Shine1i
f47c424be3
feat: integrate gradient norm tracking in training runtime and metrics
...
- Enhanced chart logic to filter and visualize finite gradient norm values.
2026-02-17 18:26:59 +01:00
Shine1i
d9e5cc89fe
refactor: streamline vision model detection and improve state persistence logic
...
- Removed redundant vision-check controllers.
- Added `NON_PERSISTED_STATE_KEYS` to manage persisted training state.
- Introduced `partializePersistedState` for cleaner state filtering.
2026-02-17 18:01:45 +01:00
Shine1i
019d52a019
feat: add default model configuration mapping and auto-apply logic
...
- Implemented backend model configuration mapping to training state.
- Added auto-apply logic for default configurations when models are selected.
- Introduced utilities for type conversion and validation within training configuration.
2026-02-17 17:58:36 +01:00
imagineer99
cf364fca19
fix: subset param name
2026-02-17 05:53:21 +00:00
Wasim Yousef Said
16f79a73de
Merge pull request #124 from unslothai/feature/bug-fixes
...
feat: support disabling top-k sampling with -1 and standardize normalization
2026-02-16 13:21:12 -08:00
Roland Tannous
108ec254cb
Merge branch 'nightly' into feature/eval-split-auto-detection
2026-02-17 01:11:30 +04:00
Shine1i
4be6eefed3
feat: support disabling top-k sampling with -1 and standardize normalization logic
...
- Updated top-k parameter range to accept -1 in models and frontend.
- Added utility to normalize top-k for backend compatibility.
2026-02-16 21:33:24 +01:00
Roland Tannous
18879a521b
feat: auto-detect model+dataset compatibility to select VLM vs LLM training path
2026-02-16 19:18:49 +00:00
Roland Tannous
367bf1ff7f
feat(frontend): auto-detect vision models via backend, separate search filter from model classification
2026-02-16 18:24:44 +00:00
Roland Tannous
7401b22c16
fix: show Back to configuration breadcrumb when training completes
2026-02-16 17:34:52 +00:00
Roland Tannous
3b117189c5
feat: add eval_enabled flag and format-first-then-split for eval dataset
2026-02-16 14:13:55 +00:00
Roland Tannous
90c3561adb
feat: add eval split auto-detection, eval_steps hyperparam, and eval_loss chart integration
2026-02-16 13:38:54 +00:00
Roland Tannous
68df1c6306
feat(onboarding): replace hardcoded system info with live /api/system/hardware data
2026-02-16 12:26:49 +00:00
Roland Tannous
d49506b7b1
feat: add live GPU monitor with nvidia-smi polling during training
2026-02-16 11:47:43 +00:00
Shine1i
d05bfa3264
feat: add index route with auth guard and redirect logic
2026-02-16 09:57:41 +01:00
Shine1i
9e42629b63
feat: add inference parameter merging for model loading and runtime updates
2026-02-16 07:14:53 +01:00
imagineer99
570ca6758f
feat: add min_p inference parameter to chat page
2026-02-16 05:41:36 +00:00
Shine1i
9fe013a4f0
feat: lock model selector during guided tour and refine open state logic
2026-02-16 02:05:53 +01:00
Shine1i
cdfa4da226
merge: nightly
2026-02-16 02:02:40 +01:00
Shine1i
be745d1637
refactor: create utilities for dataset manual mapping and improve UI logic consistency in dataset preview dialog
2026-02-16 01:54:09 +01:00
Shine1i
1420b7c28a
fix: enforce unique input/output mapping and enhance UI feedback in dataset preview dialog
2026-02-16 01:47:28 +01:00
Shine1i
4cb881313d
feat: implement dataset mapping UI and preview dialog
2026-02-16 01:44:26 +01:00
Shine1i
a6bf95d13d
feat: enable conditional confetti in guided tour and update navbar icon
2026-02-16 01:04:54 +01:00
Shine1i
0e2cd7d92f
chore: text update
2026-02-16 00:58:46 +01:00
Shine1i
34f2e5b792
feat: improve guided tour descriptions and add sidebar state management
...
- Updated step descriptions across Studio, Chat, and Export tours for better clarity.
- Added `openSidebar` state management function and integrated it into the tour logic.
- Improved target detection in guided tours with retry logic for better handling of unavailable elements.
