sinabs 3.1.0__tar.gz → 3.1.1__tar.gz

{sinabs-3.1.0 → sinabs-3.1.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sinabs
-Version: 3.1.0
+Version: 3.1.1
 Summary: SynSense Spiking Neural Network simulator for deep neural networks (DNNs).
 Author: SynSense (formerly AiCTX)
 Author-email: support@synsense.ai
@@ -87,7 +87,7 @@ Sinabs is published under Apache v2.0. See the LICENSE file for details.
 
 Contributing to Sinabs
 ------------------------
-Checkout the [contributing](https://sinabs.readthedocs.io/en/develop/about/contributing.html) page for more info.
+Checkout the [contributing](https://sinabs.readthedocs.io/develop/about/contributing.html) page for more info.
 
 
 Citation

{sinabs-3.1.0 → sinabs-3.1.1}/README.md

@@ -49,7 +49,7 @@ Sinabs is published under Apache v2.0. See the LICENSE file for details.
 
 Contributing to Sinabs
 ------------------------
-Checkout the [contributing](https://sinabs.readthedocs.io/en/develop/about/contributing.html) page for more info.
+Checkout the [contributing](https://sinabs.readthedocs.io/develop/about/contributing.html) page for more info.
 
 
 Citation

{sinabs-3.1.0 → sinabs-3.1.1}/docs/about/release_notes.md

@@ -1,15 +1,22 @@
 # Release notes
 
+## Unreleased
+
+* Fix NIR export of Conv1d layer which expected an input_shape parameter.
+* Fix broken link on documentation.
+* Fix missing plots on `Neuron Models` gallery.
+* Fix `open_device` function that would thrown an error if an unsupported SynSense device was connected to the computer.
+
 ## v3.1.0 (31/10/2025)
 
-* Fix tests crashing in case an unexpected board was found.
-* Fix tests that were not running for speck2fdevkit.
-* Fix error when initializing sinabs without matplotlib: matplotlib is not a sinabs dependency.
-* Add mapping of non-sequential networks
+* Add mapping of non-sequential networks:
     * Now, layers in Sinabs models can receive inputs from and send outputs to multiple layers.
     * Deprecate `chip_layers_ordering` from DynapCNNNetwork. Use `layer2core_map` instead.
     * Deprecate `DynapcnnCompatibleNetwork`. Use `DynapcnnNetwork` instead.
     * Deprecate `merge_conv_bn` from `sinabs.utils`. Use `merge_bn` instead.
+* Fix tests crashing in case an unexpected board was found.
+* Fix tests that were not running for speck2fdevkit.
+* Fix error when initializing sinabs without matplotlib: matplotlib is not a sinabs dependency.
 
 ## v3.0.4 (08/09/2025)
 

{sinabs-3.1.0 → sinabs-3.1.1}/docs/conf.py

@@ -31,7 +31,6 @@ sphinx_gallery_conf = {
     # "backreferences_dir": "gen_modules/backreferences",
     "doc_module": ("sinabs",),
     "download_all_examples": False,
-    "ignore_pattern": r"utils\.py",
 }
 
 # Napoleon settings

{sinabs-3.1.0 → sinabs-3.1.1}/docs/gallery/layers/utils.py

@@ -1,8 +1,14 @@
+"""
+==============================
+Utility Function
+==============================
+Utility function for plots in the documentation.
+"""
+
 import matplotlib.pyplot as plt
 import torch
 
 import sinabs
-import sinabs.layers as sl
 
 
 def plot_evolution(neuron_model: sinabs.layers, input: torch.Tensor):

