Spaces:

InstaDeepAI
/

ntv3

Running

App Files Files Community

bernardo-de-almeida commited on Dec 12, 2025

Commit

42f0385

1 Parent(s): 680e6ef

feat: add new tab

Browse files

Files changed (9) hide show

README.md +1 -1
app.py +11 -0
app_tracks.py +158 -0
index.html +148 -192
notebooks_pipelines/02_genome_annotation.ipynb +14 -127
notebooks_pipelines/NTv3_650M_pos_hg38_chr19_6700000_6831072.gff3 +107 -0
requirements.txt +7 -0
tabs/demo.html +88 -0
tabs/home.html +199 -0

README.md CHANGED Viewed

@@ -3,7 +3,7 @@ title: NTv3 — Foundation Models for Long-Range Genomics
 emoji: 🧬
 colorFrom: indigo
 colorTo: blue
-sdk: static
 pinned: false
 ---

 emoji: 🧬
 colorFrom: indigo
 colorTo: blue
+sdk: gradio
 pinned: false
 ---

app.py ADDED Viewed

	@@ -0,0 +1,11 @@

+"""
+Main Gradio app entry point for NTv3 Space.
+This file imports the track prediction demo from app_tracks.py.
+"""
+from app_tracks import demo_interface
+# For Hugging Face Spaces with Gradio SDK, the 'demo' variable must be named 'demo'
+demo = demo_interface
+if __name__ == "__main__":
+    demo.launch(server_name="0.0.0.0", share=False)

app_tracks.py ADDED Viewed

	@@ -0,0 +1,158 @@

+"""
+Gradio app for NTv3 track prediction demo.
+This module contains the interactive track prediction interface.
+"""
+import gradio as gr
+import torch
+from transformers import pipeline
+import os
+# Initialize the pipeline (will be loaded on first use)
+ntv3_tracks = None
+def load_pipeline():
+    """Load the pipeline on first use (lazy loading)."""
+    global ntv3_tracks
+    if ntv3_tracks is None:
+        model_name = "InstaDeepAI/NTv3_650M_pos"
+        ntv3_tracks = pipeline(
+            "ntv3-tracks",
+            model=model_name,
+            trust_remote_code=True,
+            device=0 if torch.cuda.is_available() else -1,
+        )
+    return ntv3_tracks
+def predict_tracks(chrom, start, end, species):
+    """Run track prediction on the specified genomic region."""
+    try:
+        # Validate inputs
+        if not chrom or not start or not end or not species:
+            return "❌ Please fill in all fields."
+        start = int(start)
+        end = int(end)
+        if start >= end:
+            return "❌ Start position must be less than end position."
+        if end - start > 1_000_000:
+            return "❌ Region size cannot exceed 1 Mb (1,000,000 bp)."
+        # Load pipeline
+        pipe = load_pipeline()
+        # Run prediction
+        out = pipe({
+            "chrom": chrom,
+            "start": start,
+            "end": end,
+            "species": species.lower()
+        })
+        # Format output
+        result = f"""✅ Prediction completed successfully!
+📊 Output Shapes:
+• BigWig tracks logits: {tuple(out.bigwig_tracks_logits.shape)}
+  → {out.bigwig_tracks_logits.shape[1]} functional tracks over the center region
+• BED tracks logits: {tuple(out.bed_tracks_logits.shape)}
+  → {out.bed_tracks_logits.shape[1]} genomic elements over the center region
+• Language model logits: {tuple(out.mlm_logits.shape)}
+  → MLM predictions for the entire sequence
+📝 Note: Predictions are made over 37.5% of the center region of the input sequence.
+"""
+        return result
+    except Exception as e:
+        return f"❌ Error: {str(e)}"
+# Create the track prediction demo interface (embedded in HTML)
+def create_demo_interface():
+    """Create the Gradio interface for track prediction."""
+    with gr.Blocks(title="NTv3 Track Prediction Demo", theme=gr.themes.Soft()) as demo_interface:
+        gr.Markdown("""
+        # 🧬 NTv3 Interactive Track Prediction Demo
+        This demo allows you to run the NTv3 650M post-trained model to predict functional tracks and genomic elements for any genomic region.
+        **Model:** `InstaDeepAI/NTv3_650M_pos`
+        """)
+        with gr.Row():
+            with gr.Column():
+                chrom = gr.Textbox(
+                    label="Chromosome",
+                    placeholder="e.g., chr19",
+                    value="chr19",
+                    info="Chromosome name (e.g., chr1, chr19)"
+                )
+                start = gr.Number(
+                    label="Start Position",
+                    placeholder="e.g., 6700000",
+                    value=6_700_000,
+                    info="Start position in base pairs"
+                )
+                end = gr.Number(
+                    label="End Position",
+                    placeholder="e.g., 6831072",
+                    value=6_831_072,
+                    info="End position in base pairs"
+                )
+                species = gr.Dropdown(
+                    label="Species",
+                    choices=[
+                        "human", "mouse", "rat", "chicken", "zebrafish",
+                        "fruitfly", "worm", "yeast", "arabidopsis", "rice",
+                        "maize", "soybean", "tomato", "potato", "grape",
+                        "poplar", "medicago", "lotus", "brachypodium", "sorghum",
+                        "barley", "wheat", "oats", "rye"
+                    ],
+                    value="human",
+                    info="Select the species (24 supported species)"
+                )
+                predict_btn = gr.Button("🚀 Run Prediction", variant="primary")
+            with gr.Column():
+                output = gr.Textbox(
+                    label="Results",
+                    lines=15,
+                    interactive=False,
+                    placeholder="Results will appear here after running prediction..."
+                )
+        gr.Markdown("""
+        ### 📝 Notes:
+        - The model predicts ~7k functional tracks and 21 genomic elements
+        - Predictions are made over 37.5% of the center region of the input sequence
+        - Maximum region size: 1 Mb (1,000,000 base pairs)
+        - First run may take longer as the model loads
+        """)
+        predict_btn.click(
+            fn=predict_tracks,
+            inputs=[chrom, start, end, species],
+            outputs=output
+        )
+        gr.Examples(
+            examples=[
+                ["chr19", 6_700_000, 6_831_072, "human"],
+                ["chr1", 100_000, 200_000, "human"],
+                ["chr2", 50_000, 150_000, "mouse"],
+            ],
+            inputs=[chrom, start, end, species]
+        )
+    return demo_interface
+# Create the demo interface
+demo_interface = create_demo_interface()
+# If running this file directly (for local testing)
+if __name__ == "__main__":
+    demo_interface.launch(server_name="0.0.0.0", share=False)

