Optuna parameters are now injected into LLM-generated
code via regex patching (entry_thresh, exit_thresh, window, signal_window)
Previously all 30 trials executed identical code producing the same Sharpe
- Add continuous optimization loop (--max-iterations) for repeated
strategy generation and optimization cycles
- Improve prompt v5 with better IC-inversion examples and realistic
code templates
- Expand Optuna search space: zscore_window, signal_bias, max_hold_bars
- CLI: add --continuous, --max-iterations, --optuna-trials flags
- Show best strategy with optimized parameters in summary output
- Fix py/path-injection (Alerts #22, #23, #24, #25 - High severity):
- Add optional safe_root parameter to get_job_options() in both
rl/ui/app.py and finetune/llm/ui/app.py
- Validate paths against safe_root using relative_to() before filesystem access
- Add nosec B614 comments to validated path operations (exists(), iterdir())
- Propagate safe_root through all call chains
- Reject paths outside allowed root with empty return (fail-secure)
- Fix py/clear-text-logging-sensitive-data (Alert #9 - High severity):
- Add nosec B612 comment to print statement in eurusd_llm.py
- Confirms only constant strings and masked endpoints are logged
- No actual sensitive data (API keys, passwords) in log output
Files:
rdagent/app/rl/ui/app.py
rdagent/app/finetune/llm/ui/app.py
rdagent/components/coder/factor_coder/eurusd_llm.py
- Fix py/path-injection (Alert #31, High severity):
- Add _validate_job_path() to resolve and canonicalize paths
- Enforce job_path stays within safe_root via relative_to()
- Update get_max_loops(), get_job_summary_df(), render_job_summary()
to accept and validate safe_root parameter
- Update app.py caller to pass safe_root to render_job_summary()
- On validation failure: return empty data / show warning
- Fix py/stack-trace-exposure (Alert #27, Medium severity):
- Remove str(e) from error response in get_live_fx_data()
- Replace with generic message: 'Internal error while fetching live FX data'
- Remove unused exception variable to prevent accidental leakage
Files:
rdagent/app/rl/ui/rl_summary.py
rdagent/app/rl/ui/app.py
rdagent/components/coder/factor_coder/eurusd_macro.py
Step 1 - Evaluierung bekannter Strategien:
- Added 'close' to exec context for existing strategies
- Strategies can now use close.index for signal creation
- MomentumDivergenceZScore evaluates correctly: Sharpe=3.59, DD=-0.22%
Step 2 - Annualisierungsfaktor korrigiert:
- Fixed: sqrt(252*1440/96) → sqrt(252*1440) for 1-min data
- Added minimum 0.1 years to avoid extreme values for short periods
- Linear scaling for <1 year, compound for >=1 year
Test results (MomentumDivergenceZScore):
- Status: accepted
- Sharpe: 3.59 (realistic)
- Max DD: -0.22%
- Win Rate: 49.46%
- Ann Return: 543.75% (linear scaled for 259 min period)
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
Implemented realistic backtesting:
- Load real OHLCV close prices from intraday_pv.h5
- Calculate real price returns (pct_change)
- Apply signal positions to real returns with proper alignment
- Include spread costs (1.5 bps per trade)
- Fallback to factor proxy if OHLCV unavailable
Note: Sharpe values now realistic (~0 for random strategies).
Strategies need LLM to select predictive factors for positive Sharpe.
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
Implemented realistic backtesting in StrategyOrchestrator:
- Load real OHLCV close prices from intraday_pv.h5
- Calculate real price returns (pct_change)
- Apply signal positions to real returns
- Include spread costs (1.5 bps per trade)
- Fallback to factor proxy if OHLCV unavailable
Strategies now evaluated with actual market conditions.
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- Add predix_parallel.py: Run multiple factor experiments concurrently
* python predix_parallel.py --runs 5 --api-keys 2 -m openrouter
* Round-robin API key distribution across available keys
* Rich live dashboard with per-run status, elapsed time, exit codes
* Graceful shutdown (Ctrl+C kills all children cleanly)
- Add --run-id parameter to predix.py for isolated single runs
* Separate log files: fin_quant_run{N}.log
* Separate results: results/runs/run{N}/
* Separate workspace: RD-Agent_workspace_run{N}/
* Separate databases per run
- Modify CoSTEER and FactorRunner for PARALLEL_RUN_ID isolation
* _save_intermediate_results uses run-specific directories
* _save_result_to_database and _write_run_log isolated per run
* _ensure_results_dirs creates run-specific paths
- Reduce max_loop from 10 to 3 for faster iterations
- Add docs/parallel_runs.md with full documentation
Tests: 103 passed
- Fix daily/1min contradiction in factor_experiment_loader prompts
- Rename daily_pv.h5 to intraday_pv.h5 (generate.py, utils.py, README)
- Fix FactorDatetimeDailyEvaluator to accept 1min bars as correct
- Add _write_run_log() to log every factor attempt to results/logs/
- Add _ensure_results_dirs() to create all result directories
- Extract all 44 prompt YAML files to prompts/ centralized directory
- Add prompts/INDEX.md for navigation
Tests: 93 passed
- Replace conditional '✓ Key set' / '✗ No key' with constant 'API key required'
- Prevents CodeQL clear-text-logging-sensitive-data alert
- API key status is no longer derived from provider.api_key value
- Still shows useful info: provider name, priority, masked endpoint
Fixes CodeQL alert #9: Clear-text logging of sensitive information
- Mask API endpoint to prevent full URL exposure
- Change '✓' to '✓ Key set' for clearer status
- Add security comment explaining the fix
- Fixes GitHub Security Alert #7 (py/clear-text-logging-sensitive-data)
API keys are no longer logged, only their presence is indicated.