{sinabs-3.1.0 → sinabs-3.1.1}/docs/speck/notebooks/nmnist_quick_start.ipynb

@@ -37,7 +37,7 @@
     "except ImportError:\n",
     "    ! pip install tonic\n",
     "    from tonic.datasets.nmnist import NMNIST\n",
-    "    \n",
+    "\n",
     "# download dataset\n",
     "root_dir = \"./NMNIST\"\n",
     "_ = NMNIST(save_to=root_dir, train=True)\n",
@@ -64,7 +64,9 @@
     "sample_data, label = NMNIST(save_to=root_dir, train=False)[0]\n",
     "\n",
     "print(f\"type of data is: {type(sample_data)}\")\n",
-    "print(f\"time length of sample data is: {sample_data['t'][-1] - sample_data['t'][0]} micro seconds\")\n",
+    "print(\n",
+    "    f\"time length of sample data is: {sample_data['t'][-1] - sample_data['t'][0]} micro seconds\"\n",
+    ")\n",
     "print(f\"there are {len(sample_data)} events in the sample data\")\n",
     "print(f\"the label of the sample data is: {label}\")"
    ]
@@ -99,15 +101,26 @@
     "# define a CNN model\n",
     "cnn = nn.Sequential(\n",
     "    # [2, 34, 34] -> [8, 17, 17]\n",
-    "    nn.Conv2d(in_channels=2, out_channels=8, kernel_size=(3, 3), padding=(1, 1), bias=False),\n",
+    "    nn.Conv2d(\n",
+    "        in_channels=2, out_channels=8, kernel_size=(3, 3), padding=(1, 1), bias=False\n",
+    "    ),\n",
     "    nn.ReLU(),\n",
     "    nn.AvgPool2d(2, 2),\n",
     "    # [8, 17, 17] -> [16, 8, 8]\n",
-    "    nn.Conv2d(in_channels=8, out_channels=16, kernel_size=(3, 3), padding=(1, 1), bias=False),\n",
+    "    nn.Conv2d(\n",
+    "        in_channels=8, out_channels=16, kernel_size=(3, 3), padding=(1, 1), bias=False\n",
+    "    ),\n",
     "    nn.ReLU(),\n",
     "    nn.AvgPool2d(2, 2),\n",
     "    # [16 * 8 * 8] -> [16, 4, 4]\n",
-    "    nn.Conv2d(in_channels=16, out_channels=16, kernel_size=(3, 3), padding=(1, 1), stride=(2, 2),  bias=False),\n",
+    "    nn.Conv2d(\n",
+    "        in_channels=16,\n",
+    "        out_channels=16,\n",
+    "        kernel_size=(3, 3),\n",
+    "        padding=(1, 1),\n",
+    "        stride=(2, 2),\n",
+    "        bias=False,\n",
+    "    ),\n",
     "    nn.ReLU(),\n",
     "    # [16 * 4 * 4] -> [10]\n",
     "    nn.Flatten(),\n",
@@ -160,7 +173,9 @@
     "\n",
     "# check the transformed data\n",
     "sample_data, label = cnn_train_dataset[0]\n",
-    "print(f\"The transformed array is in shape [Time-Step, Channel, Height, Width] --> {sample_data.shape}\")"
+    "print(\n",
+    "    f\"The transformed array is in shape [Time-Step, Channel, Height, Width] --> {sample_data.shape}\"\n",
+    ")"
    ]
   },
   {
@@ -306,8 +321,20 @@
     "\n",
     "cnn = cnn.to(device=device)\n",
     "\n",
-    "cnn_train_dataloader = DataLoader(cnn_train_dataset, batch_size=batch_size, num_workers=num_workers, drop_last=True, shuffle=shuffle)\n",
-    "cnn_test_dataloader = DataLoader(cnn_test_dataset, batch_size=batch_size, num_workers=num_workers, drop_last=True, shuffle=shuffle)\n",
+    "cnn_train_dataloader = DataLoader(\n",
+    "    cnn_train_dataset,\n",
+    "    batch_size=batch_size,\n",
+    "    num_workers=num_workers,\n",
+    "    drop_last=True,\n",
+    "    shuffle=shuffle,\n",
+    ")\n",
+    "cnn_test_dataloader = DataLoader(\n",
+    "    cnn_test_dataset,\n",
+    "    batch_size=batch_size,\n",
+    "    num_workers=num_workers,\n",
+    "    drop_last=True,\n",
+    "    shuffle=shuffle,\n",
+    ")\n",
     "\n",
     "optimizer = SGD(params=cnn.parameters(), lr=lr)\n",
     "criterion = CrossEntropyLoss()\n",
@@ -329,7 +356,9 @@
     "        loss.backward()\n",
     "        optimizer.step()\n",