index.html CHANGED Viewed

@@ -199,9 +199,56 @@
       border-radius: 12px;
     }
     .footer { margin-top: 22px; color: var(--muted); font-size: 13px; }
     @media (max-width: 860px) {
       .card { grid-column: span 12; }
       h1 { font-size: 28px; }
     }
   </style>
 </head>
@@ -223,215 +270,124 @@
       </div>
     </div>
-    <div class="summary">
-      <h2>📖 About NTv3</h2>
-      <p>
-        NTv3 is a multi-species genomic foundation model family that unifies representation learning, functional-track prediction, genome annotation, and controllable sequence generation within a single U-Net-style backbone. It models up to 1 Mb of DNA at single-base resolution, using a conv–Transformer–deconv architecture that efficiently captures both local motifs and long-range regulatory dependencies. NTv3 is first pretrained on ~9T base pairs from the OpenGenome2 corpus spanning >128k species using masked language modeling, and then post-trained with a joint objective on ~16k functional tracks and annotation labels across 24 animal and plant species, enabling state-of-the-art cross-species functional prediction and base-resolution genome annotation.
-      </p>
-      <p>
-        Beyond prediction, NTv3 can be fine-tuned into a controllable generative model via masked-diffusion language modeling, allowing targeted design of regulatory sequences (for example, enhancers with specified activity and promoter selectivity) that have been validated experimentally.
-      </p>
     </div>
-    <div class="paper-summary">
-		<!-- <h2>📄 A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction</h2> -->
-		<img src="assets/paper_summary.png" alt="NTv3 Paper Summary" />
     </div>
-    <div class="why-ntv3">
-      <h2>✨ Why NTv3?</h2>
-      <ul>
-        <li>📏 <strong>1 Mb long context at nucleotide resolution</strong> — ~100× longer than typical genomics models.</li>
-        <li>🏗️ <strong>Unified architecture</strong> for: masked language modeling, functional-track prediction, genome annotation, and sequence generation.</li>
-        <li>🌍 <strong>Cross-species generalization</strong> across 24 animals + plants with a shared conditioned representation space.</li>
-        <li>⚡ <strong>U-Net–style architecture</strong> improves stability and GPU efficiency on very long sequences.</li>
-        <li>🎯 <strong>Controllable generative modeling</strong>, enabling targeted enhancer/promoter engineering validated by experimental assays.</li>
-      </ul>
-    </div>
-    <div class="grid">
-      <div class="card">
-        <h2>🤖 Models (see <a href="https://huggingface.co/collections/InstaDeepAI/nucleotide-transformer-v3" target="_blank" rel="noopener">collection</a>)</h2>
-        <ul>
-          <li>📦 Pretrained checkpoints:
-            <div style="margin-top: 8px; margin-left: 0;">
-              <div><a href="https://huggingface.co/InstaDeepAI/NTv3_8M_pre"><code>InstaDeepAI/NTv3_8M_pre</code></a></div>
-              <div><a href="https://huggingface.co/InstaDeepAI/NTv3_100M_pre"><code>InstaDeepAI/NTv3_100M_pre</code></a></div>
-              <div><a href="https://huggingface.co/InstaDeepAI/NTv3_650M_pre"><code>InstaDeepAI/NTv3_650M_pre</code></a></div>
-            </div>
-          </li>
-          <li>🎯 Post-trained checkpoints:
-            <div style="margin-top: 8px; margin-left: 0;">
-              <div><a href="https://huggingface.co/InstaDeepAI/NTv3_100M_pos"><code>InstaDeepAI/NTv3_100M_pos</code></a></div>
-              <div><a href="https://huggingface.co/InstaDeepAI/NTv3_650M_pos"><code>InstaDeepAI/NTv3_650M_pos</code></a></div>
-            </div>
-          </li>
-        </ul>
-        <table>
-          <thead>
-            <tr>
-              <th>Model</th>
-              <th>Size</th>
-              <th>Pre-training</th>
-              <th>Post-training</th>
-              <th>Tasks</th>
-            </tr>
-          </thead>
-          <tbody>
-            <tr>
-              <td><strong>NTv3-8M</strong></td>
-              <td>8M params</td>
-              <td>MLM</td>
-              <td>❌</td>
-              <td>Embeddings, light inference</td>
-            </tr>
-            <tr>
-              <td><strong>NTv3-100M</strong></td>
-              <td>100M params</td>
-              <td>MLM</td>
-              <td><span class="checkmark">✅</span></td>
-              <td>Tracks, annotation</td>
-            </tr>
-            <tr>
-              <td><strong>NTv3-650M</strong></td>
-              <td>650M params</td>
-              <td>MLM</td>
-              <td><span class="checkmark">✅</span></td>
-              <td>Tracks, annotation, best accuracy</td>
-            </tr>
-          </tbody>
-        </table>
-      </div>
-      <div class="card-stack">
-        <div class="card">
-          <h2>📓 Tutorial notebooks (browse <a href="https://huggingface.co/spaces/InstaDeepAI/ntv3/tree/main/notebooks_tutorials" target="_blank" rel="noopener">folder</a>)</h2>
-          <ul>
-            <li><a href="https://huggingface.co/spaces/InstaDeepAI/ntv3/blob/main/notebooks_tutorials/00_quickstart_inference.ipynb" target="_blank" rel="noopener">🚀 00 — Quickstart inference</a></li>
-            <li><a href="https://huggingface.co/spaces/InstaDeepAI/ntv3/blob/main/notebooks_tutorials/01_tracks_prediction.ipynb" target="_blank" rel="noopener">📊 01 — Tracks prediction</a></li>
-            <li>🎯 02 — Fine-tune on bigwig tracks</li>
-            <li>🔍 03 — Model interpretation</li>
-            <li>🧪 04 — Training NTv3 generative </li>
-          </ul>
         </div>
-        <div class="card">
-          <h2>📓 Pipelines notebooks (browse <a href="https://huggingface.co/spaces/InstaDeepAI/ntv3/tree/main/notebooks_pipelines" target="_blank" rel="noopener">folder</a>)</h2>
-          <ul>
-            <li> 🎯 01 — Generate bigwig predictions for certain tracks</li>
-            <li><a href="https://huggingface.co/spaces/InstaDeepAI/ntv3/blob/main/notebooks_pipelines/02_genome_annotation.ipynb" target="_blank" rel="noopener">🏷️ 02 — Genome annotation / segmentation</a></li>
-            <li>🎯 03 — Fine-tune on bigwig tracks</li>
-            <li>🔍 04 — Interpret a given genomic region</li>
-            <li>🧪 05 — Sequence generation</li>
-          </ul>
-        </div>
-        <div class="card">
-          <h2>🔗 Links</h2>
-          <ul>
-            <li>📄 Paper: (add link)</li>
-            <li><a href="https://github.com/instadeepai/nucleotide-transformer">💻 JAX model code (GitHub)</a></li>
-            <li><a href="https://huggingface.co/collections/InstaDeepAI/nucleotide-transformer-v3" target="_blank" rel="noopener">🎯 HF Model Collection (all NTv3 models)</a></li>
-            <li><a href="https://huggingface.co/spaces/InstaDeepAI/ntv3/tree/main/notebooks" target="_blank" rel="noopener">📓 All notebooks</a></li>
-            <li><a href="https://huggingface.co/spaces/InstaDeepAI/ntv3_benchmark" target="_blank" rel="noopener">🏆 NTv3 benchmark leaderboard</a></li>
-          </ul>
-        </div>
-      </div>
-      <div class="card">
-        <h2>🤖 Load a pre-trained model</h2>
-        <p>Here is an example of how to load and use a pre-trained NTv3 model.</p>
-        <div class="code"><pre><code class="language-python">from transformers import AutoTokenizer, AutoModelForMaskedLM
-model_name = "InstaDeepAI/NTv3_650M_pre"
-# Load model and tokenizer
-model = AutoModelForMaskedLM.from_pretrained(model_name, trust_remote_code=True)
-tok = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
-# Tokenize input sequences
-batch = tok(["ATCGNATCG", "ACGT"], add_special_tokens=False, padding=True, pad_to_multiple_of=128, return_tensors="pt")
-# Run model
-out = model(
-  **batch,
-  output_hidden_states=True,
-  output_attentions=True
-)
-# Print output shapes
-print(out.logits.shape)       # (B, L, V = 11)
-print(len(out.hidden_states)) # convs + transformers + deconvs
-print(len(out.attentions))    # equals transformer layers = 12
-</code></pre></div>
-      <p>Model embeddings can be used for fine-tuning on downstream tasks.</p>
-      <p style="margin-top: 40px;">TO DO: add pipeline for fine-tuning on functional tracks or genome annotation.</p>
-      </div>
-      <div class="card">
-        <h2>💻 Use a post-trained model</h2>
-        <p>Here is a quick example of how to use the post-trained NTv3 650M model to predict tracks for a human genomic window.</p>
-        <div class="code"><pre><code class="language-python">from transformers import pipeline
-import torch
-model_name = "InstaDeepAI/NTv3_650M_pos"
-ntv3_tracks = pipeline(
-    "ntv3-tracks",
-    model=model_name,
-    trust_remote_code=True,
-    device=0 if torch.cuda.is_available() else -1,
-)
-# Run track prediction
-out = ntv3_tracks(
-  {
-    "chrom": "chr19",
-    "start": 6_700_000,
-    "end": 6_831_072,
-    "species": "human"
-  }
-)
-# Print output shapes
-# 7k human tracks over 37.5 % center region of the input sequence
-print("bigwig_tracks_logits:", tuple(out.bigwig_tracks_logits.shape))
-# Location of 21 genomic elements over 37.5 % center region of the input sequence
-print("bed_tracks_logits:", tuple(out.bed_tracks_logits.shape))
-# Language model logits for whole sequence over vocabulary
-print("language model logits:", tuple(out.mlm_logits.shape))</code></pre></div>
-      <p>Predictions can also be plotted for a subset of functional tracks and genomic elements:</p>
-      <div class="code"><pre><code class="language-python">tracks_to_plot = {
-    "K562 RNA-seq": "ENCSR056HPM",
-    "K562 DNAse": "ENCSR921NMD",
-    "K562 H3k4me3": "ENCSR000DWD",
-    "K562 CTCF": "ENCSR000AKO",
-    "HepG2 RNA-seq": "ENCSR561FEE_P",
-    "HepG2 DNAse": "ENCSR000EJV",
-    "HepG2 H3k4me3": "ENCSR000AMP",
-    "HepG2 CTCF": "ENCSR000BIE",
-}
-elements_to_plot = ["protein_coding_gene", "exon", "intron", "splice_donor", "splice_acceptor"]
-out = ntv3_tracks(
-    {"chrom": "chr19", "start": 6_700_000, "end": 6_831_072, "species": "human"},
-    plot=True,
-    tracks_to_plot=tracks_to_plot,
-    elements_to_plot=elements_to_plot,
-)</code></pre></div>
-      <img src="assets/output_tracks.png" alt="Output tracks visualization" style="max-width: 100%; margin-top: 20px;" />
-          </div>
-        </div>
-    <!-- <div class="paper-summary">
-		<h2>📄 A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction</h2>
-		<img src="assets/paper_summary.png" alt="NTv3 Paper Summary" />
-    </div> -->
-    <p class="footer">
-      © instadeep-ai — NTv3 companion Space.
-    </p>
-  </div>
-  <script src="https://cdnjs.cloudflare.com/ajax/libs/prism/1.29.0/components/prism-core.min.js"></script>
-  <script src="https://cdnjs.cloudflare.com/ajax/libs/prism/1.29.0/plugins/autoloader/prism-autoloader.min.js"></script>
 </body>
 </html>

