readme screenshots

This commit is contained in:
Extraltodeus
2026-07-13 23:52:57 +02:00
parent 1ff4063282
commit 790e1fa687
13 changed files with 35 additions and 8 deletions
+35 -8
View File
@@ -23,7 +23,9 @@ ship**.
│ files exist. │
└───────────────────────────────────────────────────────────────────────┘ -->
![J-Wash — chat with the live Jacobian lens](assets/animation1.gif)
![J-Wash — chat with the live Jacobian lens](assets/header.png)
<p align="center"><sub>Introducing non-expert friendly alignment!</sub></p>
---
@@ -120,6 +122,8 @@ listed automatically; **Browse** adds any model folder on disk to the list
(nothing is copied — the blue button forgets the entry, the red trash deletes
actual files). fp32 models are auto-converted to bf16 to halve disk usage.
![models tab](assets/models_tab.png)
### 2. Load a Jacobian lens
In **Lens**, J-Wash lists compatible lenses for the loaded model — local ones you
@@ -131,7 +135,8 @@ you can even pick a lens **while the model is still loading**, it chain-loads
when ready. No lens? Fit one in the **Fit** tab (see below). Manual loading by
repo / file / local path is available at the bottom of the tab.
![The Lens tab and the live Frequencies view](assets/lens-tab.png)
![models tab](assets/lens.png)
### 3. Chat with the live lens
@@ -141,9 +146,23 @@ generated token:
- **Frequencies** (default): tokens the layers "read," aggregated by how often they
appear — size ∝ frequency. Click a token to **pin** it (rank curves + a rank
heatmap per layer); right-click to hide noise.
- **Heatmap**: layers × positions, top token per cell (reading = amber, thinking =
![The Lens tab and the live Frequencies view](assets/freq_animation.gif)
Once a token is pinned, you can see it's activation (vertical axis) along the tokens generated (horizontal axis). The tokens related to the column hovered can be seen on the upper part :
![Graphs view](assets/animation1.gif)
- **Heatmap view**: layers × positions, top token per cell (reading = amber, thinking =
blue).
![Heatmap view and a pinned token's per-layer rank curves](assets/heatmap_animation.gif)
Selecting a token will display related activations in all layers:
![Activations](assets/activations.png)
Leading/trailing spaces are rendered with `˽` (so `˽Euro``Euro`). Replies
render as markdown (toggleable), can be **edited in place** (✎ — later turns use
the edited text) and **continued** (the model picks up exactly where it
@@ -151,13 +170,17 @@ stopped). Conversations are persisted (SQLite + full-text search), branchable
from any node, and replayable offline. Export a conversation as JSON or
Markdown, with or without lens frames. The lens view's height is draggable.
![Heatmap view and a pinned token's per-layer rank curves](assets/heatmap-pins.png)
### 4. Edit tokens ☢
Open the **token editor** (the ☢ button in the composer, or the ☢ on a pinned
Open the **token editor** (the **☢** button in the composer, or the ☢ on a pinned
token). Add rules:
![Edit](assets/edit.png)
![Edit2](assets/edit2.png)
- **multiply ×f** — `×0` removes a token's direction, `×0.5` attenuates, `×2`
amplifies;
- **replace** — rewrite token A's component onto token B's direction
@@ -169,7 +192,9 @@ editing. A mode toggle switches between:
- **Per-layer steering** (default) — the most expressive way to *explore*, but it
does not export faithfully.
- **Read projection** (pure-weights) — a change of basis of the downstream reads so
the **live preview matches the exported checkpoint exactly**. Use this to ship.
the **live preview matches the exported checkpoint exactly**. Use this to save a model and preview the result.
![projection](assets/projection.png)
> **Architecture note**: models whose layers normalize their *writes* into the
> residual stream (Gemma 2/3 style, `pre/post_feedforward_layernorm`) can't take
@@ -177,8 +202,6 @@ editing. A mode toggle switches between:
> abliteration) instead — still pure weights, faithful for full removals and
> replacements (a rule's layer range is ignored: the projection is global).
![The token editor: rules, mode toggle, and export](assets/token-editor.png)
### 5. Export the edit
Save a set of rules as a **preset** and re-apply it in one click. Export an edit
@@ -189,6 +212,10 @@ Save a set of rules as a **preset** and re-apply it in one click. Export an edit
- **LoRA** (PEFT) — the exact low-rank diff between the edited weights and the
originals (the edit is low-rank by construction, so nothing is approximated).
![export](assets/export.png)
Exports are standard safetensors weights — everything that follows from that
(quantizing, converting to other runtimes' formats, publishing on the Hub)
works exactly as it would for any other model.
Binary file not shown.

After

Width:  |  Height:  |  Size: 112 KiB

BIN
View File
Binary file not shown.

After

Width:  |  Height:  |  Size: 1.1 KiB

BIN
View File
Binary file not shown.

After

Width:  |  Height:  |  Size: 8.9 KiB

BIN
View File
Binary file not shown.

After

Width:  |  Height:  |  Size: 9.3 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 1.4 MiB

BIN
View File
Binary file not shown.

After

Width:  |  Height:  |  Size: 2.1 MiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 2.5 MiB

BIN
View File
Binary file not shown.

After

Width:  |  Height:  |  Size: 12 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 9.2 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 22 KiB

BIN
View File
Binary file not shown.

After

Width:  |  Height:  |  Size: 148 KiB

BIN
View File
Binary file not shown.

After

Width:  |  Height:  |  Size: 14 KiB