     "        # set progressing bar\n",
-    "        train_p_bar.set_description(f\"Epoch {e} - Training Loss: {round(loss.item(), 4)}\")\n",
+    "        train_p_bar.set_description(\n",
+    "            f\"Epoch {e} - Training Loss: {round(loss.item(), 4)}\"\n",
+    "        )\n",
     "\n",
     "    # validate\n",
     "    correct_predictions = []\n",
@@ -348,9 +377,11 @@
     "            correct_predictions.append(pred.eq(label.view_as(pred)))\n",
     "            # set progressing bar\n",
     "            test_p_bar.set_description(f\"Epoch {e} - Testing Model...\")\n",
-    "    \n",
+    "\n",
     "        correct_predictions = torch.cat(correct_predictions)\n",
-    "        print(f\"Epoch {e} - accuracy: {correct_predictions.sum().item()/(len(correct_predictions))*100}%\")"
+    "        print(\n",
+    "            f\"Epoch {e} - accuracy: {correct_predictions.sum().item()/(len(correct_predictions))*100}%\"\n",
+    "        )"
    ]
   },
   {
@@ -391,7 +422,9 @@
    "source": [
     "from sinabs.from_torch import from_model\n",
     "\n",
-    "snn_convert = from_model(model=cnn, input_shape=(2, 34, 34), batch_size=batch_size).spiking_model\n",
+    "snn_convert = from_model(\n",
+    "    model=cnn, input_shape=(2, 34, 34), batch_size=batch_size\n",
+    ").spiking_model\n",
     "snn_convert"
    ]
   },
@@ -434,7 +467,13 @@
     "n_time_steps = 100\n",
     "to_raster = ToFrame(sensor_size=NMNIST.sensor_size, n_time_bins=n_time_steps)\n",
     "snn_test_dataset = NMNIST(save_to=root_dir, train=False, transform=to_raster)\n",
-    "snn_test_dataloader = DataLoader(snn_test_dataset, batch_size=batch_size, num_workers=num_workers, drop_last=True, shuffle=False)\n",
+    "snn_test_dataloader = DataLoader(\n",
+    "    snn_test_dataset,\n",
+    "    batch_size=batch_size,\n",
+    "    num_workers=num_workers,\n",
+    "    drop_last=True,\n",
+    "    shuffle=False,\n",
+    ")\n",
     "\n",
     "snn_convert = snn_convert.to(device)\n",
     "\n",
@@ -459,7 +498,9 @@
     "        test_p_bar.set_description(f\"Testing SNN Model...\")\n",
     "\n",
     "    correct_predictions = torch.cat(correct_predictions)\n",
-    "    print(f\"accuracy of converted SNN: {correct_predictions.sum().item()/(len(correct_predictions))*100}%\")"
+    "    print(\n",
+    "        f\"accuracy of converted SNN: {correct_predictions.sum().item()/(len(correct_predictions))*100}%\"\n",
+    "    )"
    ]
   },
   {
@@ -511,20 +552,39 @@
     "# just replace the ReLU layer with the sl.IAFSqueeze\n",
     "snn_bptt = nn.Sequential(\n",
     "    # [2, 34, 34] -> [8, 17, 17]\n",
-    "    nn.Conv2d(in_channels=2, out_channels=8, kernel_size=(3, 3), padding=(1, 1), bias=False),\n",
-    "    sl.IAFSqueeze(batch_size=batch_size, min_v_mem=-1.0, surrogate_grad_fn=PeriodicExponential()),\n",
+    "    nn.Conv2d(\n",
+    "        in_channels=2, out_channels=8, kernel_size=(3, 3), padding=(1, 1), bias=False\n",
+    "    ),\n",
+    "    sl.IAFSqueeze(\n",
+    "        batch_size=batch_size, min_v_mem=-1.0, surrogate_grad_fn=PeriodicExponential()\n",
+    "    ),\n",
     "    nn.AvgPool2d(2, 2),\n",
     "    # [8, 17, 17] -> [16, 8, 8]\n",
-    "    nn.Conv2d(in_channels=8, out_channels=16, kernel_size=(3, 3), padding=(1, 1), bias=False),\n",
-    "    sl.IAFSqueeze(batch_size=batch_size, min_v_mem=-1.0, surrogate_grad_fn=PeriodicExponential()),\n",
+    "    nn.Conv2d(\n",
+    "        in_channels=8, out_channels=16, kernel_size=(3, 3), padding=(1, 1), bias=False\n",
+    "    ),\n",
+    "    sl.IAFSqueeze(\n",