       border-radius: 12px;
     }
     .footer { margin-top: 22px; color: var(--muted); font-size: 13px; }
+    /* Tab navigation styles */
+    .tabs {
+      margin-top: 24px;
+      display: flex;
+      gap: 8px;
+      border-bottom: 2px solid var(--border);
+      overflow-x: auto;
+    }
+    .tab-button {
+      padding: 12px 20px;
+      background: transparent;
+      border: none;
+      border-bottom: 2px solid transparent;
+      color: var(--muted);
+      font-family: var(--sans);
+      font-size: 14px;
+      font-weight: 500;
+      cursor: pointer;
+      transition: all 0.2s ease;
+      white-space: nowrap;
+      margin-bottom: -2px;
+    }
+    .tab-button:hover {
+      color: var(--text);
+      background: rgba(255, 255, 255, 0.03);
+    }
+    .tab-button.active {
+      color: var(--link);
+      border-bottom-color: var(--link);
+    }
+    .tab-content {
+      display: none;
+      animation: fadeIn 0.3s ease;
+    }
+    .tab-content.active {
+      display: block;
+    }
+    @keyframes fadeIn {
+      from { opacity: 0; transform: translateY(8px); }
+      to { opacity: 1; transform: translateY(0); }
+    }
     @media (max-width: 860px) {
       .card { grid-column: span 12; }
       h1 { font-size: 28px; }
+      .tab-button {
+        padding: 10px 16px;
+        font-size: 13px;
+      }
     }
   </style>
 </head>
       </div>
     </div>
+    <!-- Tab Navigation -->
+    <div class="tabs">
+      <button class="tab-button active" data-tab="home">🏠 Home</button>
+      <button class="tab-button" data-tab="demo">💻 Code Demo</button>
     </div>
+    <!-- Home Tab (Content loaded from tabs/home.html) -->
+    <div id="home" class="tab-content active">
+      <!-- Content will be loaded dynamically -->
     </div>
+    <!-- Code Demo Tab (Content loaded from tabs/demo.html) -->
+    <div id="demo" class="tab-content">
+      <!-- Content will be loaded dynamically -->
         </div>
+    <!-- <div class="paper-summary">
+		<h2>📄 A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction</h2>
+		<img src="assets/paper_summary.png" alt="NTv3 Paper Summary" />
+    </div> -->
+    <p class="footer">
+      © instadeep-ai — NTv3 companion Space.
+    </p>
+  </div>
+  <script src="https://cdnjs.cloudflare.com/ajax/libs/prism/1.29.0/components/prism-core.min.js"></script>
+  <script src="https://cdnjs.cloudflare.com/ajax/libs/prism/1.29.0/plugins/autoloader/prism-autoloader.min.js"></script>
+  <script>
+    // Tab content mapping
+    const tabFiles = {
+      'home': 'tabs/home.html',
+      'demo': 'tabs/demo.html'
+    };
+    // Cache for loaded tab content
+    const tabCache = {};
+    // Function to load tab content
+    async function loadTabContent(tabId) {
+      // Return cached content if available
+      if (tabCache[tabId]) {
+        return tabCache[tabId];
+      }
+      // Load content from file
+      const filePath = tabFiles[tabId];
+      if (!filePath) {
+        console.error(`No file path defined for tab: ${tabId}`);
+        return '';
+      }
+      try {
+        const response = await fetch(filePath);
+        if (!response.ok) {
+          throw new Error(`Failed to load ${filePath}: ${response.statusText}`);
+        }
+        const content = await response.text();
+        tabCache[tabId] = content;
+        return content;
+      } catch (error) {
+        console.error(`Error loading tab content for ${tabId}:`, error);
+        return `<div class="summary"><p>Error loading content. Please refresh the page.</p></div>`;
+      }
+    }
+    // Function to show a tab
+    async function showTab(tabId) {
+      const tabContent = document.getElementById(tabId);
+      if (!tabContent) {
+        console.error(`Tab element not found: ${tabId}`);
+        return;
+      }
+      // Load content if not already loaded
+      if (!tabContent.dataset.loaded) {
+        tabContent.innerHTML = await loadTabContent(tabId);
+        tabContent.dataset.loaded = 'true';
+        // Re-run Prism.js syntax highlighting for code blocks in the loaded content
+        if (typeof Prism !== 'undefined') {
+          // Find all code blocks in the loaded content and highlight them
+          const codeBlocks = tabContent.querySelectorAll('code[class*="language-"]');
+          codeBlocks.forEach(block => {
+            Prism.highlightElement(block);
+          });
+        }
+      }
+    }
+    // Tab switching functionality
+    document.addEventListener('DOMContentLoaded', function() {
+      const tabButtons = document.querySelectorAll('.tab-button');
+      const tabContents = document.querySelectorAll('.tab-content');
+      // Load the default active tab (home)
+      const activeTab = document.querySelector('.tab-content.active');
+      if (activeTab) {
+        showTab(activeTab.id);
+      }
+      tabButtons.forEach(button => {
+        button.addEventListener('click', async () => {
+          const targetTab = button.getAttribute('data-tab');
+          // Remove active class from all buttons and contents
+          tabButtons.forEach(btn => btn.classList.remove('active'));
+          tabContents.forEach(content => content.classList.remove('active'));
+          // Add active class to clicked button and corresponding content
+          button.classList.add('active');
+          const tabElement = document.getElementById(targetTab);
+          tabElement.classList.add('active');
+          // Load and show the tab content
+          await showTab(targetTab);
+        });
+      });
+    });
+  </script>
 </body>
 </html>

