release-0.5.0 by you-n-g · Pull Request #1 · microsoft/qlib

you-n-g · 2020-09-16T00:56:59Z

typo

you-n-g

Make this part optional and put it at last.

Make it possible that user could leave the model.score not implemented.
Maybe a try catch NotImplement can solve the problem

you-n-g · 2020-09-19T09:16:10Z

 ====================

-Users can get stock data by Qlib, the following are some examples.
+Users can get stock data by ``Qlib``. Following examples will demonstrate the basic user interface.


We will only provide csi300 and all stock pool in our demo. Please change this tutorial

you-n-g · 2020-09-19T13:38:05Z

            feature_label_config: /data/qlib/feature_config/feature_config.yaml

-The class `SomeDataHandler` should be in the module `custom_data_handler`, and Qlib could parse the `module_path` to load the class.
+The class `SomeDataHandler` should be in the module `custom_data_handler`, and ``Qlib`` could parse the `module_path` to load the class.


Did you test feature_label_config ?

deletefeature_label_config

you-n-g · 2020-09-19T13:44:33Z



-User can specify `trainer` through the configuration file:
+Users can specify `trainer` through the configuration file:


yaml will not accept such style of comments

you-n-g · 2020-09-19T13:47:17Z

+To Know more about ``Data Handler``, please refer to `Data Framework&Usage <data.html>`_.

-About training
+Trainer Field


Trainer section maybe better for such title

you-n-g · 2020-09-19T14:09:28Z

+---------------
+Users can use APIs in ``qlib.data`` to retrieve data, please refer to `Data Retrieval <../start/getdata.html>`_.
+
+Filter


Put the Filter behind feature.
Feature is Much more important

you-n-g · 2020-09-19T14:11:48Z

+Cache
+==========
+
+Local Cache


spli the memory cache and disk cache into two sub-sections in Cache section

How to split the Cache? what's the disk cache?

you-n-g · 2020-09-19T14:15:17Z

+backtest:
+  normal_backtest_args:
+    verbose: False
+    limit_threshold: 0.095


@zhupr
Please add limit_threshold=0.0095 as the default option in China Stock market setting

you-n-g · 2020-09-19T14:20:29Z

+
+To know more about backtesting with specific strategy, please refer to `Strategy <strategy.html>`_.
+
+Score File


Please rename all the content like Score XXX which indicates the predction of model to Prediction.
Such consistant name will help user understand the doc

you-n-g · 2020-09-19T14:22:13Z

+    sub_cost        mean    0.000718
+                    std     0.004694
+                    annual  0.181003
+                    ir      2.428964


@bxdd Only the docs are changed. The code is not changed

@zhupr It seems to have been fixed by Pengrong

you-n-g · 2020-09-19T14:24:13Z

+                    mdd    -0.072977
+
+- `sub_bench`
+    Income without deduction of fees


Income -> returns of the portfolio

you-n-g · 2020-09-19T14:29:53Z

+   Interday Model: Model Training & Prediction <component/model.rst>
+   Interday Strategy: Portfolio Management <component/strategy.rst>
+   Intraday Trading: Model&Strategy Testing <component/backtest.rst>
+   Aanalysis: Evaluation & Results Analysis <component/report.rst>


The content in report.rst is too simple.
There are no docs about the graph output by the functions
It's better to put all the graphs in the docs and exaplain then one by one
@bxdd @zhupr