+    "        batch_size=batch_size, min_v_mem=-1.0, surrogate_grad_fn=PeriodicExponential()\n",
+    "    ),\n",
     "    nn.AvgPool2d(2, 2),\n",
     "    # [16 * 8 * 8] -> [16, 4, 4]\n",
-    "    nn.Conv2d(in_channels=16, out_channels=16, kernel_size=(3, 3), padding=(1, 1), stride=(2, 2),  bias=False),\n",
-    "    sl.IAFSqueeze(batch_size=batch_size, min_v_mem=-1.0, surrogate_grad_fn=PeriodicExponential()),\n",
+    "    nn.Conv2d(\n",
+    "        in_channels=16,\n",
+    "        out_channels=16,\n",
+    "        kernel_size=(3, 3),\n",
+    "        padding=(1, 1),\n",
+    "        stride=(2, 2),\n",
+    "        bias=False,\n",
+    "    ),\n",
+    "    sl.IAFSqueeze(\n",
+    "        batch_size=batch_size, min_v_mem=-1.0, surrogate_grad_fn=PeriodicExponential()\n",
+    "    ),\n",
     "    # [16 * 4 * 4] -> [10]\n",
     "    nn.Flatten(),\n",
     "    nn.Linear(16 * 4 * 4, 10, bias=False),\n",
-    "    sl.IAFSqueeze(batch_size=batch_size, min_v_mem=-1.0, surrogate_grad_fn=PeriodicExponential()),\n",
+    "    sl.IAFSqueeze(\n",
+    "        batch_size=batch_size, min_v_mem=-1.0, surrogate_grad_fn=PeriodicExponential()\n",
+    "    ),\n",
     ")\n",
     "\n",
     "# init the model weights\n",
@@ -585,6 +645,7 @@
    "source": [
     "try:\n",
     "    from sinabs.exodus import conversion\n",
+    "\n",
     "    snn_bptt = conversion.sinabs_to_exodus(snn_bptt)\n",
     "except ImportError:\n",
     "    print(\"Sinabs-exodus is not intalled.\")\n",
@@ -668,8 +729,20 @@
     "device = \"cuda:0\"\n",
     "shuffle = True\n",
     "\n",
-    "snn_train_dataloader = DataLoader(snn_train_dataset, batch_size=batch_size, num_workers=num_workers, drop_last=True, shuffle=True)\n",
-    "snn_test_dataloader = DataLoader(snn_test_dataset, batch_size=batch_size, num_workers=num_workers, drop_last=True, shuffle=False)\n",
+    "snn_train_dataloader = DataLoader(\n",
+    "    snn_train_dataset,\n",
+    "    batch_size=batch_size,\n",
+    "    num_workers=num_workers,\n",
+    "    drop_last=True,\n",
+    "    shuffle=True,\n",
+    ")\n",
+    "snn_test_dataloader = DataLoader(\n",
+    "    snn_test_dataset,\n",
+    "    batch_size=batch_size,\n",
+    "    num_workers=num_workers,\n",
+    "    drop_last=True,\n",
+    "    shuffle=False,\n",
+    ")\n",
     "\n",
     "snn_bptt = snn_bptt.to(device=device)\n",
     "\n",
@@ -695,15 +768,17 @@
     "        # backward\n",
     "        loss.backward()\n",
     "        optimizer.step()\n",
-    "        \n",
+    "\n",
     "        # detach the neuron states and activations from current computation graph(necessary)\n",
     "        for layer in snn_bptt.modules():\n",
     "            if isinstance(layer, sl.StatefulLayer):\n",
     "                for name, buffer in layer.named_buffers():\n",
     "                    buffer.detach_()\n",
-    "        \n",
+    "\n",
     "        # set progressing bar\n",
-    "        train_p_bar.set_description(f\"Epoch {e} - BPTT Training Loss: {round(loss.item(), 4)}\")\n",
+    "        train_p_bar.set_description(\n",
+    "            f\"Epoch {e} - BPTT Training Loss: {round(loss.item(), 4)}\"\n",
+    "        )\n",
     "\n",
     "    # validate\n",
     "    correct_predictions = []\n",
@@ -725,9 +800,11 @@
     "            correct_predictions.append(pred.eq(label.view_as(pred)))\n",
     "            # set progressing bar\n",
     "            test_p_bar.set_description(f\"Epoch {e} - BPTT Testing Model...\")\n",
-    "    \n",
+    "\n",
     "        correct_predictions = torch.cat(correct_predictions)\n",
