Dobromir Popov
d136f9d79c
data anotation ui/ testcase generation
2025-10-18 16:14:58 +03:00
Dobromir Popov
0c28a0997c
more cleanup
2025-10-13 16:11:06 +03:00
Dobromir Popov
6cf4d902df
update requirements
2025-10-09 15:22:49 +03:00
Dobromir Popov
a86e07f556
tools
2025-10-09 13:58:43 +03:00
Dobromir Popov
270ba2e52b
fix broken merge
2025-10-08 20:02:41 +03:00
Dobromir Popov
a468c75c47
fix merge
2025-10-02 23:50:08 +03:00
Dobromir Popov
8654e08028
renames
2025-10-01 00:34:13 +03:00
Dobromir Popov
388334e4a8
Merge commit 'd49a473ed6f4aef55bfdd47d6370e53582be6b7b' into cleanup
2025-10-01 00:32:19 +03:00
Dobromir Popov
a03b9c5701
improve predictions
2025-10-01 00:30:37 +03:00
Dobromir Popov
0a28cee58d
docs: Add prediction loop integration guide
...
- Document the root cause: models never called for real-time predictions
- Provide step-by-step integration instructions
- Show expected behavior before/after
- Guide for monitoring and validation
2025-10-01 00:17:17 +03:00
Dobromir Popov
510a657092
feat: Add real-time prediction loop - the MISSING piece
...
Create RealtimePredictionLoop to continuously call model.predict() on incoming data
Features:
- Detects new 1s/1m candles and triggers predictions
- Detects pivot points for prediction triggers
- Calls ALL models (CNN, DQN, COB-RL) continuously
- Combines predictions into trading signals
- NO SYNTHETIC DATA - only real market data
This is why model.predict() was only called once at startup - there was NO
continuous prediction loop running on new market data!
2025-10-01 00:16:33 +03:00
Dobromir Popov
0d08339d98
docs: Add comprehensive training fix implementation plan
...
- Document critical issues and fixes applied
- Detail proper training loop architecture
- Outline signal-position linking system
- Define comprehensive reward calculation
- List implementation phases and next steps
2025-10-01 00:08:46 +03:00
Dobromir Popov
49529d564d
fix: Critical error fixes - division by zero and mock data removal
...
- Add zero price validation in trading_executor to prevent division by zero
- Remove mock prediction fallback - enforce NO SYNTHETIC DATA policy
- System now fails gracefully without predictions rather than using fake data
Fixes division by zero errors when executing trades
Enforces REAL_MARKET_DATA_POLICY.md compliance
2025-10-01 00:07:19 +03:00
Dobromir Popov
608da8233f
main cleanup
2025-09-30 23:56:36 +03:00
Dobromir Popov
468a2c2a66
revision, pending fixes
2025-09-26 10:49:45 +03:00
Dobromir Popov
2b09e7fb5a
mtp
2025-09-26 02:42:10 +03:00
Dobromir Popov
00ae5bd579
NPU (wip); docker
2025-09-25 00:46:08 +03:00
Dobromir Popov
d9a66026c6
docker container, inference chaining
2025-09-25 00:32:59 +03:00
Dobromir Popov
1f35258a66
show dummy references
2025-09-09 22:27:07 +03:00
Dobromir Popov
2e1b3be2cd
increase prediction horizon
2025-09-09 09:50:14 +03:00
Dobromir Popov
34780d62c7
better logging
2025-09-09 09:41:30 +03:00
Dobromir Popov
47d63fddfb
dash fix wip
2025-09-09 03:59:06 +03:00
Dobromir Popov
2f51966fa8
update dash with model performance
2025-09-09 03:51:04 +03:00
Dobromir Popov
55fb865e7f
training metrics . fix cnn model
2025-09-09 03:43:20 +03:00
Dobromir Popov
a3029d09c2
full RL training pass
2025-09-09 03:41:06 +03:00
Dobromir Popov
17e18ae86c
more elaborate RL training
2025-09-09 03:33:49 +03:00
Dobromir Popov
8c17082643
immedite training imp
2025-09-09 02:57:03 +03:00
Dobromir Popov
729e0bccb1
cob ma data for models
2025-09-09 02:07:04 +03:00
Dobromir Popov
317c703ea0
unify model names
2025-09-09 01:10:35 +03:00
Dobromir Popov
0e886527c8
models load
2025-09-09 00:51:33 +03:00
Dobromir Popov
9671d0d363
dedulicae model storage
2025-09-09 00:45:49 +03:00
Dobromir Popov
c3a94600c8
refactoring
2025-09-08 23:57:21 +03:00
Dobromir Popov
98ebbe5089
cleanup
2025-09-08 15:22:01 +03:00
Dobromir Popov
96b0513834
ignore mcp
2025-09-08 14:58:04 +03:00
Dobromir Popov
32d54f0604
model selector
2025-09-08 14:53:46 +03:00
Dobromir Popov
e61536e43d
additional logging for data stream
2025-09-08 14:08:13 +03:00
Dobromir Popov
56e857435c
cleanup
2025-09-08 13:41:22 +03:00
Dobromir Popov
c9fba56622
model checkpoint manager
2025-09-08 13:31:11 +03:00
Dobromir Popov
060fdd28b4
enable training
2025-09-08 12:13:50 +03:00
Dobromir Popov
4fe952dbee
wip
2025-09-08 11:44:15 +03:00
Dobromir Popov
fe6763c4ba
prediction database
2025-09-02 19:25:42 +03:00
Dobromir Popov
226a6aa047
training wip
2025-09-02 19:25:13 +03:00
Dobromir Popov
6dcb82c184
data normalizations
2025-09-02 18:51:49 +03:00
Dobromir Popov
1c013f2806
improve stream
2025-09-02 18:15:12 +03:00
Dobromir Popov
c55175c44d
data stream working
2025-09-02 17:59:12 +03:00
Dobromir Popov
8068e554f3
data stream
2025-09-02 17:29:18 +03:00
Dobromir Popov
e0fb76d9c7
removed COB 400M Model, text data stream wip
2025-09-02 16:16:01 +03:00
Dobromir Popov
15cc694669
fix models loading /saving issue
2025-09-02 16:05:44 +03:00
Dobromir Popov
1b54438082
dash and training wip
2025-09-02 15:30:05 +03:00
Dobromir Popov
443e8e746f
req notes
2025-08-29 18:50:53 +03:00