…e-0.5.0

…o release-0.5.0

…e-0.5.0

Update README.md

…e-0.5.0

qlib auto init basedon project & black format

add import order

High freq execution

Fix test_start setting in `Benchmark.basic_task()`

CRITICAL FIX microsoft#1: Portfolio Chinese Stocks → ASX Stocks ❌ BEFORE: Chinese stocks (000001.SZ, 600036.SH, Kweichow Moutai) ✅ AFTER: ASX stocks (CBA.AX, BHP.AX, CSL.AX, WBC.AX, RIO.AX) ✅ Portfolio: 8 major ASX holdings worth A72K total ✅ Sectors: Financials 45.9%, Materials 25.8%, Healthcare 8.3% CRITICAL FIX microsoft#2: Email Authentication Features ✅ /api/auth/send-verification - 6-digit email codes ✅ /api/auth/verify-email - Email verification process ✅ /api/auth/forgot-password - Password reset links ✅ /api/auth/reset-password - Secure password updates ✅ Development mode with fallback codes ✅ Production-ready with SMTP integration SECURITY IMPROVEMENTS: - Secure token generation (secrets.token_urlsafe) - Rate limiting (max 3 verification attempts) - Expiration times (10min verification, 1hr reset) - No email enumeration (security best practice) Portfolio now shows Australian companies only. Authentication now production-ready for launch. 🎉 Generated with Claude Code Co-Authored-By: Claude <noreply@anthropic.com>

* init commit * change the version number * rich the docs&fix cache docs * update index readme * Modify cache class name * Modify sharpe to information_ratio * Modify Group- to Group * add the description of graphical results & fix the backtest docs * fix docs in details * update docs * Update introduction.rst * Update README.md * Update introduction.rst * Update introduction.rst * Update introduction.rst * Update installation.rst * Update installation.rst * Update initialization.rst * Update getdata.rst * Update integration.rst * Update initialization.rst * Update getdata.rst * Update estimator.rst Modify some typos. * Update README.md Modify the typos. * Update initialization.rst * Update data.rst * Update report.rst * Update estimator.rst * Update cumulative_return.py * Update model.rst * Update rank_label.py * Update cumulative_return.py * Update strategy.rst * Update getdata.rst * Update backtest.rst * Update integration.rst * Update getdata.rst * Update introduction.rst * Update introduction.rst * Update README.md * Update report.rst * Update integration.rst Fix typos * Update installation.rst Fix typos * Update getdata.rst * Update initialization.rst Fix typos. * add quick start docs&fix detials * fix estimator docs & fix strategy docs * fix the cahce in data.rst * update documents * Fix Corr && Rsquare * fix data retrival example to csi300 & fix a data bug * fix filter bug * Fix data collector * Modift model args * add the log & fix README.md\quick.rst * add enviroment depend & add intoduction of qlib-server online mode * fix image center fomat & set log_only of docs is True * fix README.md format * update data preparation & readme logo image * get_data support version * Modify analysis names * Modify analysis graph * update report.rst & data.rst * commmit estimator for merge * minimal requirements * Update README.md * Update README.md * Update README.md * Update README.md * Update README.md * Update README.md * Update README.md * Update READEME.md * Update READEME.md * update estimator * Fix doc urls * fix get_data.py docstring * update test_get_data.py * Upate docs * Upate docs * Upate docs Co-authored-by: bxdd <bxddream@gmail.com> Co-authored-by: zhupr <zhu.pengrong@foxmail.com> Co-authored-by: Wendi Li <wendili.academic@qq.com> Co-authored-by: Dingsu Wang <dingsu.wang@gmail.com> Co-authored-by: bxdd <45119470+bxdd@users.noreply.github.com> Co-authored-by: cslwqxx <cslwqxx@users.noreply.github.com>

qlib auto init basedon project & black format

在 CLAUDE.md 新增"已知问题与修复记录"章节，详细记录： Bug microsoft#1: iQuant 实盘下单失败 - passorder 返回 0 - 问题描述：passorder 返回 0，订单未进入系统，无错误提示 - 根本原因：缺少 is_last_bar() 检查，在历史回放阶段执行下单 - 修复方案：添加 is_last_bar() 检查，移除 set_account() 调用 - 经验教训：iQuant 策略生命周期、passorder 返回值含义、调试要点包含完整的问题诊断过程、对比分析、修复代码示例和验证结果。 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>