-    "        print(f\"Epoch {e} - BPTT accuracy: {correct_predictions.sum().item()/(len(correct_predictions))*100}%\")"
+    "        print(\n",
+    "            f\"Epoch {e} - BPTT accuracy: {correct_predictions.sum().item()/(len(correct_predictions))*100}%\"\n",
+    "        )"
    ]
   },
   {
@@ -775,6 +852,7 @@
    "source": [
     "try:\n",
     "    from sinabs.exodus import conversion\n",
+    "\n",
     "    snn_bptt = conversion.exodus_to_sinabs(snn_bptt)\n",
     "except ImportError:\n",
     "    print(\"Sinabs-exodus is not intalled.\")\n",
@@ -812,7 +890,9 @@
     "\n",
     "# cpu_snn = snn_convert.to(device=\"cpu\")\n",
     "cpu_snn = snn_bptt.to(device=\"cpu\")\n",
-    "dynapcnn = DynapcnnNetwork(snn=cpu_snn, input_shape=(2, 34, 34), discretize=True, dvs_input=False)\n",
+    "dynapcnn = DynapcnnNetwork(\n",
+    "    snn=cpu_snn, input_shape=(2, 34, 34), discretize=True, dvs_input=False\n",
+    ")\n",
     "devkit_name = \"speck2fdevkit\"\n",
     "\n",
     "# use the `to` method of DynapcnnNetwork to deploy the SNN to the devkit\n",
@@ -886,10 +966,10 @@
     "    samna_event_stream = []\n",
     "    for ev in events:\n",
     "        spk = samna.speck2f.event.Spike()\n",
-    "        spk.x = ev['x']\n",
-    "        spk.y = ev['y']\n",
-    "        spk.timestamp = ev['t'] - events['t'][0]\n",
-    "        spk.feature = ev['p']\n",
+    "        spk.x = ev[\"x\"]\n",
+    "        spk.y = ev[\"y\"]\n",
+    "        spk.timestamp = ev[\"t\"] - events[\"t\"][0]\n",
+    "        spk.feature = ev[\"p\"]\n",
     "        # Spikes will be sent to layer/core #0, since the SNN is deployed on core: [0, 1, 2, 3]\n",
     "        spk.layer = 0\n",
     "        samna_event_stream.append(spk)\n",
@@ -897,7 +977,7 @@
     "    # inference on chip\n",
     "    # output_events is also a list of Spike, but each Spike.layer is 3, since layer#3 is the output layer\n",
     "    output_events = dynapcnn(samna_event_stream)\n",
-    "    \n",
+    "\n",
     "    # use the most frequent output neruon index as the final prediction\n",
     "    neuron_index = [each.feature for each in output_events]\n",
     "    if len(neuron_index) != 0:\n",
@@ -905,14 +985,16 @@
     "        prediction = frequent_counter.most_common(1)[0][0]\n",
     "    else:\n",
     "        prediction = -1\n",
-    "    inferece_p_bar.set_description(f\"label: {label}, prediction: {prediction}， output spikes num: {len(output_events)}\") \n",
+    "    inferece_p_bar.set_description(\n",
+    "        f\"label: {label}, prediction: {prediction}， output spikes num: {len(output_events)}\"\n",
+    "    )\n",
     "\n",
     "    if prediction == label:\n",
     "        correct_samples += 1\n",
     "\n",
     "    test_samples += 1\n",
-    "    \n",
-    "print(f\"On chip inference accuracy: {correct_samples / test_samples}\")        "
+    "\n",
+    "print(f\"On chip inference accuracy: {correct_samples / test_samples}\")"
    ]
   },
   {
@@ -1004,6 +1086,8 @@
    ],
    "source": [
     "import samnawe\n",
+    "\n",
+    "\n",
     "# first define a callback function to modify the devkit configuration\n",
     "# the callback function should only has 1 devkit config instance as its input argument\n",
     "def config_modify_callback(devkit_cfg):\n",
@@ -1014,24 +1098,32 @@
     "    devkit_cfg.dvs_layer.raw_monitor_enable = False\n",
     "    # prevent the events generated by the embedded dvs been feed to the DynapCNN Core.\n",
     "    devkit_cfg.dvs_layer.pass_sensor_events = False\n",