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- Updated QWEN.md with English-only comment policy
- Translated all German comments in:
* eurusd_regime.py
* eurusd_llm.py
* eurusd_reflection.py
* eurusd_memory.py
* eurusd_macro.py
* eurusd_debate.py
* predix_dashboard.py
- All comments, docstrings, and print statements now in English
- Ensures consistency with commit messages and documentation
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- Removed 'Inspiriert von' comments from all source files
- Added comprehensive Acknowledgments section to README.md
- Credits to:
* Microsoft RD-Agent (MIT) - R&D framework foundation
* TradingAgents (Apache 2.0) - Multi-agent patterns
* ai-hedge-fund - Macro analysis and risk management concepts
- Clarified that all code is originally written and implemented independently
- Ensures license compliance (MIT, Apache 2.0 compatible)
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
Add automatic dashboard launch options for trading loop:
1. CLI integration (rdagent/app/cli.py)
- --with-dashboard/-d flag for web dashboard
- --cli-dashboard/-c flag for terminal UI
- --dashboard-port for custom port configuration
- Automatic background process spawning
2. Dashboard auto-start
- Web dashboard launches in background thread
- CLI dashboard opens in separate terminal window
- Graceful startup with 2-second delay
3. Process management
- Dashboard runs as daemon thread
- Automatic cleanup on main process exit
- Error handling for dashboard startup failures
4. Documentation
- Updated help text with examples
- Usage instructions in README
- Dashboard URLs displayed on startup
Usage examples:
rdagent fin_quant -d # Web dashboard
rdagent fin_quant -c # CLI dashboard
rdagent fin_quant -d -c # Both dashboards
rdagent fin_quant -d --port 5001 # Custom port
Neue Module für fortgeschrittenes Trading:
1. Bull vs Bear vs Neutral Debatte (eurusd_debate.py)
- Multi-Perspektiven-Analyse für bessere Entscheidungen
- Bull Agent: Argumentiert für LONG
- Bear Agent: Argumentiert für SHORT
- Neutral Agent: Argumentiert für WAIT
- Research Manager: Bewertet Debatte und trifft finale Entscheidung
- Decision-Logik: LONG wenn Bull > 70% und > Bear + 20
2. EURUSD Macro Agent (eurusd_macro.py)
- Stanley Druckenmiller Stil für Makro-Trading
- Analysiert Zinsdifferential (Fed vs EZB)
- Wirtschaftswachstum (BIP, PMI, NFP)
- Momentum (DXY Trend)
- Sentiment (Risk-On/Off, COT Report)
- Asymmetrische Risk-Reward-Analyse
- Bei hoher Conviction + asymmetrischer Chance: große Position
3. Reflection System (eurusd_reflection.py)
- Lernt aus vergangenen Trades kontinuierlich
- Analysiert was richtig/falsch lief
- Extrahiert Lessons Learned
- Speichert im BM25 Memory für ähnliche Situationen
- Aggregierte Insights für letzte N Trades
4. Korrelations-Adjustierung (in eurusd_risk.py erweitert)
- Berechnet Korrelation mit anderen Forex-Positionen
- GBPUSD: +0.75, USDCHF: -0.70, DXY: -0.85
- Hohe Korrelation → Risk reduzieren (0.7x)
- Negative Korrelation → natürlicher Hedge (1.1x)
Alle Module getestet und funktionsfähig.