notebooks_pipelines/02_genome_annotation.ipynb CHANGED Viewed

@@ -29,16 +29,7 @@
    "execution_count": 1,
    "id": "2e2f5963",
    "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "\u001b[33mWARNING: 401 Error, Credentials not correct for https://gitlab.com/api/v4/projects/36813343/packages/pypi/simple/igv-notebook/\u001b[0m\u001b[33m\n",
-      "\u001b[0m"
-     ]
-    }
-   ],
    "source": [
     "# Install dependencies\n",
     "!pip -q install \"transformers>=4.55\" \"huggingface_hub>=0.23\" safetensors torch pyfaidx requests seaborn matplotlib igv_notebook"
@@ -127,28 +118,14 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
    "id": "4857d15c",
    "metadata": {},
    "outputs": [
     {
      "data": {
       "application/vnd.jupyter.widget-view+json": {
-       "model_id": "92629742bf7e419b9aaad0c8c14867d5",
-       "version_major": 2,
-       "version_minor": 0
-      },
-      "text/plain": [
-       "config.json:   0%|          | 0.00/338k [00:00<?, ?B/s]"
-      ]
-     },
-     "metadata": {},
-     "output_type": "display_data"
-    },
-    {
-     "data": {
-      "application/vnd.jupyter.widget-view+json": {
-       "model_id": "2468d781d0b7409791c5079ee9860a81",
        "version_major": 2,
        "version_minor": 0
       },
@@ -165,105 +142,15 @@
      "text": [
       "A new version of the following files was downloaded from https://huggingface.co/InstaDeepAI/NTv3_650M_pos:\n",
       "- ntv3_gff_pipeline.py\n",
-      ". Make sure to double-check they do not contain any added malicious code. To avoid downloading new versions of the code file, you can pin a revision.\n"
-     ]
-    },
-    {
-     "data": {
-      "application/vnd.jupyter.widget-view+json": {
-       "model_id": "fabadaa764ba4da799c0d43b12ac42b1",
-       "version_major": 2,
-       "version_minor": 0
-      },
-      "text/plain": [
-       "model.safetensors:   0%|          | 0.00/2.72G [00:00<?, ?B/s]"
-      ]
-     },
-     "metadata": {},
-     "output_type": "display_data"
-    },
-    {
-     "data": {
-      "application/vnd.jupyter.widget-view+json": {
-       "model_id": "9188715aa52d48a2b54b6b89a015f1da",
-       "version_major": 2,
-       "version_minor": 0
-      },
-      "text/plain": [
-       "tokenizer_config.json:   0%|          | 0.00/1.47k [00:00<?, ?B/s]"
-      ]
-     },
-     "metadata": {},
-     "output_type": "display_data"
-    },
-    {
-     "data": {
-      "application/vnd.jupyter.widget-view+json": {
-       "model_id": "f457beef2cdf4ecca076b589a95edf2b",
-       "version_major": 2,
-       "version_minor": 0
-      },
-      "text/plain": [
-       "vocab.json:   0%|          | 0.00/138 [00:00<?, ?B/s]"
-      ]
-     },
-     "metadata": {},
-     "output_type": "display_data"
-    },
-    {
-     "data": {
-      "application/vnd.jupyter.widget-view+json": {
-       "model_id": "b900fc305af84983b820e385d239dc29",
-       "version_major": 2,
-       "version_minor": 0
-      },
-      "text/plain": [
-       "special_tokens_map.json:   0%|          | 0.00/149 [00:00<?, ?B/s]"
-      ]
-     },
-     "metadata": {},
-     "output_type": "display_data"
-    },
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
       "Device set to use cpu\n"
      ]
     },
-    {
-     "data": {
-      "application/vnd.jupyter.widget-view+json": {
-       "model_id": "7fbd5f15218142e1b4a14474e96189b8",
-       "version_major": 2,
-       "version_minor": 0
-      },
-      "text/plain": [
-       "tokenizer_config.json:   0%|          | 0.00/1.49k [00:00<?, ?B/s]"
-      ]
-     },
-     "metadata": {},
-     "output_type": "display_data"
-    },
-    {
-     "data": {
-      "application/vnd.jupyter.widget-view+json": {
-       "model_id": "c3ce29bcfd6b4f0681b0ac94809ef9ab",
-       "version_major": 2,
-       "version_minor": 0
-      },
-      "text/plain": [
-       "vocab.json:   0%|          | 0.00/693 [00:00<?, ?B/s]"
-      ]
-     },
-     "metadata": {},
-     "output_type": "display_data"
-    },
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Inference + decoding time: 47.49 seconds\n"
      ]
     }
    ],
@@ -302,7 +189,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 7,
    "id": "959cf79f",
    "metadata": {},
    "outputs": [
@@ -336,7 +223,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 8,
    "id": "84f013f6",
    "metadata": {},
    "outputs": [
@@ -379,14 +266,14 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 9,
    "id": "0904a5cb",
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/html": [
-       "<div id=\"jb_2472686_buttons\"></div><div id=\"jb_2472686_igvcontainer\"></div>"
       ],
       "text/plain": [
        "<IPython.core.display.HTML object>"
@@ -397,7 +284,7 @@
     },
     {
      "data": {
-      "application/javascript": "window.igv.MessageHandler.on({\"id\": \"jb_2472686\", \"command\": \"createBrowser\", \"data\": {\"genome\": \"hg38\", \"locus\": \"chr19:6700000-6831072\", \"id\": \"jb_2472686\"}})",
       "text/plain": [
        "<IPython.core.display.Javascript object>"
       ]
@@ -407,7 +294,7 @@
     },
     {
      "data": {
-      "application/javascript": "window.igv.MessageHandler.on({\"id\": \"jb_2472686\", \"command\": \"loadTrack\", \"data\": {\"name\": \"NTv3 annotations\", \"format\": \"gff3\", \"type\": \"annotation\", \"url\": \"NTv3_650M_pos_hg38_chr19_6700000_6831072.gff3\"}})",
       "text/plain": [
        "<IPython.core.display.Javascript object>"
       ]
@@ -417,7 +304,7 @@
     },
     {
      "data": {
-      "application/javascript": "window.igv.MessageHandler.on({\"id\": \"jb_2472686\", \"command\": \"search\", \"data\": \"chr19:6700000-6831072\"})",
       "text/plain": [
        "<IPython.core.display.Javascript object>"
       ]
@@ -428,10 +315,10 @@
     {
      "data": {
       "text/plain": [
-       "<igv_notebook.browser.Browser at 0x1047ec880>"
       ]
      },
-     "execution_count": 9,
      "metadata": {},
      "output_type": "execute_result"
     }