Reads the trained LightGBM's gain importance + TreeSHAP (lightgbm pred_contrib, no external shap lib) — the correct attribution that captures nonlinear/interaction use univariate IC misses. Our 6 custom factors = 19.5% of total SHAP vs 3.7% fair share (~5.3x). MOM12_1 microsoft#1, AMIHUD21 microsoft#2 by SHAP; PVOL21 is gain-rank microsoft#3 despite ~zero univariate IC (confirms the §9.7 lesson); only LOTTERY21 ranks bottom by both methods. Fork: gain/SHAP 是树模型正确的因子归因法，单因子 IC 会误判（见 §9.7/§9.8）

…ulk pulls The 429 path was an unbounded `while: continue` (hangs forever if persistently rate-limited) and 5xx/timeout hit raise_for_status() with no retry. Since upstream BaseCollector runs get_data under joblib.Parallel with no per-symbol try/except, one flaky response would hang or crash an entire S&P500 / minute-level pull. Add _fetch() with bounded retry (MAX_RETRIES=6, backoff) on 429/5xx/network errors; on final failure get_data skips the symbol instead of crashing the batch. Also cache the API key once instead of re-reading .env per pagination page. Fork: 自研 collector 鲁棒性修复（评估 §鲁棒性 microsoft#1），美股是主战场，批量拉取不能被单点请求拖死

…ruments + new tasks - Fix microsoft#1+microsoft#2: scheduler job now spawns rolling_train as a subprocess so the FastAPI event loop is never blocked by the 1.5-4h ML training. Adds AlreadyRunning guard + in-flight asyncio.Lock to prevent overlap. - Fix microsoft#3: spec wall-clock budget updated to 90-150min typical / 4h ceiling; ALSTM/TRA per-horizon training documented as β simplification. - Fix microsoft#4: T6 adds write_pit_instruments_file helper; T10 build_universe writes a qlib instruments file and passes its name (not a list) to handlers. - Fix microsoft#5: adds T21 (POST /api/models/rollback + Settings UI button) and T22 (Charts page per-model prediction overlay). - Renumbers acceptance script from T21 → T23. Plan now has 23 tasks / 148 steps / 23 commits. - Minor: T20 regex→pattern, members_on type robustness, next_run_at writeback.

- Run honest_baseline_daily.json: daily/k=30 → net_ir=-0.431, turnover=0.320, net_cagr=-14.1% (cost drag 36.9%/yr kills the gross +24.2% signal at ¥100k small account) - Sweep 120 configs; best at ¥100k: fixed/k=10/period=5 → net_ir=+0.902, net_cagr=+27.0% - improved_banded.json (fixed/10/5): net_ir=+0.902, turnover=0.161 (-49.7% vs baseline) - improved_banded_neutralized.json: neutralization hurt on this 12-week window (net_ir=-0.581) - P1 acceptance: microsoft#1 net_ir↑ PASS; microsoft#2 turnover ≤50% NARROWLY FAIL (50.3%) - Fix data.py: add sys.path purelib fixup + Path() wrap for load_config (same pattern as rolling_train.py) - All 29 backtest tests still green

you-n-g commented Sep 19, 2020

View reviewed changes

you-n-g assigned bxdd and zhupr Sep 19, 2020

you-n-g commented Sep 19, 2020

View reviewed changes

zhupr force-pushed the release-0.5.0 branch from 216f2ca to e01553f Compare September 20, 2020 17:18

you-n-g force-pushed the release-0.5.0 branch from 0c90528 to d3e22a6 Compare September 22, 2020 01:40

you-n-g closed this Sep 22, 2020

you-n-g force-pushed the release-0.5.0 branch from d3e22a6 to aa51e5a Compare September 22, 2020 01:42

