{% extends "base.html" %} {% block title %}{{ project.name }} — {{ app_name }}{% endblock %} {% block content %}

{{ project.name }}

{{ project.description or project.project_type.value | replace('_', ' ') | title }}

{% if project.location %}

Location: {{ project.location }}

{% endif %}
{{ chrome.label_cta }}
{% for step in chrome.workflow %} {% if not loop.first %} {% endif %} {% if step.href == 'upload' %} {% elif step.href == 'annotate' %} {% else %} {% endif %} {% endfor %}
{% for tab in chrome.tabs %} {% endfor %}
{% for card in chrome.overview_cards %} {% if card.href == 'upload' %} {% elif card.href == 'annotate' %} {% else %}
{% endif %}
{{ card.step }}

{{ card.title }}

{{ card.body }}

{% if card.href %} {% else %}
{% endif %} {% endfor %} {% if modality == 'image' %} {% endif %}

{{ chrome.assets_label }}

{% if modality == 'image' %} {% endif %} Annotate

{% if chrome.quality_profile == 'image' %}Dataset quality & cleaning{% else %}Quality & cleaning{% endif %}

{% if chrome.quality_profile == 'image' %}Configure checks, then preview or apply cleaning before generating a version or training.{% else %}Check label readiness, then clean empty / duplicate samples before packaging or training.{% endif %}

Checks to apply

Thresholds & gate

Config is saved in this browser and reused when starting training with the readiness gate on.

Training readiness

Click “Run checks” to score this project.

Passed

  • —

Blocked

  • —

Warnings

  • —

Data cleaning

Remediations run before versioning / packaging / training. Preview is non-destructive; Apply may delete rows. 

—

Profile

—

Active learning

Suggests images or samples to annotate next. Embeddings-based diversity is not implemented yet.

{{ chrome.classes_hint or 'Define label classes for this task.' }}

Generate Dataset Version

Configure splits, preprocessing, and augmentations for detection / segmentation training.

Thumbnail source
Loading preview…

Train / Valid / Test Split

Train 70%
Valid 20%
Test 10%

Total: 100%

Preprocessing

Applied once to every image. Click a card to toggle.

Augmentation

Enable augmentation
Creates extra labeled variants for training

Export Dataset

{% if project.project_type.value == 'rag_eval' %}Package for evaluation{% elif modality == 'llm' %}Package / Fine-tune{% else %}Train / Fine-tune{% endif %}

{% if project.project_type.value == 'rag_eval' %} Package query / contexts / answer JSONL for RAGAS, DeepEval, or a custom RAG eval harness. There is no model fine-tune in this flow. {% elif modality == 'image' %} Train YOLO, Face (MTCNN / RetinaFace-style), or Torchvision detectors locally. {% elif modality == 'tabular' %} Train sklearn classifiers / regressors on labeled rows. {% elif modality == 'text' %} Train classical NLP or prepare Hugging Face packages. {% elif modality == 'audio' %} Train audio-label models or prepare Whisper / ASR fine-tune JSONL. {% elif modality == 'video' %} Train video-label models or prepare action-recognition packages. {% elif modality == 'llm' %} Package instruction or preference data for LLM fine-tuning (TRL, Axolotl…). {% else %} Choose a training backend for this workspace. {% endif %}

Team & review

Members, image assignments, and the review queue.

Members

  • Loading…

Assign images

  • —

Review queue

  • —

Experiments

Training runs are recorded as experiments. Select two or more to compare metrics side-by-side.

Pipelines

Build workflows as a visual DAG or YAML. The runner executes validation, cleaning, splits, export, training, and active learning; other step types are skipped for now.

Import YAML

Latest run

No run yet.
{% endblock %} {% block scripts %} {% endblock %}