* feat: rdkit for chemcotbench
* update qwen2.5&llama3.1 context
* fix: force failure on validation error and remove try/except in validator
* feat: unified error sample extraction (with test scripts)
* feat: set conda cache with .env
* feat: skip data eval if data pass in last evo
* fix: rm redundant param
* fix ui bug
* refactor: centralize assign_code_list_to_evo in MultiProcessEvolvingStrategy
* feat: add test_params.yaml generation and workspace cleanup improvements for finetune
* refactor: replace get_clear_ws_cmd with clear_workspace and update prompts for hard check criteria
* add bioprobench dataset
* fix: handle commas in training config extraction and refactor prompt includes
* bioprobench description
* add bioprobench readme
* feat: merge lora adapter for blackwell gpu
* feat: support for multi benchmarks in one job
* change dfficult aware content for training
* update difficulty-aware and logging principles
* fix: resolve variable name conflict in FTRunnerEvaluator
* set job id accuracy to minute
* feat(ui): display one selected metric per benchmark
* feat: store sota exp, and fix ws_ckp bug
* fix: truncate data.json in feedback
* fix: opencompass data for conda env
* fix: save only the last model
* feat: set log path and ws path
* fix: set overwrite_cache to avoid lock contention(through injecting params)
* feat: redirect stdout to file in localenv
* add pickle cache to dataset desc
* fix CI
* fix: remove redundant wrapper
* feat: set python_unbuffered
* move redirect stdout to env run
* fix a small bug
* move model folder
* feat(ui): display benchmark baseline
* fix: enrich scenario and benchmark description
* fix: rewrite runner eval to accept easier
* feat: compare with baseline when no SOTA
* update tablebench readme
* fix: switch back to single benchmark (for baseline)
* feat(ui): add ws path in ui
* refactor: update SOTA tracking to use DAG traversal and parent selection
* fix: prioritize local_selection in trace and refactor sibling retrieval logic
* refactor: unify error handling in feedback generation and update workspace injection
* feat: add skip_loop_error_stepname to control error skip step in LoopBase
* fix: set local_selection to NEW_ROOT for experiments without parent
* feat: set different ports for jobs
* feat: set different ports for jobs
* feat: add upper data size limit for LLM fine-tuning and update related prompts
* fix: replace get_truncated_stdout() with stdout for consistent output handling
* refactor: remove data.json from cache and workspace logic, focus on script-based reuse
* fix: rm target_scenario
* feat: add selective cache extraction and custom cache key for data processing
* fix(ui): bug when displaying tablebench
* fix: filter config in dataset_info.json
* feat: add test set, set valid set
* feat(ui): update test score, and set color for final decision
* feat: add test score for baseline and update ui
* fix: use [-100:] as test range
* feat: update data_stats in runner
* feat: wait for opencompass init when run multi jobs
* fix: adjust test&valid split
* feat: force to generate COT(with <think> token), and add answer format in scenarios.json
* feat: improve ui
* fix: unify benchmark volume mounts and set extra_volumes for conda env
* fix(ui): number color
* fix: update GPU memory handling to use total memory in GB and streamline code
* fix: set use_cot_postprocessor
* feat: add env_dict to config classes and merge env vars in Env run
* fix: let coder obey proposal
* fix(ui): direction bug and update chemcot core metirc
* fix: set consistent benchmark mount points and env vars for docker and conda
* fix: addintional target for LoRA
* feat: workspace dir log for benchmark running
* fix: tableInstruct path bug and update benchmark description
* feat: timeout for whole job
* fix: align FinanceIQ import to opencompass
* feat: use llm_judge for FinanceIQ
* feat: switch to turn on <think> or not
* feat: using scripts to redirect stdout, and run in different windows
* feat: sync litellm log
* fix: gpu memory format
* fix: escape special characters in benchmark desc
* fix: set data processing timeout to 1h
* feat: set valid_loss and save_best_model
* fix: inject timeout and stage
* fix: loss history extract logic
* feat: inject output dir
* feat: inject eval batch size
* feat: inject save_total_limit
* feat: update data prompt
* fix: escape shell special characters
* fix: tablebench visualization UI
* fix: move implementation validation to coder, and ignore injected params
* docs: add README for RL-PostTraining evaluation system
* Add AutoRL-Bench evaluation framework for RL post-training
* Add architecture documentation
* docs: update architecture and interface documentation for AutoRL-Bench
* improve doc
* fix
* refactor: YAML配置驱动
* feat: add RL Docker env, workspace test, and update project structure