you-n-g added 2 commits September 22, 2020 01:43

init commit

99ebd87

change the version number

6c41acf

you-n-g reopened this Sep 22, 2020

bxdd and others added 19 commits September 22, 2020 16:31

rich the docs&fix cache docs

c3b4a9a

update index readme

8e888c7

Merge branch 'release-0.5.0' of github.com:microsoft/qlib into releas…

b5570ed

…e-0.5.0

Modify cache class name

dfdcee9

Modify sharpe to information_ratio

0fb5d36

Modify Group- to Group

d92329b

merge origin

f43465d

add the description of graphical results & fix the backtest docs

446b6c3

Merge branch 'release-0.5.0' of https://github.com/microsoft/qlib int…

77e78c7

…o release-0.5.0

fix docs in details

d417c4f

update docs

45cf806

Merge branch 'release-0.5.0' of github.com:microsoft/qlib into releas…

d7b5833

…e-0.5.0

Update introduction.rst

0951895

Update README.md

b395ec5

Update introduction.rst

9812a0a

Merge pull request #10 from wendili-cs/patch-1

b3e36e4

Update README.md

Update introduction.rst

40c61c3

Update introduction.rst

690a36f

Update installation.rst

adec04a

zhupr added 6 commits September 24, 2020 11:12

fix get_data.py docstring

abeb14a

Merge branch 'release-0.5.0' of github.com:microsoft/qlib into releas…

1c69bd9

…e-0.5.0

update test_get_data.py

2f81797

Upate docs

fc9bbc4

Upate docs

654195b

Upate docs

9454da2

you-n-g merged commit de9e13b into main Sep 24, 2020

you-n-g deleted the release-0.5.0 branch September 25, 2020 14:46

tracoco mentioned this pull request Jan 7, 2021

Online client sample codes failed #184

Open

you-n-g pushed a commit that referenced this pull request May 17, 2021

Merge pull request #1 from you-n-g/online_srv

c4733f6

qlib auto init basedon project & black format

wangwenxi-handsome added a commit that referenced this pull request Aug 26, 2021

Merge pull request #1 from you-n-g/nested_decision_exe

5f22970

add import order

Mingzhe-Han added a commit that referenced this pull request Oct 3, 2021

Merge pull request #1 from microsoft/high-freq-execution

202bbea

High freq execution

orchardc mentioned this pull request Jan 3, 2022

How to define a list of shares (20 us shares for example) for predicting? Rather than the entire nasdaq100 or sp500? #794

Closed

2young-2simple-sometimes-naive mentioned this pull request Feb 23, 2022

double ensemble training bug after recent updates #928

Closed

2young-2simple-sometimes-naive mentioned this pull request Jul 7, 2022

highfreq data get calendar failure #1184

Closed

winglight mentioned this pull request Jul 8, 2022

Why dataset can't load alpha168/360 data since 2020-09-26? #1189

Closed

This was referenced Jul 13, 2022

AttributeError: module 'collector' has no attribute 'YahooNormalizeUS1dExtend' while running update_data_to_bin #1196

Closed

Crash normalizing data #1206

Closed

2young-2simple-sometimes-naive mentioned this pull request Jul 25, 2022

dump_fix mode of dump_bin.py #1231

Closed

ZhongHaoAustin pushed a commit to ZhongHaoAustin/qlib that referenced this pull request Apr 25, 2024

Merge pull request microsoft#1 from zhiyuan5986/patch-1

4fced29

Fix test_start setting in `Benchmark.basic_task()`

ChristoCai mentioned this pull request Feb 5, 2025

AttributeError: 'BaostockCollectorHS3005min' object has no attribute 'DEFAULT_START_DATETIME_5MIN' #1885

Open

gaogen123 pushed a commit to gaogen123/qlib that referenced this pull request Sep 13, 2025

Merge pull request microsoft#1 from you-n-g/online_srv

1cbdad1

qlib auto init basedon project & black format

Olcmyk mentioned this pull request May 23, 2026

DDG-DA: TypeError: unhashable type: 'list' in InternalData.setup() #2229

Closed



		User can specify `trainer` through the configuration file:
		Users can specify `trainer` through the configuration file:


		To know more about backtesting with specific strategy, please refer to `Strategy <strategy.html>`_.

		Score File

Conversation

you-n-g commented Sep 16, 2020

Uh oh!

you-n-g left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants