Abstract: A method and system are provided a dataflow using a reconfigurable dataflow architecture (RDA) with repeated layered structure having synchronization boundaries between layers for a large language model used in machine learning for artificial intelligence (AI). The method first identifies in the dataflow consecutive calls to a same kernel, wherein the consecutive calls are made in an inference process involving generating a sequence of output tokens based on a given set of input tokens. The method then applies kernel looping to transform the identified consecutive calls in the dataflow into a single call to the kernel by identifying from a list of pattern matched calls ones to include in the single call to the kernel.

Abstract: A system and method provide kernel looping to eliminate synchronization boundaries to achieve peak inference performance in dataflow accelerators. The system provides a dataflow using a reconfigurable dataflow architecture (RDA) with repeated layered structure having synchronization boundaries between layers for a large language model used in machine learning. The method provides kernel looping between layers by first identifying in the dataflow consecutive calls to a same kernel, wherein the consecutive calls are made in an inference process involving generating a sequence of output tokens based on a given set of input tokens. The method then applies kernel looping to transform the identified consecutive calls in the dataflow into a single call to the kernel. The method further modifies the kernel to contain a pipelined outer loop.

Abstract: In the description and drawings a fiberglass product is disclosed. The fiberglass product may have one or more fiber curl layers. The fiber curl layers may have a plurality of generally sinusoidal fiber bundles therein.

Abstract: In the description and drawings a fiberglass product is disclosed. The fiberglass product may have one or more fiber curl layers. The fiber curl layers may have a plurality of generally sinusoidal fiber bundles therein.

Abstract: Provided herein is a tack coat for use with paving applications and a method of making the same. The tack coat is substantially non-tracking, and includes asphalt, a pH adjustor, an emulsifier, acrylic, and water. In certain embodiments, the tack coat may also include a stabilizer to prevent settling of the emulsion during storage. The method of producing the tack coat involves heating the asphalt in one tank, and heating the water and emulsifier in a second, separate tank. This mixture of the water and emulsifier is referred to as soap, and is then mixed with asphalt, and the acrylic may be added to the mixture immediately thereafter, with a stabilizer, or later, just prior to shipment of the product, in which case a stabilizer may not be required. The tack coat exhibits moderate softness, with a penetration value above 40 dmm.