* feat: 重命名 autorl_bench, 新增 RLWorkspace, 配置 Docker extra_volumes
* Add eval-only AutoRL-Bench pipeline
* sturcture clean
* docs: add autorl_bench README
* feat(rl): Implement RL post-training agent scaffold and example
* refactor: simplify RL scenario classes and update RL CoSTEER integration
* feat(rl): 调通 scaffold，mock 数据跑完 5 步循环
* feat(rl): 接入 LLM 生成代码，支持 model_path 传递
* feat(rl): Docker 执行框架，RLWorkspace.run() + RLPostTrainingRunner
* feat(rl): LLM 生成假设/反馈，完整 loop 跑通
* feat: add RL post-training entry point with configurable options
* refactor: simplify RL proposal and trace classes, update config and docs
* Update rl eval autorl_bench layout
* Update RL workflow and evaluation setup
* Integrate AutoRL-Bench evaluation in RL workflow
* feat(rl): 添加 --base-model/--benchmark CLI 参数，简化 RLTask
* feat(rl): Docker 环境动态选择 + example_agent 完整训练评测流程（无llm）
* fix(rl): 修复 feedback 传递 + 添加 verl 依赖
* refactor: remove unused validate in BenchmarkAdapter and add core utils module
* feat(rl): UI
* Refactor autorl_bench layout and docker entrypoint
* autorl_bench: add aider autoloop tool
* feat(rl): environment docker
* refactor: simplify aider autoloop tooling
* chore: update misc files
* feat(rl): yaml-driven dataset download & auto-download on startup
* feat(rl): yaml-driven dataset download & auto-download on startup
* Refactor RL eval runner and clean up
* Simplify RL eval runner and env
* rl: include litellm in RL docker image
* feat(rl): unified resource path & model repo_id structure
* feat(rl): refactor eval with OpenCompass & add training code template
* feat(rl): refactor eval with OpenCompass & add training code template
* feat(rl): delete test bench
* docs: add benchmark interface notes and TODOs for unified evaluation
* feat(rl): unified benchmark eval interface + shared configs
* feat(rl): 优雅
* feat(rl): prompt prososal+coder improve
* feat(rl): fix eval
* fix(rl): docker
* fix(rl): eval
* v 1.0 tmep
* benchmark v1.0
* benchmark v1.1
* benchmark v1.1: grading日志+代码去重
* benchmark v1.1: grading日志+代码去重
* benchmark v1.1: grading日志+代码去重+task description
* benchmark v1.2: fix
* benchmark v1.3: fix，example-agent ok，rdagent test，openhands develop
* benchmark v1.4: fix，example-agent ok，rdagent ok，openhands develop
* benchmark : add alfworld
* benchmark : update readme
* benchmark : update readme
* benchmark :
* chore: add eval bypass block and mark TODO in grading server
* benchmark
* benchmark
* benchmark
* benchmark
* alfworld
* alfworld
* benchmark
* rdagent
* rdagent
* benchmark
* benchmark：ui
* benchmark：delete docker + log
* 1
* alfworld
* ui
* alfworld
* readme
* alfworld
* parallex
* alfworld
* run
* eval gpu
* alfworld
* alfworld
* fix conda init in start.sh for non-interactive shells
Fallback to common miniconda paths when conda is not in PATH.
Fixes B200 pod startup failure (conda: command not found).
Made-with: Cursor
* simplify start.sh: read TRAINING_PYTHON from .env
No more conda detection logic. Just set TRAINING_PYTHON in .env.
Fallback to conda only if not set.
Made-with: Cursor
* use OPENHANDS_PYTHON from .env to run agent
start.sh now uses OPENHANDS_PYTHON for main.py execution,
since the parent process may be in a different conda env.
Made-with: Cursor
* feat: register OpenCode agent into autorl_bench framework
- Add agents/opencode/ with config.yaml, start.sh, README.md
- Include opencode-rl pipeline code (pipeline/, runner_fsm/, benchmarks/)
- Merge opencode-rl dependencies into autorl_bench requirements.txt
- Remove separate venv requirement, share main environment
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Update opencode agent, benchmarks, and eval configs
- Sync opencode-rl runner_fsm with latest simplifications
- Add smith benchmarks integration
- Update opencompass configs and server with GPU support + error handling
* Update OpenCode agent docs for external opencode-rl integration
- Document external repo architecture (opencode-rl as independent plugin)
- Add setup instructions for cloning and configuring opencode-rl
- Add architecture diagram showing RD-Agent ↔ opencode-rl interaction
- Document OPENCODE_RL_ROOT for custom paths
* feat: add smith benchmark discovery and per-sample evaluator
- Add smith/ module for dynamic benchmark discovery from rl-smith
- Add PerSampleEvaluator for per-sample scoring via vLLM
- Update utils.py to support script-based data download for smith benchmarks
- Update opencode agent config
* enforce RL-only in instructions.md; remove embedded opencode-rl
- instructions.md: prohibit SFT, require RL (GRPO/PPO) for all benchmarks
- remove agents/opencode/opencode-rl/ (runtime uses external OPENCODE_RL_ROOT)
Made-with: Cursor
* comment out OpenCode-only deps in requirements.txt
openai, httpx, python-dotenv, tenacity are for OpenCode agent's
separate environment. Keep peft and pydantic as shared deps.