-    "    # point the dvs layer output destination to the core#0 \n",
+    "    # point the dvs layer output destination to the core#0\n",
     "    devkit_cfg.dvs_layer.destinations[0].enable = True\n",
     "    devkit_cfg.dvs_layer.destinations[0].layer = 0\n",
     "\n",
     "    # the callback must return the modified devkit config\n",
     "    return devkit_cfg\n",
     "\n",
+    "\n",
     "# close the devkit before reopen\n",
     "samna.device.close_device(dynapcnn.samna_device)\n",
     "\n",
     "# init DynapcnnNetwork instance\n",
-    "dynapcnn = DynapcnnNetwork(snn=cpu_snn, input_shape=(2, 34, 34), discretize=True, dvs_input=True)\n",
+    "dynapcnn = DynapcnnNetwork(\n",
+    "    snn=cpu_snn, input_shape=(2, 34, 34), discretize=True, dvs_input=True\n",
+    ")\n",
     "\n",
     "devkit_name = \"speck2fdevkit\"\n",
     "# define which layers output you want to monitor\n",
     "layers_to_monitor = [0, 1, 2, 3]\n",
     "# pass the callback function into the `.to` method\n",
-    "dynapcnn.to(device=devkit_name, chip_layers_ordering=[0, 1, 2, 3], monitor_layers=layers_to_monitor, config_modifier=config_modify_callback)\n",
+    "dynapcnn.to(\n",
+    "    device=devkit_name,\n",
+    "    chip_layers_ordering=[0, 1, 2, 3],\n",
+    "    monitor_layers=layers_to_monitor,\n",
+    "    config_modifier=config_modify_callback,\n",
+    ")\n",
     "print(f\"The SNN is deployed on the core: {dynapcnn.chip_layers_ordering}\")"
    ]
   },
@@ -1160,16 +1252,16 @@
     "    samna_event_stream = []\n",
     "    for ev in events:\n",
     "        dvs_ev = samna.speck2f.event.DvsEvent()\n",
-    "        dvs_ev.x = ev['x']\n",
-    "        dvs_ev.y = ev['y']\n",
-    "        dvs_ev.timestamp = ev['t'] - events['t'][0]\n",
-    "        dvs_ev.p = ev['p']\n",
+    "        dvs_ev.x = ev[\"x\"]\n",
+    "        dvs_ev.y = ev[\"y\"]\n",
+    "        dvs_ev.timestamp = ev[\"t\"] - events[\"t\"][0]\n",
+    "        dvs_ev.p = ev[\"p\"]\n",
     "        samna_event_stream.append(dvs_ev)\n",
     "\n",
     "    # inference on chip\n",
     "    # output_events is also a list of Spike, but .layer will have 0, 1, 2, 3 since we choose to monitor all layers' output\n",
     "    output_events = dynapcnn(samna_event_stream)\n",
-    "    \n",
+    "\n",
     "    # get each layers output spikes\n",
     "    layer0_spks = [each.feature for each in output_events if each.layer == 0]\n",
     "    layer1_spks = [each.feature for each in output_events if each.layer == 1]\n",
@@ -1181,14 +1273,16 @@
     "        prediction = frequent_counter.most_common(1)[0][0]\n",
     "    else:\n",
     "        prediction = -1\n",
-    "    inferece_p_bar.set_description(f\"label: {label} prediction: {prediction}，layer 0 output spks: {len(layer0_spks)},layer 1 output spikes num: {len(layer1_spks)}, layer 2 output spikes num: {len(layer2_spks)},layer 3 output spikes num: {len(layer3_spks)}\") \n",
+    "    inferece_p_bar.set_description(\n",
+    "        f\"label: {label} prediction: {prediction}，layer 0 output spks: {len(layer0_spks)},layer 1 output spikes num: {len(layer1_spks)}, layer 2 output spikes num: {len(layer2_spks)},layer 3 output spikes num: {len(layer3_spks)}\"\n",
+    "    )\n",
     "\n",
     "    if prediction == label:\n",
     "        correct_samples += 1\n",
     "\n",
     "    test_samples += 1\n",
-    "    \n",
-    "print(f\"On chip inference accuracy: {correct_samples / test_samples}\")    "
+    "\n",
+    "print(f\"On chip inference accuracy: {correct_samples / test_samples}\")"
    ]
   },
   {