    "execution_count": 1,
    "id": "2e2f5963",
    "metadata": {},
+   "outputs": [],
    "source": [
     "# Install dependencies\n",
     "!pip -q install \"transformers>=4.55\" \"huggingface_hub>=0.23\" safetensors torch pyfaidx requests seaborn matplotlib igv_notebook"
   },
   {
    "cell_type": "code",
+   "execution_count": 5,
    "id": "4857d15c",
    "metadata": {},
    "outputs": [
     {
      "data": {
       "application/vnd.jupyter.widget-view+json": {
+       "model_id": "cead875ae8c34250b6929e22283652e1",
        "version_major": 2,
        "version_minor": 0
       },
      "text": [
       "A new version of the following files was downloaded from https://huggingface.co/InstaDeepAI/NTv3_650M_pos:\n",
       "- ntv3_gff_pipeline.py\n",
+      ". Make sure to double-check they do not contain any added malicious code. To avoid downloading new versions of the code file, you can pin a revision.\n",
       "Device set to use cpu\n"
      ]
     },
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
+      "Inference + decoding time: 53.09 seconds\n"
      ]
     }
    ],
   },
   {
    "cell_type": "code",
+   "execution_count": 6,
    "id": "959cf79f",
    "metadata": {},
    "outputs": [
   },
   {
    "cell_type": "code",
+   "execution_count": 7,
    "id": "84f013f6",
    "metadata": {},
    "outputs": [
   },
   {
    "cell_type": "code",
+   "execution_count": 8,
    "id": "0904a5cb",
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/html": [
+       "<div id=\"jb_471625_buttons\"></div><div id=\"jb_471625_igvcontainer\"></div>"
       ],
       "text/plain": [
        "<IPython.core.display.HTML object>"
     },
     {
      "data": {
+      "application/javascript": "window.igv.MessageHandler.on({\"id\": \"jb_471625\", \"command\": \"createBrowser\", \"data\": {\"genome\": \"hg38\", \"locus\": \"chr19:6700000-6831072\", \"id\": \"jb_471625\"}})",
       "text/plain": [
        "<IPython.core.display.Javascript object>"
       ]
     },
     {
      "data": {
+      "application/javascript": "window.igv.MessageHandler.on({\"id\": \"jb_471625\", \"command\": \"loadTrack\", \"data\": {\"name\": \"NTv3 annotations\", \"format\": \"gff3\", \"type\": \"annotation\", \"url\": \"NTv3_650M_pos_hg38_chr19_6700000_6831072.gff3\"}})",
       "text/plain": [
        "<IPython.core.display.Javascript object>"
       ]
     },
     {
      "data": {
+      "application/javascript": "window.igv.MessageHandler.on({\"id\": \"jb_471625\", \"command\": \"search\", \"data\": \"chr19:6700000-6831072\"})",
       "text/plain": [
        "<IPython.core.display.Javascript object>"
       ]
     {
      "data": {
       "text/plain": [
+       "<igv_notebook.browser.Browser at 0x30d4e3e50>"
       ]
      },
+     "execution_count": 8,
      "metadata": {},
      "output_type": "execute_result"
     }