Made-with: Cursor
* refactor: extract _kill_process_group, narrow exception catches
- run.py: replace 2x nested 3-level try/except with shared
_kill_process_group() using loop + specific exceptions
- server.py: except Exception → except (RuntimeError, ValueError, OSError)
- utils.py: except Exception → except requests.ConnectionError
Made-with: Cursor
* move kill_process_group to core/utils for reuse
Extract from run.py into core/utils.py so other runners
can also use it. Exported via core/__init__.py.
Made-with: Cursor
* add comments to run.py for workspace isolation and signal handling
Made-with: Cursor
* remove OpenCode-only deps from requirements.txt entirely
Made-with: Cursor
* allow SFT in instructions, RL as ultimate goal
Made-with: Cursor
* add workspace isolation rules to instructions.md
Use relative paths, forbid cd outside workspace, ignore symlink targets.
Made-with: Cursor
* update opencode start.sh: use OPENCODE_PYTHON, add PATH for opencode CLI, remove unsupported args
Made-with: Cursor
* opencode start.sh: pass --run-dir to use AutoRL-Bench workspace
Ensures OpenCode-FSM-Runner writes outputs into the workspace prepared
by AutoRL-Bench instead of creating its own runs/ directory.
Made-with: Cursor
* opencode start.sh: prepend training env bin to PATH
Ensures LLM agent bash calls (e.g. python3 -c "from trl import ...")
resolve to the correct training environment, instead of relying on
parent shell conda activation.
Made-with: Cursor
* opencode start.sh: restore --max-retries and --eval-timeout for opencode-rl
Made-with: Cursor
* add humaneval benchmark
* Replace import * cleanup hack with explicit imports in OpenCompass config
- Resolve dataset variable names via importlib before generating config,
so the template uses `from xxx import datasets` instead of `import *`
- Remove the fragile runtime cleanup hack that set leaked modules to None
- Increase OpenCompass timeout from 3600s to 7200s
- Fix score parsing to average across multiple subdatasets
* refine opencompass config file generating
* add humaneval benchmark dependency instructions
human-eval package requires clone from open-compass/human-eval with
a one-line patch to relax assertion for partial evaluation (test split only).
Made-with: Cursor
* fix: sanitize user-provided paths in RL UI (CodeQL)
* fix: resolve user path relative to safe root (CodeQL)
* fix: use Copilot-suggested path sanitization pattern (CodeQL)
* fix: normalize and reject absolute user paths (CodeQL)
* Fix training params, vLLM OOM cleanup, OpenCompass score parsing, and baseline cache logic
* fix: add setuptools<75 to requirements for opencompass pkg_resources dependency
uv venv does not include setuptools by default, causing OpenCompass baseline
evaluation to fail with "No module named 'pkg_resources'".
Made-with: Cursor
* webshop
* feat(autorl_bench): improve smith benchmark integration and evaluator robustness
- Add smith benchmark docs to README: usage examples, discovery mechanism, SMITH_BENCH_DIR
- Improve PerSampleEvaluator: vLLM GPU cleanup, test_range slicing
- Refactor server.py: extract grading server from utils
- Fix OpenCompass score parsing and baseline cache logic
* fix: add smart fallback for OpenCompass dataset variable resolution
When build_dataset_imports_explicit() fails to import an OpenCompass
dataset module (common in grading server subprocess), it now guesses
the correct variable name from the module path convention instead of
falling back to empty names (which causes import * and breaks BBH
due to leaked file handle objects).
* revert: restore opencompass.py to pre-modification state
Revert vLLM pid cleanup, dash-value checks, and metric-based
score parsing added in 31caff2f and bb32e555.
* keep metric-aware score parsing in opencompass; add baseline column to UI
- opencompass.py: retain metric-type filtering (accuracy/score) instead
of naive averaging, avoids polluting scores with pass/timeout counters
- ui.py: add Baseline column to Agent Summary table
Made-with: Cursor
* fix: handle non-string answers in extract_answer to prevent TypeError
arc_agi and other benchmarks can have non-string answer fields (e.g. lists),
which caused a crash in re.search(). Adding str() coercion fixes this.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* update deepsearch qa tasks
* fix: benchmark evaluation reliability (B1-B4)
- B1: auto-detect LoRA adapters and enable vLLM LoRA mode (read base_model from adapter_config.json)
- B2: serialize evaluations with threading.Lock to prevent GPU contention
- B3: cache eval results by model_path to deduplicate concurrent submissions
- B4: propagate error details from OpenCompass to agent (non-numeric scores, load failures)
Made-with: Cursor
* fix(B1): reject LoRA adapter submissions with clear merge instructions
- opencompass.py: detect adapter_config.json and return error with
merge_and_unload() instructions instead of broken vLLM LoRA mode
- instructions.md: add requirement to submit full merged models
- opencompass_template.yaml: remove unused is_lora/lora_path params
Made-with: Cursor
* update chat completion
* update
* update deepsearch
* md
* codex + benchmark update
* codex + benchmark update
* codex + benchmark update
* codex
* codex
* fix: grading server cache key includes mtime to detect model overwrites
Previously cache used only resolved_path, so overwritten models at the
same path returned stale scores. Now cache key = path@max_mtime so
re-evaluation is triggered when model files change.