{sinabs-3.1.0 → sinabs-3.1.1}/docs/speck/notebooks/using_readout_layer.ipynb

@@ -190,7 +190,7 @@
    "outputs": [],
    "source": [
     "def remapping_output_index(\n",
-    "    output_layer: Union[nn.Conv2d, nn.Linear]\n",
+    "    output_layer: Union[nn.Conv2d, nn.Linear],\n",
     ") -> Union[nn.Conv2d, nn.Linear]:\n",
     "    \"\"\"\n",
     "    Since the mapping of output channel's index from last cnn layer to the readout layer is not correct\n",

{sinabs-3.1.0 → sinabs-3.1.1}/sinabs/backend/dynapcnn/io.py

@@ -133,8 +133,11 @@ def get_device_map() -> Dict:
     # Group by device_type_name
     device_groups = groupby(devices, lambda x: x.device_type_name)
     # Switch keys from samna's device_type_name to device_type names
+    # -- guarantee is a supported device
     device_groups = {
-        device_type_map[k]: sort_devices(list(v)) for k, v in device_groups
+        device_type_map[k]: sort_devices(list(v))
+        for k, v in device_groups
+        if k in device_type_map
     }
     # Flat map
     for dev_type, dev_list in device_groups.items():

{sinabs-3.1.0 → sinabs-3.1.1}/sinabs/nir.py

@@ -46,8 +46,8 @@ def _import_sinabs_module(
             groups=node.groups,
             bias=True,
         )
-        conv.weight.data = torch.tensor(node.weight).float()
-        conv.bias.data = torch.tensor(node.bias).float()
+        conv.weight.data = node.weight.detach().clone().to(float)
+        conv.bias.data = node.bias.detach().clone().to(float)
         return conv
 
     elif isinstance(node, nir.Conv2d):
@@ -184,6 +184,7 @@ def _extract_sinabs_module(module: torch.nn.Module) -> Optional[nir.NIRNode]:
             return nir.Affine(module.weight.detach(), module.bias.detach())
     elif isinstance(module, torch.nn.Conv1d):
         return nir.Conv1d(
+            input_shape=None,
             weight=module.weight.detach(),
             stride=module.stride,
             padding=module.padding,
@@ -191,7 +192,7 @@ def _extract_sinabs_module(module: torch.nn.Module) -> Optional[nir.NIRNode]:
             groups=module.groups,
             bias=(
                 module.bias.detach()
-                if module.bias
+                if isinstance(module.bias, torch.Tensor)
                 else torch.zeros((module.weight.shape[0]))
             ),
         )

{sinabs-3.1.0 → sinabs-3.1.1}/sinabs.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sinabs
-Version: 3.1.0
+Version: 3.1.1
 Summary: SynSense Spiking Neural Network simulator for deep neural networks (DNNs).
 Author: SynSense (formerly AiCTX)
 Author-email: support@synsense.ai
@@ -87,7 +87,7 @@ Sinabs is published under Apache v2.0. See the LICENSE file for details.
 
 Contributing to Sinabs
 ------------------------
-Checkout the [contributing](https://sinabs.readthedocs.io/en/develop/about/contributing.html) page for more info.
+Checkout the [contributing](https://sinabs.readthedocs.io/develop/about/contributing.html) page for more info.
 
