Fusion 360 Sample Files

[Carolina Schmalzried]

Deartrivedi.nagaraj,

layer fusion optimization happens automatically during build phase. Please see if Best Practices For TensorRT Performance :: NVIDIA Deep Learning TensorRT Documentation helps to clarify.

This is the second clarification I have. Let us take an example code provided in the path /usr/src/tensorrt/samples/python/network_api_pytorch_mnist/sample.py

In this file it manually creates the network with the similar layers, weights and bias taken from the pretrained model (present in the file model.py in the same directory).

Dear @trivedi.nagaraj,

The layer fusion happen at the time of TRT engine building step. You can use verbose flag in trtexec to get more information about engine building. Please check trtexec --help for details.

My query is how I can make use of the logger to get layer fusion information. How and from where the logger gets this information about the builder that it has fused few layers, particularly for the programs present in the python sample directory

/usr/src/tensorrt/samples/python/network_api_pytorch_mnist/

It would be really helpful if the example files were made available in the tutorials in self-paced learning. In the getting started tutorials, the example files are available, and it makes the lessons much more helpful and meaningful. PLEASE MAKE THE EXAMPLE FILES AVAILABLE FOR ALL THE TUTORIALS since the tutorials build on the example files.

I'll ask about this one today. Let me know if you think any others are missing data or have faulty paths to the samples. I'll also ask if they can create a uniform experience rather than sending you different places for each tutorial.

Yes, I agree. We should present the tutorial designs in a consistent manner, wherever they are. As I showed above, you might have to look around. I believe that, with one exception, all the tutorial files are readily available.

Regarding the change in UI that Fusion got last year. We are aware of the old content and working to replace it. Keep in mind those old videos represent hundreds of hours of work over several years of development. It will take hundreds of hours to replace them, hopefully not several years.

By look around, I mean look at the screen when the tutorial is on screen. That is where tutorial designs are either represented by download links, or instructions are given about where to find them in your Fusion data panel.

In general, as shown above in the images, start with looking for the download link or instructions in the tutorial page. If Fusion 360 is open, it is also possible browse in "Basic Training" to find designs based on the folders that represent lessons, e.g. "06 Assemblies". All the assembly tutorial files are in that folder.

This page describes the file formats that cancer study data should assume in order to be successfully imported into the database. Unless otherwise noted, all data files are in tabular-TSV (tab separated value) format and have an associated metadata file which is in a multiline record format. The metadata and data files should follow a few rules documented at the Data Loading page.

As of March 2016, the clinical file is split into a patient clinical file and a sample clinical file. The sample file is required, whereas the patient file is optional. cBioPortal has specific functionality for a core set of patient and sample columns, but can also display custom columns (see section "Custom columns in clinical data").

For both patients and samples, the clinical data file is a two dimensional matrix with multiple clinical attributes. When the attributes are defined in the patient file they are considered to be patient attributes; when they are defined in the sample file they are considered to be sample attributes.

The first four rows of the clinical data file contain tab-delimited metadata about the clinical attributes. These rows have to start with a '#' symbol. Each of these four rows contain different type of information regarding each of the attributes that are defined in the fifth row:

Row 4: The attribute Priority: A number which indicates the importance of each attribute. In the future, higher priority attributes will appear in more prominent places than lower priority ones on relevant pages (such as the Study View). A higher number indicates a higher priority.

Please note:Priority is not the sole factor determining which chart will be displayed first.A layout algorithm in study view also makes a minor adjustment on the layout.The algorithm tries to fit all charts into a 2 by 2 matrix (Mutated Genes Table occupies 2 by 2 space).When a chart can not be fitted in the first matrix, the second matrixed will be generated.And the second matrix will have lower priority than the first one.If later chart can fit into the first matrix, then its priority will be promoted.

The following columns are used by the study view as well as the patient view. In the study view they are used to create the survival plots. In the patient view they are used to add information to the header.

Note on survival plots: to generate the survival plots successfully, the columns are required to be in pairs, which means the file should have a pair of columns that have the same prefix but ending with _STATUS and _MONTHS individually. For example, PFS_STATUS and PFS_MONTHS are a valid pair of columns that can generate the survival plots.