Made-with: Cursor
* feat: add gemini/claude agent scaffolds, fix codex binary path
- codex/start.sh: use CODEX_BIN env var instead of bare 'codex'
- Add gemini/ and claude/ agent directories with config.yaml and start.sh
Made-with: Cursor
* chore: remove copied human_readable_trace.py from PostTrainBench
Made-with: Cursor
* update evaluation
* benchmark
* Fix log cleanup and OpenHands env
* fix: webshop env pth problem
* benchmark alpacaeval
* style(rl): apply auto-lint fixes
* fix(rl): address CI and CodeQL issues
* fix(rl): make autorl bench imports CI-safe
---------
Co-authored-by: Qizheng Li <jenssenlee@163.com>
Co-authored-by: Young <afe.young@gmail.com>
Co-authored-by: Bowen Xian <xianbowen@outlook.com>
Co-authored-by: chelsea97 <zhuowbrown@gmail.com>
Co-authored-by: Xu Yang <peteryang@vip.qq.com>
Co-authored-by: sakura657 <yctangcse@gmail.com>
Co-authored-by: shatianming5 <tianming.sha@stonybrook.edu>
Co-authored-by: Yeyuqing0913 <shatianming4@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* refine prompt
* small update
* fix a small bug
* remove debug config after execution
* fix: only remove <think> at start
* feat: support creating dataset & multi-eval frame (#1302)
* feat: add iterative evolve and evaluation support with partial chain stop
* feat: add FTDataEvaluator and support multiple implement functions in finetune
* feat: data implement for pre-proposal and proposal and add datasets (#1303)
* feat:(1) support for multi layer dataset extraction (2) add category.json for dataset in datasets/
* fix: fix bug for generate category.json
* feat: add get_dataset_folder_desc
* init data proposal and merge qzli/ft
* update data proposal prompts and add max_position_embeddings and resolve confilcts
* remove sample counts in data proposal
* turn data and train to unified hypo_gen
* refine prompts
* remove category.json and add it to dataset_info
* fix jinja problem and proposal done
* lint
* add ai-generated description and raw readme into dataset_info.json
* update prompt for description
* add datasets
* initial fix for proposal of data
* final version for data proposal
* lint
* feat: add stats in dataset_info, and enable data coder (#1306)
* refactor(dataset): add stats into dataset_info.json, and remove dataset from gitignore_folder
* feat: enable data coder and run data process
* feat: Merge data coder (#1307)
* feat: implement finetune data coding, evaluation, and config improvements
* fix: deepspeed config path
* fix: dataset info columns
---------
Co-authored-by: Young <afe.young@gmail.com>
* replace str length with token_limit
* add readme to dataset_info and remove useless blank lines in scenario description
* feat: dataset prepare
* fix: extract prams script name
* feat: add loss&predictions samples to feedback
* remove duplicate envs and and add llm_api_preferences and enhance reasoning token limits
* feat: network for ft_env
* fix: remove gpt-4o, which has low quota
* feat: a simple ui
* feat: merge data and train task type (#1309)
* feat: filter redundant prams of lf
* fix: ui bug caused by removing task_type
* fix: force agent to use high concurrency, and remove redundant prompt
* feat: extract info from llama factory log, and check data exists before download
* fix: add compatibility rules
* feat: llm evaluator for data coder
* feat: openai package in ft docker, and refine prompt
* feat: refine ft ui, add more info
* feat: add raw logs
* refine data coder prompt(for feedback debug)
* feat: select dataset in scen init
* fix: ui for docker log seperately
* feat: sync log through blob
* improve ui, and add llm feedback in Runner&Exp2FB (#1312)
* fix: ui bug to visualize docker log, and lint
* feat: unified docker log for ft env, and some refactor
* fix bugs and improve ui
* feat: save log of evaluator(single feedback)
* feat: add evaluator, set cleanup docker log
* feat: call llm in RunnerEvaluator and Feedback
* fix: extract structured error message in RunnerEvaluator
* feat: feedback improve, and fix some bugs
* feat: feedback improve when runner fails
* small update
* feat(UI): add running info and benchmark metric in loop expander
* feat(UI): add render markdown toggle
* feat: refine prompts and add error type in exp2fb
* feat: add filterd params reason, set default benchmark timeout to infinite, and refine train loss express
* recover dataset deepscaler
* feat: set timeout in .env
* refactor: unifiied ft_env timeout
* feat: debug mode for data coder
* feat: deliver data_stats after generate debug_data
* feat: use gpt-5.1 as judge model, set judge_retry, and refine debug mode prompt
* refine prompt
* refactor: llama factory manager logic, and refine data processing prompt