notebooks_pipelines/NTv3_650M_pos_hg38_chr19_6700000_6831072.gff3 ADDED Viewed

	@@ -0,0 +1,107 @@

+##gff-version 3
+# model: InstaDeepAI/NTv3_650M_pos
+# window: chr19:6700000-6831072 (hg38); predictions on central 37.5%: chr19:6740960-6790112
+chr19	NTv3_HMM	intron	6740961	6740995	0.975	.	.	ID=INTRON_1;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	start_codon	6740996	6741013	0.355	.	.	ID=START_CODON_2;Name=START_CODON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,191,255
+chr19	NTv3_HMM	exon	6741014	6741124	0.673	.	.	ID=EXON_3;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_donor_site	6741125	6741125	0.857	.	.	ID=SPLICE_DONOR_4;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6741126	6741224	0.974	.	.	ID=INTRON_5;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_acceptor_site	6741225	6741225	0.930	.	.	ID=SPLICE_ACCEPTOR_6;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6741226	6741280	0.693	.	.	ID=EXON_7;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_donor_site	6741281	6741281	0.837	.	.	ID=SPLICE_DONOR_8;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6741282	6742966	0.959	.	.	ID=INTRON_9;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_acceptor_site	6742967	6742967	0.958	.	.	ID=SPLICE_ACCEPTOR_10;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6742968	6743113	0.841	.	.	ID=EXON_11;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_donor_site	6743114	6743114	0.779	.	.	ID=SPLICE_DONOR_12;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6743115	6743193	0.963	.	.	ID=INTRON_13;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_acceptor_site	6743194	6743194	0.910	.	.	ID=SPLICE_ACCEPTOR_14;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6743195	6743255	0.845	.	.	ID=EXON_15;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_donor_site	6743256	6743256	0.782	.	.	ID=SPLICE_DONOR_16;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6743257	6743493	0.970	.	.	ID=INTRON_17;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_acceptor_site	6743494	6743494	0.780	.	.	ID=SPLICE_ACCEPTOR_18;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6743495	6743597	0.876	.	.	ID=EXON_19;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_donor_site	6743598	6743598	0.856	.	.	ID=SPLICE_DONOR_20;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6743599	6743707	0.951	.	.	ID=INTRON_21;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_acceptor_site	6743708	6743708	0.856	.	.	ID=SPLICE_ACCEPTOR_22;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6743709	6743835	0.812	.	.	ID=EXON_23;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_donor_site	6743836	6743836	0.887	.	.	ID=SPLICE_DONOR_24;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6743837	6744553	0.989	.	.	ID=INTRON_25;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_acceptor_site	6744554	6744554	0.972	.	.	ID=SPLICE_ACCEPTOR_26;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6744555	6744700	0.977	.	.	ID=EXON_27;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	intron	6744701	6744799	0.972	.	.	ID=INTRON_28;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_acceptor_site	6744800	6744800	0.954	.	.	ID=SPLICE_ACCEPTOR_29;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6744801	6744993	0.977	.	.	ID=EXON_30;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_donor_site	6744994	6744994	0.886	.	.	ID=SPLICE_DONOR_31;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6744995	6746451	0.979	.	.	ID=INTRON_32;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_acceptor_site	6746452	6746452	0.938	.	.	ID=SPLICE_ACCEPTOR_33;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6746453	6746560	0.840	.	.	ID=EXON_34;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_donor_site	6746561	6746561	0.947	.	.	ID=SPLICE_DONOR_35;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6746562	6749933	0.973	.	.	ID=INTRON_36;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_acceptor_site	6749934	6749934	0.693	.	.	ID=SPLICE_ACCEPTOR_37;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6749935	6750065	0.918	.	.	ID=EXON_38;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_donor_site	6750066	6750066	0.783	.	.	ID=SPLICE_DONOR_39;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6750067	6750291	0.955	.	.	ID=INTRON_40;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_acceptor_site	6750292	6750292	0.960	.	.	ID=SPLICE_ACCEPTOR_41;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6750293	6750430	0.959	.	.	ID=EXON_42;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_donor_site	6750431	6750431	0.723	.	.	ID=SPLICE_DONOR_43;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6750432	6750511	0.939	.	.	ID=INTRON_44;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_acceptor_site	6750512	6750512	0.750	.	.	ID=SPLICE_ACCEPTOR_45;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6750513	6750632	0.902	.	.	ID=EXON_46;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_donor_site	6750633	6750633	0.917	.	.	ID=SPLICE_DONOR_47;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6750634	6751062	0.961	.	.	ID=INTRON_48;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_acceptor_site	6751063	6751063	0.694	.	.	ID=SPLICE_ACCEPTOR_49;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6751064	6751199	0.558	.	.	ID=EXON_50;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	stop_codon	6751200	6751212	0.332	.	.	ID=STOP_CODON_51;Name=STOP_CODON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=220,20,60
+chr19	NTv3_HMM	three_prime_UTR	6751213	6751488	0.965	+	.	ID=UTR3_PLUS_52;Name=UTR3_PLUS;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=34,139,34
+chr19	NTv3_HMM	polyA_signal	6751489	6751507	0.355	.	.	ID=POLYA_SIGNAL_53;Name=POLYA_SIGNAL;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=139,69,19
+chr19	NTv3_HMM	start_codon	6751508	6752169	0.002	.	.	ID=START_CODON_54;Name=START_CODON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,191,255
+chr19	NTv3_HMM	polyA_signal	6752170	6752187	0.432	.	.	ID=POLYA_SIGNAL_55;Name=POLYA_SIGNAL;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=139,69,19
+chr19	NTv3_HMM	three_prime_UTR	6752188	6752571	0.839	-	.	ID=UTR3_MINUS_56;Name=UTR3_MINUS;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=34,139,34
+chr19	NTv3_HMM	stop_codon	6752572	6752752	0.136	.	.	ID=STOP_CODON_57;Name=STOP_CODON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=220,20,60
+chr19	NTv3_HMM	splice_acceptor_site	6752753	6752753	0.798	.	.	ID=SPLICE_ACCEPTOR_58;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6752754	6753455	0.910	.	.	ID=INTRON_59;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_donor_site	6753456	6753456	0.766	.	.	ID=SPLICE_DONOR_60;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6753457	6753640	0.953	.	.	ID=EXON_61;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_acceptor_site	6753641	6753641	0.939	.	.	ID=SPLICE_ACCEPTOR_62;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6753642	6754051	0.985	.	.	ID=INTRON_63;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_donor_site	6754052	6754052	0.844	.	.	ID=SPLICE_DONOR_64;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6754053	6754161	0.908	.	.	ID=EXON_65;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_acceptor_site	6754162	6754163	0.633	.	.	ID=SPLICE_ACCEPTOR_66;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6754164	6754250	0.962	.	.	ID=INTRON_67;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_donor_site	6754251	6754251	0.875	.	.	ID=SPLICE_DONOR_68;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6754252	6754424	0.965	.	.	ID=EXON_69;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_acceptor_site	6754425	6754425	0.791	.	.	ID=SPLICE_ACCEPTOR_70;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6754426	6754615	0.975	.	.	ID=INTRON_71;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_donor_site	6754616	6754616	0.953	.	.	ID=SPLICE_DONOR_72;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6754617	6754730	0.731	.	.	ID=EXON_73;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_acceptor_site	6754731	6754731	0.822	.	.	ID=SPLICE_ACCEPTOR_74;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6754732	6754830	0.975	.	.	ID=INTRON_75;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_donor_site	6754831	6754831	0.944	.	.	ID=SPLICE_DONOR_76;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6754832	6755314	0.757	.	.	ID=EXON_77;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_acceptor_site	6755315	6755315	0.713	.	.	ID=SPLICE_ACCEPTOR_78;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6755316	6759593	0.988	.	.	ID=INTRON_79;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_donor_site	6759594	6759594	0.928	.	.	ID=SPLICE_DONOR_80;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6759595	6759669	0.840	.	.	ID=EXON_81;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_acceptor_site	6759670	6759670	0.901	.	.	ID=SPLICE_ACCEPTOR_82;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6759671	6760637	0.985	.	.	ID=INTRON_83;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_donor_site	6760638	6760638	0.928	.	.	ID=SPLICE_DONOR_84;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6760639	6760985	0.748	.	.	ID=EXON_85;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_acceptor_site	6760986	6760987	0.603	.	.	ID=SPLICE_ACCEPTOR_86;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6760988	6763679	0.984	.	.	ID=INTRON_87;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_donor_site	6763680	6763680	0.759	.	.	ID=SPLICE_DONOR_88;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	exon	6763681	6763732	0.663	.	.	ID=EXON_89;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	five_prime_UTR	6763733	6763815	0.840	-	.	ID=UTR5_MINUS_90;Name=UTR5_MINUS;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,128,0
+chr19	NTv3_HMM	splice_acceptor_site	6763816	6763816	0.869	.	.	ID=SPLICE_ACCEPTOR_91;Name=SPLICE_ACCEPTOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6763817	6767386	0.976	.	.	ID=INTRON_92;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072
+chr19	NTv3_HMM	splice_donor_site	6767387	6767387	0.902	.	.	ID=SPLICE_DONOR_93;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	start_codon	6767388	6767411	0.051	.	.	ID=START_CODON_94;Name=START_CODON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,191,255
+chr19	NTv3_HMM	five_prime_UTR	6767412	6767514	0.578	-	.	ID=UTR5_MINUS_95;Name=UTR5_MINUS;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,128,0
+chr19	NTv3_HMM	start_codon	6767515	6769347	0.009	.	.	ID=START_CODON_96;Name=START_CODON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,191,255
+chr19	NTv3_HMM	TF_binding_site	6769348	6769521	0.506	.	.	ID=CTCF_97;Name=CTCF;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=148,0,211
+chr19	NTv3_HMM	start_codon	6769522	6772696	0.002	.	.	ID=START_CODON_98;Name=START_CODON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,191,255
+chr19	NTv3_HMM	five_prime_UTR	6772697	6772806	0.885	+	.	ID=UTR5_PLUS_99;Name=UTR5_PLUS;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,128,0
+chr19	NTv3_HMM	start_codon	6772807	6772810	0.694	.	.	ID=START_CODON_100;Name=START_CODON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,191,255
+chr19	NTv3_HMM	five_prime_UTR	6772811	6772922	0.748	+	.	ID=UTR5_PLUS_101;Name=UTR5_PLUS;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,128,0
+chr19	NTv3_HMM	exon	6772923	6773010	0.635	.	.	ID=EXON_102;Name=EXON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,255
+chr19	NTv3_HMM	splice_donor_site	6773011	6773011	0.884	.	.	ID=SPLICE_DONOR_103;Name=SPLICE_DONOR;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072;color=0,0,0
+chr19	NTv3_HMM	intron	6773012	6790112	0.972	.	.	ID=INTRON_104;Name=INTRON;model=InstaDeepAI/NTv3_650M_pos;assembly=hg38;window=chr19:6700000-6831072