 
 Citation

sinabs-3.1.1/sinabs.egg-info/pbr.json

@@ -0,0 +1 @@
+{"git_version": "cf8590b", "is_release": false}

{sinabs-3.1.0 → sinabs-3.1.1}/tests/test_dynapcnn/hw_utils.py

@@ -62,9 +62,12 @@ def reset_all_connected_boards():
     import samna
 
     devs = samna.device.get_unopened_devices()
+    reverse_dict = {v: k for k, v in supported_device_types_for_testing.items()}
+
     if len(devs) > 0:  # check if the connected board is found.
         for device in devs:
-            handle = samna.device.open_device(device)
-            handle.reset_board_soft(True)
-            samna.device.close_device(handle)
-            print(f"Resetted board: {device.device_type_name}")
+            if device.device_type_name in reverse_dict:
+                handle = samna.device.open_device(device)
+                handle.reset_board_soft(True)
+                samna.device.close_device(handle)
+                print(f"Resetted board: {device.device_type_name}")

{sinabs-3.1.0 → sinabs-3.1.1}/tests/test_dynapcnn/test_visualizer.py

@@ -62,9 +62,6 @@ def get_demo_dynapcnn_network():
     return dynapcnn_network
 
 
-@pytest.mark.skip(
-    "This test was being skiiped all the time and now needs modification in the visualizer"
-)
 def test_jit_compilation():
     dvs_shape = (128, 128)
     spike_collection_interval = 500
@@ -76,22 +73,28 @@ def test_jit_compilation():
 
     dynapcnn_network = get_demo_dynapcnn_network()
     for device_name, _ in devices.items():
-        dynapcnn_network.to(device=device_name)
-
-        visualizer = DynapcnnVisualizer(
-            dvs_shape=dvs_shape, spike_collection_interval=spike_collection_interval
-        )
-        visualizer.create_visualizer_process(visualizer_id=visualizer_id)
-
-        streamer_graph = samna.graph.EventFilterGraph()
-        # Streamer graph
-        # Dvs node
-        (_, dvs_member_filter, _, streamer_node) = streamer_graph.sequential(
-            [
-                # samna.graph.JitSource(samna.speck2e.event.OutputEvent),
-                dynapcnn_network.samna_device.get_model_source_node(),
-                samna.graph.JitMemberSelect(),
-                samna.graph.JitDvsEventToViz(samna.ui.Event),
-                "VizEventStreamer",
-            ]
-        )
+        if device_name in [
+            "speck2e",
+            "speck2edevkit",
+            "speck2fmodule",
+        ]:
+
+            dynapcnn_network.to(device=device_name)
+
+            visualizer = DynapcnnVisualizer(
+                dvs_shape=dvs_shape, spike_collection_interval=spike_collection_interval
+            )
+            visualizer.create_visualizer_process(visualizer_id=visualizer_id)
+
+            streamer_graph = samna.graph.EventFilterGraph()
+            # Streamer graph
+            # Dvs node
+            (_, dvs_member_filter, _, streamer_node) = streamer_graph.sequential(
+                [
+                    # samna.graph.JitSource(samna.speck2e.event.OutputEvent),
+                    dynapcnn_network.samna_device.get_model_source_node(),
+                    samna.graph.JitMemberSelect(),
+                    samna.graph.JitDvsEventToViz(samna.ui.Event),
+                    "VizEventStreamer",
+                ]
+            )

{sinabs-3.1.0 → sinabs-3.1.1}/tests/test_nir.py

@@ -105,3 +105,18 @@ def test_2dcnn_network():
     nir_graph = to_nir(orig_model, torch.rand(1, 2, 10, 10))
 
     loaded_model = from_nir(nir_graph, batch_size=1)
+
+
+def test_conv1d():
+    batch_size = 2
+    conv1d = nn.Conv1d(16, 16, 3)
+    graph = to_nir(conv1d, torch.randn(batch_size, 16, 32))
+    converted = from_nir(graph, batch_size=batch_size)
+
+    assert len(graph.nodes) == 1 + 2
+    assert isinstance(graph.nodes["model"], nir.Conv1d)
+    assert len(graph.edges) == 0 + 2
+    assert conv1d.kernel_size == converted.model.kernel_size
+    assert conv1d.stride == converted.model.stride
+    assert conv1d.padding == converted.model.padding
+    assert conv1d.dilation == converted.model.dilation

sinabs-3.1.0/sinabs.egg-info/pbr.json

@@ -1 +0,0 @@
-{"git_version": "8b87dc3", "is_release": false}