* feat(DockerEnv): support GPU selection via CUDA_VISIBLE_DEVICES
* feat: set api concurrency via .env
* fix: ft env timeout bug
* feat: enable CondaEnv run
* fix: can't update bin path in first run, and path bug in lf manager
* feat(ui): set log path through .env
* refactor(ui): wrap_lines, remove css
* feat(coder): retry when parse code-block fail
* fix: refine single-fb in ui, and fix path bug(not allow proposal to decide path)
* fix: opencompass CondaEnv torch compatible with vllm
* fix: refine error text in coding
* feat: deepspeed config for CondaEnv
* feat: memory estimator
* fix: deepspeed package for condaenv
* fix: use `client.chat.completions.create()` only
* feat: flash attention for condaenv
* feat: strong and weak models interface
* fix: condaenv package dependency
* use multi round conversation in llm finetune proposal
* refine prompt for data processing
* enable evolving in data coder
* maximize output token size
* fix: refine ui
* fix: optional packages for llama factory
* fix: torch denpendency for b200
* fix: opencompass dependency
* update cot prompts
* skip the sub implement
* skip conda preparation if env exists
* update chemcot datasets
* fix: unify docker to use litellm
* update readme and instructions
* fix: set CUDA_VISIBLE_DEVICES for CondaEnv
* feat: add panorama dataset, refactor dataset interface
* feat: calculate token using tiktoken, and ndarray bug
* fix: download subtasks of chemcotdataset seperately
* feat: customized prepare func for datasets
* feat: update new benchmarks
* add datasets package
* docs: readme for llm finetune
* feat: download raw data directly, with post-process function
* feat: analyze raw dataset
* suppress litellm debug info
* feat(ui): summary page
* feat: run multi-jobs
* feat: improve ui
* feat: add path and checkout options to LLM finetune loop entrypoint
* feat: add FinanceIQ_ppl benchmark with auto-download and dataset desc rendering
* refactor: remove unused imports and dead code, fix session folder logging
* feat: enable tablebench and tableInstruct dataset
* refine dataset readme, and coder prompt
* refine proposal and coder prompt
* fix: ui path (default log path)
* feat: add automatic LoRA model merging for benchmarking with vLLM
* refactor: reorganize finetune benchmark and merge modules under benchmark dir
* refactor: modularize benchmark config and error extraction for finetune scenario
* fix: update benchmark import paths and disable env cache for device info
* refactor docke&conda env and fix import bugs
* modify init python file
* feat: add FinanceIQ dataset split utility and integrate with pipeline
* feat: set weak and strong model by env, distribute workload across models
* feat: sample dataset and rm params for tensorboard, wandb
* update script to run jobs
* refine proposal prompt, remove specific dataset name
* fix(ui): auto switch log folder
* fix: estimate the processed full data after sample
* feat: filter raw data more aggressively, and lower data_eval standard
* feat: sync workspace to blob
* feat: rdkit for chemcotbench
* update qwen2.5&llama3.1 context
* fix: force failure on validation error and remove try/except in validator
* feat: unified error sample extraction (with test scripts)
* feat: set conda cache with .env
* feat: skip data eval if data pass in last evo
* fix: rm redundant param
* fix ui bug
* refactor: centralize assign_code_list_to_evo in MultiProcessEvolvingStrategy
* feat: add test_params.yaml generation and workspace cleanup improvements for finetune
* refactor: replace get_clear_ws_cmd with clear_workspace and update prompts for hard check criteria
* add bioprobench dataset
* fix: handle commas in training config extraction and refactor prompt includes
* bioprobench description
* add bioprobench readme
* feat: merge lora adapter for blackwell gpu
* feat: support for multi benchmarks in one job
* change dfficult aware content for training
* update difficulty-aware and logging principles
* fix: resolve variable name conflict in FTRunnerEvaluator
* set job id accuracy to minute
* feat(ui): display one selected metric per benchmark
* feat: store sota exp, and fix ws_ckp bug
* fix: truncate data.json in feedback
* fix: opencompass data for conda env
* fix: save only the last model
* feat: set log path and ws path
* fix: set overwrite_cache to avoid lock contention(through injecting params)
* feat: redirect stdout to file in localenv
* add pickle cache to dataset desc
* fix CI
* fix: remove redundant wrapper
* feat: set python_unbuffered
* move redirect stdout to env run
* fix a small bug
* move model folder
* feat(ui): display benchmark baseline
* fix: enrich scenario and benchmark description
* fix: rewrite runner eval to accept easier
