Improve Inferentia docs/examples

docs/deployments/inferentia.md

@@ -44,7 +44,8 @@ import tensorflow.neuron as tfn
 tfn.saved_model.compile(
     model_dir,
     compiled_model_dir,
-    compiler_args=["--num-neuroncores", "1"]
+    batch_size,
+    compiler_args=["--num-neuroncores", "1"],
 )
 ```
 
@@ -54,6 +55,7 @@ Here is an example of compiling a PyTorch model for Inferentia:
 import torch_neuron, torch
 
 model.eval()
+example_input = torch.zeros([batch_size] + input_shape, dtype=torch.float32)
 model_neuron = torch.neuron.trace(
     model,
     example_inputs=[example_input],

examples/pytorch/image-classifier-resnet50/README.md

@@ -43,7 +43,9 @@ This example deploys models that we have built and uploaded to a public S3 bucke
 Run the following command to install the dependencies required for the [generate_resnet50_models.ipynb](generate_resnet50_models.ipynb) notebook:
 
 ```bash
-pip install neuron-cc==1.0.9410.0+6008239556 torch-neuron==1.0.825.0
+pip install --extra-index-url=https://pip.repos.neuron.amazonaws.com \
+ neuron-cc==1.0.9410.0+6008239556 \
+ torch-neuron==1.0.825.0
 ```
 
 Also, `torchvision` has to be installed, but without any dependencies:

examples/pytorch/image-classifier-resnet50/generate_resnet50_models.ipynb

@@ -67,7 +67,8 @@
    ],
    "source": [
     "model.eval()\n",
-    "image = torch.zeros([1, 3, 224, 224], dtype=torch.float32)\n",
+    "batch_size = 1\n",
+    "image = torch.zeros([batch_size, 3, 224, 224], dtype=torch.float32)\n",
     "model_neuron = torch.neuron.trace(model, example_inputs=[image], compiler_args=[\"--num-neuroncores\", \"2\"])"
    ]
   },

examples/tensorflow/image-classifier-resnet50/README.md

@@ -84,7 +84,9 @@ This example deploys models that we have built and uploaded to a public S3 bucke
 Run the following command to install the dependencies required for the [generate_resnet50_models.ipynb](generate_resnet50_models.ipynb) notebook:
 
 ```bash
-pip install neuron-cc==1.0.9410.0+6008239556 tensorflow-neuron==1.15.0.1.0.1333.0
+pip install --extra-index-url=https://pip.repos.neuron.amazonaws.com \
+  neuron-cc==1.0.9410.0+6008239556 \
+  tensorflow-neuron==1.15.0.1.0.1333.0
 ```
 
 The [generate_resnet50_models.ipynb](generate_resnet50_models.ipynb) notebook will generate 2 SavedModels. One will be saved in the `resnet50` directory which can be run on GPU or on CPU and another in the `resnet50_neuron` directory which can only be run on `inf1` instances.

examples/tensorflow/image-classifier-resnet50/generate_resnet50_models.ipynb

@@ -149,7 +149,8 @@
    ],
    "source": [
     "compiler_args = ['--static-weights', '--num-neuroncores', '1']\n",
-    "tfn.saved_model.compile(model_dir, compiled_model_dir)"
+    "batch_size = 1\n",
+    "tfn.saved_model.compile(model_dir, compiled_model_dir, batch_size)"
    ]
   }
  ],