2026-02-16 00:55:13 +01:00
Shine1i
c7ff1687e6
feat: implement guided tours and refactor model selector components
2026-02-15 23:43:10 +01:00
Wasim Yousef Said
f2b3eb9a46
Merge pull request #101 from unslothai/feature/guided-tour
...
feat: studio: guided tour (1st visit, skippable)
2026-02-15 12:33:07 -08:00
Wasim Yousef Said
d33791f173
Merge pull request #100 from unslothai/feature/chat-compare
...
feat: chat compare + inference stream cancel fix
2026-02-15 12:29:54 -08:00
Wasim Yousef Said
17c98f5a4b
Merge pull request #98 from unslothai/feat/dataset-config-splits
...
feat: check dataset configs and splits before hitting check-format
2026-02-15 12:17:22 -08:00
Wasim Yousef Said
f99e4318e5
Merge pull request #99 from unslothai/feature/vram-estimation
...
feat: VRAM-based model filtering in frontend
2026-02-15 12:17:07 -08:00
Shine1i
b8c5ad2839
rm vitest
2026-02-15 21:13:15 +01:00
Shine1i
56461bd9a7
cfg->subset
2026-02-15 21:09:45 +01:00
Shine1i
579e515afd
fix hf cfg/split ui
2026-02-15 20:53:59 +01:00
Shine1i
def787e6d4
feat: add confetti fireworks effect on tour completion
2026-02-15 20:25:23 +01:00
Shine1i
78a9a5c8d0
refactor: reorganize tour steps and relocate ReadMore component for improved structure
2026-02-15 20:22:17 +01:00
Shine1i
825fe4b870
refactor: define explicit prop types for GuidedTour and SpotlightOverlay components
2026-02-15 20:17:28 +01:00
imagineer99
26ae7033ce
feat: VRAM-based model filtering in frontend
2026-02-15 19:12:28 +00:00
Shine1i
7eb2b91da7
refactor: extract tour utils into separate modules for cleaner structure
2026-02-15 20:11:28 +01:00
Shine1i
721e5e0fc5
tour readmore
2026-02-15 19:58:10 +01:00
Shine1i
a074f5e1c7
tour steps split
2026-02-15 19:57:07 +01:00
Shine1i
ffa9595fbd
rm shine + perf
2026-02-15 19:41:15 +01:00
Shine1i
a117d9cf94
tour light card
2026-02-15 19:31:49 +01:00
imagineer99
0ae1a73e46
feat: check dataset configs and splits before hitting check-format
2026-02-15 18:30:54 +00:00
Shine1i
3bba579768
feat: add guided tour component and integrate with Studio UI elements
2026-02-15 19:20:37 +01:00
Shine1i
4da43ae6b1
refactor: streamline chat runtime logic and remove warming indicator
...
- Replaced `setThreadWarming` logic with streamlined token settlement functions (`settleFirstTokenOk` and `settleFirstTokenErr`) for improved readability and reliability.
- Simplified model loading/unloading functions with reusable `performLoad` and `performUnload` patterns.
- Removed `warmingByThreadId` from runtime store and associated code for reduced complexity.
- Enhanced title generation flow by consolidating logic for persisting and streaming titles.
2026-02-15 18:59:03 +01:00
Shine1i
e02662c309
feat: improve model loading/unloading UX and remove _WarmupIndicator_ from thread UI
...
- Refactored loading/unloading logic to provide detailed toast notifications with statuses (loading, success, error).
- Removed unused `WarmupIndicator` component from thread UI to simplify interface.
- Introduced better error handling for model refresh and inference tasks.
2026-02-15 18:48:02 +01:00
Shine1i
5aeae854fd
feat: conditionally render "Compare" button based on active checkpoint and lora selection, improve fallback title generation, and update default settings
2026-02-15 18:38:25 +01:00
Shine1i
4f1c67bb02
chore: remove unused ComponentExample and associated imports and auto title generate
2026-02-15 18:08:46 +01:00
Shine1i
b707e4096f
fix: ensure consistent message order in chat runtime by improving sort logic and adding fallback for createdAt
2026-02-15 17:34:32 +01:00
Shine1i
bbee40172c
chat compare: send use_adapter
2026-02-15 16:44:14 +01:00