requirements.txt ADDED Viewed

	@@ -0,0 +1,7 @@

+gradio>=4.0.0
+torch>=2.0.0
+transformers>=4.55.0
+accelerate>=0.20.0
+safetensors>=0.3.0
+huggingface_hub>=0.23.0

tabs/demo.html ADDED Viewed

	@@ -0,0 +1,88 @@

+<div class="summary">
+  <h2>💻 Interactive Code Demo</h2>
+  <p>Run the NTv3 650M post-trained model interactively to predict functional tracks and genomic elements for any genomic region.</p>
+  <p><strong>Model:</strong> <code>InstaDeepAI/NTv3_650M_pos</code></p>
+</div>
+<div class="grid">
+  <div class="card" style="grid-column: span 12;">
+    <h2>🚀 NTv3 Track Prediction Pipeline</h2>
+    <p>Enter a genomic region to get predictions for functional tracks and genomic elements. The model will predict ~7k functional tracks and 21 genomic elements over the center 37.5% of your input region.</p>
+    <!-- Gradio app embedded here -->
+    <!-- Note: With Gradio SDK, the app.py serves as the main interface -->
+    <!-- The HTML interface can still be accessed, but the Gradio demo is the primary interface -->
+    <div id="gradio-container" style="margin-top: 20px; min-height: 600px;">
+      <p style="color: var(--muted); margin-bottom: 15px;">
+        <strong>Note:</strong> With Gradio SDK enabled, the interactive demo is now the main interface of this Space.
+        You can interact with it directly, or use the code example below to run predictions programmatically.
+      </p>
+      <div style="background: rgba(0,0,0,0.3); padding: 20px; border-radius: 12px; border: 1px solid var(--border);">
+        <p style="color: var(--link); margin: 0;">
+          💡 The Gradio interactive demo is now available as the main interface of this Space.
+          Refresh the page to see it, or use the code example below.
+        </p>
+      </div>
+    </div>
+    <p style="margin-top: 20px; color: var(--muted); font-size: 13px;">
+      <strong>Note:</strong> The first run may take longer as the model loads. Maximum region size: 1 Mb (1,000,000 base pairs).
+    </p>
+  </div>
+  <div class="card" style="grid-column: span 12;">
+    <h2>📝 Code Example</h2>
+    <p>Here's the Python code that powers the demo above. You can run this in a notebook or Python script:</p>
+    <div class="code"><pre><code class="language-python">from transformers import pipeline
+import torch
+model_name = "InstaDeepAI/NTv3_650M_pos"
+ntv3_tracks = pipeline(
+    "ntv3-tracks",
+    model=model_name,
+    trust_remote_code=True,
+    device=0 if torch.cuda.is_available() else -1,
+)
+# Run track prediction
+out = ntv3_tracks(
+  {
+    "chrom": "chr19",
+    "start": 6_700_000,
+    "end": 6_831_072,
+    "species": "human"
+  }
+)
+# Print output shapes
+# 7k human tracks over 37.5 % center region of the input sequence
+print("bigwig_tracks_logits:", tuple(out.bigwig_tracks_logits.shape))
+# Location of 21 genomic elements over 37.5 % center region of the input sequence
+print("bed_tracks_logits:", tuple(out.bed_tracks_logits.shape))
+# Language model logits for whole sequence over vocabulary
+print("language model logits:", tuple(out.mlm_logits.shape))</code></pre></div>
+    <p style="margin-top: 15px;">To run the interactive Gradio app locally:</p>
+    <div class="code"><pre><code class="language-bash">pip install -r requirements.txt
+python app.py</code></pre></div>
+  </div>
+</div>
+<script>
+// Try to detect if Gradio app is available
+window.addEventListener('load', function() {
+  const iframe = document.getElementById('gradio-iframe');
+  iframe.onerror = function() {
+    // If iframe fails to load, keep showing the instructions
+    document.getElementById('gradio-loading').style.display = 'block';
+    iframe.style.display = 'none';
+  };
+  // Set a timeout to show instructions if iframe doesn't load
+  setTimeout(function() {
+    if (iframe.style.display === 'none') {
+      document.getElementById('gradio-loading').style.display = 'block';
+    }
+  }, 2000);
+});
+</script>