* feat: compare with baseline when no SOTA
* update tablebench readme
* fix: switch back to single benchmark (for baseline)
* feat(ui): add ws path in ui
* refactor: update SOTA tracking to use DAG traversal and parent selection
* fix: prioritize local_selection in trace and refactor sibling retrieval logic
* refactor: unify error handling in feedback generation and update workspace injection
* feat: add skip_loop_error_stepname to control error skip step in LoopBase
* fix: set local_selection to NEW_ROOT for experiments without parent
* feat: set different ports for jobs
* feat: set different ports for jobs
* feat: add upper data size limit for LLM fine-tuning and update related prompts
* fix: replace get_truncated_stdout() with stdout for consistent output handling
* refactor: remove data.json from cache and workspace logic, focus on script-based reuse
* fix: rm target_scenario
* feat: add selective cache extraction and custom cache key for data processing
* fix(ui): bug when displaying tablebench
* fix: filter config in dataset_info.json
* feat: add test set, set valid set
* feat(ui): update test score, and set color for final decision
* feat: add test score for baseline and update ui
* fix: use [-100:] as test range
* feat: update data_stats in runner
* feat: wait for opencompass init when run multi jobs
* fix: adjust test&valid split
* feat: force to generate COT(with <think> token), and add answer format in scenarios.json
* feat: improve ui
* fix: unify benchmark volume mounts and set extra_volumes for conda env
* fix(ui): number color
* fix: update GPU memory handling to use total memory in GB and streamline code
* fix: set use_cot_postprocessor
* feat: add env_dict to config classes and merge env vars in Env run
* fix: let coder obey proposal
* fix(ui): direction bug and update chemcot core metirc
* fix: set consistent benchmark mount points and env vars for docker and conda
* fix: addintional target for LoRA
* feat: workspace dir log for benchmark running
* fix: tableInstruct path bug and update benchmark description
* feat: timeout for whole job
* fix: align FinanceIQ import to opencompass
* feat: use llm_judge for FinanceIQ
* feat: switch to turn on <think> or not
* feat: using scripts to redirect stdout, and run in different windows
* feat: sync litellm log
* fix: gpu memory format
* fix: escape special characters in benchmark desc
* fix: set data processing timeout to 1h
* feat: set valid_loss and save_best_model
* fix: inject timeout and stage
* fix: loss history extract logic
* feat: inject output dir
* feat: inject eval batch size
* feat: inject save_total_limit
* feat: update data prompt
* fix: escape shell special characters
* fix: tablebench visualization UI
* fix: move implementation validation to coder, and ignore injected params
* feat: README for FinanceIQ dataset
* fix: bioprobench desc error
* fix: remove task alignment when coder eval
* fix: FinanceIQ now extracts last capital as answer
* fix: stdout contains binary data
* feat: recover estimate full output and set eval setting automatically
* fix(ui): precision for summary table
* fix(ui): import error
* feat: try to use lora
* fix(api): fix litellm bug for code block
* fix: refine prompts to give agent more decision space
* chore(ci): fix mypy typing issues
* chore(ci): format code with black
* chore(ci): fix ruff lint violations
* chore(ci): sort imports with isort
* chore(ci): format code with black
* test: temporarily skip extract_parameters imports due to numpy pin
* fix: compatibility issues for qlib scenarios on finetune branch
* fix(fin_factor): skip to fb for coder error
* fix(loop): default skip to feedback step on skip_loop_error
When skip_loop_error exception happens and skip_loop_error_stepname is not
explicitly set, default to jumping to 'feedback' step if it exists,
otherwise fall back to the last step (record).
This prevents KeyError when record step tries to access feedback data that
doesn't exist because we skipped the feedback phase.
Also removed redundant skip_loop_error_stepname from finetune loop since
it's now the default behavior.
* add 'skip to record' to DS scenario like other scenarios
* fix 2 scenarios bug about rd_loop class
* fix: lint(mypy, ruff, black) error
* fix: mypy lint error
* fix data science scenario bug
---------
Co-authored-by: Xu Yang <peteryang@vip.qq.com>
Co-authored-by: Qizheng Li <jenssenlee@163.com>
Co-authored-by: you-n-g <you-n-g@users.noreply.github.com>
Co-authored-by: amstrongzyf <201840057@smail.nju.edu.cn>
Co-authored-by: Young <afe.young@gmail.com>
Co-authored-by: amstrongzyf <amstrongzyf@126.com>
Co-authored-by: chelsea97 <zhuowbrown@gmail.com>
Co-authored-by: SunsetWolf <Lv.Linlang@hotmail.com>
TPTBusiness