Search Results for author:
Found 12 papers, 6 papers with code
Minions: Cost-efficient Collaboration Between On-device and Cloud Language Models
MinionS reduces costs by 5. 7x on average while recovering 97. 9% of the performance of the remote model alone.
The Mamba in the Llama: Distilling and Accelerating Hybrid Models
The resulting hybrid model, which incorporates a quarter of the attention layers, achieves performance comparable to the original Transformer in chat benchmarks and outperforms open-source hybrid Mamba models trained from scratch with trillions of tokens in both chat benchmarks and general benchmarks.
SpecExec: Massively Parallel Speculative Decoding for Interactive LLM Inference on Consumer Devices
We propose SpecExec (Speculative Execution), a simple parallel decoding method that can generate up to 20 tokens per target model iteration for popular LLM families.
Sequoia: Scalable, Robust, and Hardware-aware Speculative Decoding
This paper introduces Sequoia, a scalable, robust, and hardware-aware algorithm for speculative decoding.
Audio-visual fine-tuning of audio-only ASR models
Audio-visual automatic speech recognition (AV-ASR) models are very effective at reducing word error rates on noisy speech, but require large amounts of transcribed AV training data.
Contextual Embeddings: When Are They Worth It?
We study the settings for which deep contextual embeddings (e. g., BERT) give large improvements in performance relative to classic pretrained embeddings (e. g., GloVe), and an even simpler baseline---random word embeddings---focusing on the impact of the training set size and the linguistic properties of the task.
Understanding the Downstream Instability of Word Embeddings
To theoretically explain this tradeoff, we introduce a new measure of embedding instability---the eigenspace instability measure---which we prove bounds the disagreement in downstream predictions introduced by the change in word embeddings.
On the Downstream Performance of Compressed Word Embeddings
Finally, we show that by using the eigenspace overlap score as a selection criterion between embeddings drawn from a representative set we compressed, we can efficiently identify the better performing embedding with up to $2\times$ lower selection error rates than the next best measure of compression quality, and avoid the cost of training a model for each task of interest.
Low-Precision Random Fourier Features for Memory-Constrained Kernel Approximation
We investigate how to train kernel approximation methods that generalize well under a memory budget.
Kernel Approximation Methods for Speech Recognition
First, in order to reduce the number of random features required by kernel models, we propose a simple but effective method for feature selection.
A Comparison between Deep Neural Nets and Kernel Acoustic Models for Speech Recognition
We study large-scale kernel methods for acoustic modeling and compare to DNNs on performance metrics related to both acoustic modeling and recognition.
How to Scale Up Kernel Methods to Be As Good As Deep Neural Nets
The computational complexity of kernel methods has often been a major barrier for applying them to large-scale learning problems.
Automatic Speech Recognition
Automatic Speech Recognition (ASR)
+2