tabs/home.html ADDED Viewed

	@@ -0,0 +1,199 @@

+<div class="summary">
+  <h2>📖 About NTv3</h2>
+  <p>
+    NTv3 is a multi-species genomic foundation model family that unifies representation learning, functional-track prediction, genome annotation, and controllable sequence generation within a single U-Net-style backbone. It models up to 1 Mb of DNA at single-base resolution, using a conv–Transformer–deconv architecture that efficiently captures both local motifs and long-range regulatory dependencies. NTv3 is first pretrained on ~9T base pairs from the OpenGenome2 corpus spanning >128k species using masked language modeling, and then post-trained with a joint objective on ~16k functional tracks and annotation labels across 24 animal and plant species, enabling state-of-the-art cross-species functional prediction and base-resolution genome annotation.
+  </p>
+  <p>
+    Beyond prediction, NTv3 can be fine-tuned into a controllable generative model via masked-diffusion language modeling, allowing targeted design of regulatory sequences (for example, enhancers with specified activity and promoter selectivity) that have been validated experimentally.
+  </p>
+</div>
+<div class="paper-summary">
+  <!-- <h2>📄 A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction</h2> -->
+  <img src="assets/paper_summary.png" alt="NTv3 Paper Summary" />
+</div>
+<div class="why-ntv3">
+  <h2>✨ Why NTv3?</h2>
+  <ul>
+    <li>📏 <strong>1 Mb long context at nucleotide resolution</strong> — ~100× longer than typical genomics models.</li>
+    <li>🏗️ <strong>Unified architecture</strong> for: masked language modeling, functional-track prediction, genome annotation, and sequence generation.</li>
+    <li>🌍 <strong>Cross-species generalization</strong> across 24 animals + plants with a shared conditioned representation space.</li>
+    <li>⚡ <strong>U-Net–style architecture</strong> improves stability and GPU efficiency on very long sequences.</li>
+    <li>🎯 <strong>Controllable generative modeling</strong>, enabling targeted enhancer/promoter engineering validated by experimental assays.</li>
+  </ul>
+</div>
+<div class="grid">
+  <div class="card">
+    <h2>🤖 Models (see <a href="https://huggingface.co/collections/InstaDeepAI/nucleotide-transformer-v3" target="_blank" rel="noopener">collection</a>)</h2>
+    <ul>
+      <li>📦 Pretrained checkpoints:
+        <div style="margin-top: 8px; margin-left: 0;">
+          <div><a href="https://huggingface.co/InstaDeepAI/NTv3_8M_pre"><code>InstaDeepAI/NTv3_8M_pre</code></a></div>
+          <div><a href="https://huggingface.co/InstaDeepAI/NTv3_100M_pre"><code>InstaDeepAI/NTv3_100M_pre</code></a></div>
+          <div><a href="https://huggingface.co/InstaDeepAI/NTv3_650M_pre"><code>InstaDeepAI/NTv3_650M_pre</code></a></div>
+        </div>
+      </li>
+      <li>🎯 Post-trained checkpoints:
+        <div style="margin-top: 8px; margin-left: 0;">
+          <div><a href="https://huggingface.co/InstaDeepAI/NTv3_100M_pos"><code>InstaDeepAI/NTv3_100M_pos</code></a></div>
+          <div><a href="https://huggingface.co/InstaDeepAI/NTv3_650M_pos"><code>InstaDeepAI/NTv3_650M_pos</code></a></div>
+        </div>
+      </li>
+    </ul>
+    <table>
+      <thead>
+        <tr>
+          <th>Model</th>
+          <th>Size</th>
+          <th>Pre-training</th>
+          <th>Post-training</th>
+          <th>Tasks</th>
+        </tr>
+      </thead>
+      <tbody>
+        <tr>
+          <td><strong>NTv3-8M</strong></td>
+          <td>8M params</td>
+          <td>MLM</td>
+          <td>❌</td>
+          <td>Embeddings, light inference</td>
+        </tr>
+        <tr>
+          <td><strong>NTv3-100M</strong></td>
+          <td>100M params</td>
+          <td>MLM</td>
+          <td><span class="checkmark">✅</span></td>
+          <td>Tracks, annotation</td>
+        </tr>
+        <tr>
+          <td><strong>NTv3-650M</strong></td>
+          <td>650M params</td>
+          <td>MLM</td>
+          <td><span class="checkmark">✅</span></td>
+          <td>Tracks, annotation, best accuracy</td>
+        </tr>
+      </tbody>
+    </table>
+  </div>
+  <div class="card-stack">
+    <div class="card">
+      <h2>📓 Tutorial notebooks (browse <a href="https://huggingface.co/spaces/InstaDeepAI/ntv3/tree/main/notebooks_tutorials" target="_blank" rel="noopener">folder</a>)</h2>
+      <ul>
+        <li><a href="https://huggingface.co/spaces/InstaDeepAI/ntv3/blob/main/notebooks_tutorials/00_quickstart_inference.ipynb" target="_blank" rel="noopener">🚀 00 — Quickstart inference</a></li>
+        <li><a href="https://huggingface.co/spaces/InstaDeepAI/ntv3/blob/main/notebooks_tutorials/01_tracks_prediction.ipynb" target="_blank" rel="noopener">📊 01 — Tracks prediction</a></li>
+        <li>🎯 02 — Fine-tune on bigwig tracks</li>
+        <li>🔍 03 — Model interpretation</li>
+        <li>🧪 04 — Training NTv3 generative </li>
+      </ul>
+    </div>
+    <div class="card">
+      <h2>📓 Pipelines notebooks (browse <a href="https://huggingface.co/spaces/InstaDeepAI/ntv3/tree/main/notebooks_pipelines" target="_blank" rel="noopener">folder</a>)</h2>
+      <ul>
+        <li> 🎯 01 — Generate bigwig predictions for certain tracks</li>
+        <li><a href="https://huggingface.co/spaces/InstaDeepAI/ntv3/blob/main/notebooks_pipelines/02_genome_annotation.ipynb" target="_blank" rel="noopener">🏷️ 02 — Genome annotation / segmentation</a></li>
+        <li>🎯 03 — Fine-tune on bigwig tracks</li>
+        <li>🔍 04 — Interpret a given genomic region</li>
+        <li>🧪 05 — Sequence generation</li>
+      </ul>
+    </div>
+    <div class="card">
+      <h2>🔗 Links</h2>
+      <ul>
+        <li>📄 Paper: (add link)</li>
+        <li><a href="https://github.com/instadeepai/nucleotide-transformer">💻 JAX model code (GitHub)</a></li>
+        <li><a href="https://huggingface.co/collections/InstaDeepAI/nucleotide-transformer-v3" target="_blank" rel="noopener">🎯 HF Model Collection (all NTv3 models)</a></li>
+        <li><a href="https://huggingface.co/spaces/InstaDeepAI/ntv3/tree/main/notebooks" target="_blank" rel="noopener">📓 All notebooks</a></li>
+        <li><a href="https://huggingface.co/spaces/InstaDeepAI/ntv3_benchmark" target="_blank" rel="noopener">🏆 NTv3 benchmark leaderboard</a></li>
+      </ul>
+    </div>
+  </div>
+  <div class="card">
+    <h2>🤖 Load a pre-trained model</h2>
+    <p>Here is an example of how to load and use a pre-trained NTv3 model.</p>
+    <div class="code"><pre><code class="language-python">from transformers import AutoTokenizer, AutoModelForMaskedLM
+model_name = "InstaDeepAI/NTv3_650M_pre"
+# Load model and tokenizer
+model = AutoModelForMaskedLM.from_pretrained(model_name, trust_remote_code=True)
+tok = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
+# Tokenize input sequences
+batch = tok(["ATCGNATCG", "ACGT"], add_special_tokens=False, padding=True, pad_to_multiple_of=128, return_tensors="pt")
+# Run model
+out = model(
+  **batch,
+  output_hidden_states=True,
+  output_attentions=True
+)
+# Print output shapes
+print(out.logits.shape)       # (B, L, V = 11)
+print(len(out.hidden_states)) # convs + transformers + deconvs
+print(len(out.attentions))    # equals transformer layers = 12
+</code></pre></div>
+    <p>Model embeddings can be used for fine-tuning on downstream tasks.</p>
+    <p style="margin-top: 40px;">TO DO: add pipeline for fine-tuning on functional tracks or genome annotation.</p>
+  </div>
+  <div class="card">
+    <h2>💻 Use a post-trained model</h2>
+    <p>Here is a quick example of how to use the post-trained NTv3 650M model to predict tracks for a human genomic window.</p>
+    <div class="code"><pre><code class="language-python">from transformers import pipeline
+import torch
+model_name = "InstaDeepAI/NTv3_650M_pos"
+ntv3_tracks = pipeline(
+    "ntv3-tracks",
+    model=model_name,
+    trust_remote_code=True,
+    device=0 if torch.cuda.is_available() else -1,
+)
+# Run track prediction
+out = ntv3_tracks(
+  {
+    "chrom": "chr19",
+    "start": 6_700_000,
+    "end": 6_831_072,
+    "species": "human"
+  }
+)
+# Print output shapes
+# 7k human tracks over 37.5 % center region of the input sequence
+print("bigwig_tracks_logits:", tuple(out.bigwig_tracks_logits.shape))
+# Location of 21 genomic elements over 37.5 % center region of the input sequence
+print("bed_tracks_logits:", tuple(out.bed_tracks_logits.shape))
+# Language model logits for whole sequence over vocabulary
+print("language model logits:", tuple(out.mlm_logits.shape))</code></pre></div>
+    <p>Predictions can also be plotted for a subset of functional tracks and genomic elements:</p>
+    <div class="code"><pre><code class="language-python">tracks_to_plot = {
+    "K562 RNA-seq": "ENCSR056HPM",
+    "K562 DNAse": "ENCSR921NMD",
+    "K562 H3k4me3": "ENCSR000DWD",
+    "K562 CTCF": "ENCSR000AKO",
+    "HepG2 RNA-seq": "ENCSR561FEE_P",
+    "HepG2 DNAse": "ENCSR000EJV",
+    "HepG2 H3k4me3": "ENCSR000AMP",
+    "HepG2 CTCF": "ENCSR000BIE",
+}
+elements_to_plot = ["protein_coding_gene", "exon", "intron", "splice_donor", "splice_acceptor"]
+out = ntv3_tracks(
+    {"chrom": "chr19", "start": 6_700_000, "end": 6_831_072, "species": "human"},
+    plot=True,
+    tracks_to_plot=tracks_to_plot,
+    elements_to_plot=elements_to_plot,
+)</code></pre></div>
+    <img src="assets/output_tracks.png" alt="Output tracks visualization" style="max-width: 100%; margin-top: 20px;" />
+  </div>
+</div>