By adding PATIENT_ID here, cBioPortal will map the given sample to this patient. This enables one to associate multiple samples to one patient. For example, a single patient may have had multiple biopsies, each of which has been genomically profiled. See this example for a patient with multiple samples.

cBioPortal supports custom columns with clinical data in either the patient or sample file. They should follow the previously described 5-row header format. Be sure to provide the correct Datatype, for optimal search, sorting, filtering (in clinical data tab) and visualization.

The Clinical Data Dictionary from MSKCC is used to normalize clinical data, and should be followed to make the clinical data comparable between studies. This dictionary provides a definition whether an attribute should be defined on the patient or sample level, as well as provides a name, description and datatype. The data curator can choose to ignore these proposed definitions, but not following this dictionary might make comparing data between studies more difficult. It should however not break any cBioPortal functionality. See GET /api/ at -ui.html#/ for the data dictionary of all known clinical attributes.

The discrete copy number data file contain values that would be derived from copy-number analysis algorithms like GISTIC 2.0 or RAE. GISTIC 2.0 can be installed or run online using the GISTIC 2.0 module on GenePattern. For some help on using GISTIC 2.0, check the Data Loading: Tips and Best Practices page. When loading case list data, the _cna case list is required. See the case list section.

Custom driver annotations can be defined for discrete copy number data. These annotations can be used to complement or replace default driver annotation resources OncoKB and HotSpots.Custom driver annotations can be placed in a separate file that is referenced by the pd_annotations_file field of the meta file. The annotation file can hold the following columns:

The cbp_driver column flags the mutation as either driver or passenger. In cBioPortal, passenger mutations are also known as variants of unknown significance (VUS). The cbp_driver_tiers column assigns an annotation tier to the mutation, such as "Driver", "Highly actionable" or "Potential drug target". When a tier is selected, mutations with that annotation are highlighted as driver. Both types of custom annotations contain a second column with the suffix _annotation, to add a description. This is displayed in the tooltip that appears when hovering over the sample's custom annotation icon in the OncoPrint view.

You can learn more about configuring these annotations in the application.properties documentation. When properly configured, the customized annotations appear in the "Mutation Color" menu of the OncoPrint view:

schreenshot mutation color menu

GISTIC 2.0 outputs a tabular file similarly formatted to the cBioPortal format, called _all_thresholded.by_genes.txt.In this file the gene symbol is found in the Gene Symbol column, while Entrez gene IDs are in the Gene ID orLocus ID column. Please rename Gene Symbol to Hugo_Symbol and Gene ID or Locus ID to Entrez_Gene_Id. TheCytoband column can be kept in the table, but note that these values are ignored in cBioPortal. cBioPortal usescytoband annotations from the map_location column in NCBI's Homo_sapiens.gene_info.gz when loading genes intothe seed database.

GISTIC 2.0 outputs a tabular file similarly formatted to the cBioPortal format, called _all_data_by_genes.txt.In this file the gene symbol is found in the Gene Symbol column, while Entrez gene IDs are in the Gene ID orLocus ID column. Please rename Gene Symbol to Hugo_Symbol and Gene ID or Locus ID to Entrez_Gene_Id. TheCytoband column can be kept in the table, but note that these values are ignored in cBioPortal. cBioPortal usescytoband annotations from the map_location column in NCBI's Homo_sapiens.gene_info.gz when loading genes intothe seed database.

A SEG file (segmented data; .seg or .cbs) is a tab-delimited text file that lists loci and associated numeric values. The segmented data file format is the output of the Circular Binary Segmentation algorithm (Olshen et al., 2004). This Segment data enables the 'CNA' lane in the Genomic overview of the Patient view (as can be seen in this example).

For historical reasons, cBioPortal expects the stable_id to be one of those listed in the following static set.The stable_id for continuous RNA-seq data has two options: rna_seq_mrna or rna_seq_v2_mrna. These options were added to distinguish between two different TCGA pipelines, which perform different types of normalization (RPKM and RSEM). However, for custom datasets either one of these stable_id can be chosen.

3a8082e126