Dongwei Jiang

I'm a bit confused about the experimental setup for speaker recognition described in the original PASE paper. If my understanding is correct, only 15s * 2484, or 10.35 hours of...

Interpolation of different language models in kenlm

1

Is there any easy-to-use binary like 'compute-best-mix' from srilm? I've glanced through the [code](https://github.com/kpu/kenlm/blob/master/lm/builder/interpolate.cc) in kenlm but I'm not sure how can I use it.

Is there any other work that combines contrastive objective and reconstruction objective

1

Hi Yao-Hung, it seems like the method of combining contrastive objective and reconstruction objective is really effective in your work. I'm wondering if this method can be extended to similar...

I'm trying to finetune ATLAS on HotPotQA with 4 80G A100 GPU. I'm wondering which script should I use to train the model. Also, am I correct in understanding that...

Problem transforming tensorflow model to tensorrt

1

We're having issue transforming tensorflow model (with embedding) to try. But the following error occurred. ``` File "/home/research/jiangdongwei/miniconda3/envs/cuda11/lib/python3.6/site-packages/tensorflow/python/compiler/tensorrt/trt_convert.py", line 1108, in convert self._converted_graph_def = self._run_conversion(grappler_meta_graph_def) File "/home/research/jiangdongwei/miniconda3/envs/cuda11/lib/python3.6/site-packages/tensorflow/python/compiler/tensorrt/trt_convert.py", line 1042, in...

What is the meaning of [lower_control_flow](https://github.com/tensorflow/tensorflow/blob/765ac8d16eff6d6ff997ee73809b402d8b1194ae/tensorflow/python/framework/convert_to_constants.py#L380) in TensorRT? We're seeing different issues setting lower_control_flow toTrue or False.

I'm wondering whether TensorRT supports while loop under tensorflow 2.4. We want to convert a transformer based model to TensorRT but we're having issue with the while loop in transformer...

How to serve fine-tuned model with vllm

2

After training, the output folder only contain files like `meta_model_0.pt`. If I try to use vllm server to serve this model like this: `python -m vllm.entrypoints.openai.api_server --model finetuned_model_path --dtype bfloat16...

Will Pyke always give the correct answer for the query

2

I'm conducting trials with ProofWriter and stumbled upon an instance where, despite the formalization appearing accurate, it's yielding inaccurate results. The Problem is is `ProofWriter_RelNoneg-OWA-D5-775_Q9`, it's taken from the development...