| People |
This Week |
Next Week |
Task Tracking (DeadLine)
|
| Dong Wang
|
- AI middle school hand book / English version almost completed.
|
|
|
| Lantian Li
|
- Recheck AI textbooks for senior middle school.
- MoE daily work (busy week)
|
- Go on the final review of my MLA book (6/10)
|
|
| Wenqiang Du
|
- Project work about AIbabel
- Upgraded online model
- New demand technology survey
|
|
|
| Yang Wei
|
- Analyze uy/ch_child data annotation and exp result.
- Test the av speech separation model with a dataset with more speakers.
|
|
|
| Ying Shi
|
|
|
|
| Yue Gu
|
|
|
|
| Lily
|
- Organized course materials production （小初高分册）
- Participated in reviewing the AI handbooks
- Miyun 3-high school AI report
|
|
|
| Pengqi Li
|
- Experimental analysis revealed the unreliability of the insertion and deletion metrics adapted from face verification.[1]
- Proposed and implemented a novel framework for assessing insertion and deletion.(still test)
|
|
|
| Junming Yuan
|
- learn-not-to-listen MT-HuBERT
- PR(PER%), LNTL-MT-HuBERT: 7.24%, Clean-HuBERT: 5.41%, WavLM: 4.84%
- ASR(WER%), LNTL-MT-HuBERT: 8.43%, Clean-HuBERT: 6.42%, WavLM: 6.21%
- SD(DER%), LNTL-MT-HuBERT: 4.31%, Cocktail-HuBERT: 4.07%, MT-HuBERT: 3.93%
- revise ICASSP final paper
|
|
|
| Yu Zhang
|
- GPU Util: [2]
- LLM
- Completed the code for Swarm MMLU metric-based optimization together with @chenjunhui
- Still working on modifying the Swarm topology code (for more complex swarm topology)
|
|
|
| Junhui Chen
|
- (caught a cold)
- MMLU
- difference in the ECS distribution between good nodes and bad nodes.
- metrics loss helps select high-quality nodes and eliminate low-quality ones. [3]
|
|
|
| Jiaying Wang
|
- spk order separation：2-mix，3-mix training finished，testing； 4-mix are training
- loudness order verification done
- content order verification 1/3, will be finished these days
|
|
|
| Xiaoxue Luo
|
- prepare 2mix speech separation results for ASR test
- 2-5mix multi_head separation model for Huawei project
- data: remove data that the target audio is mixed audio(input equals output, model does not learn)
- loss: add MSE_loss to the original SISDR_loss to control the energy amplitude of output audio
|
|
|
| Bochao Hu
|
- read papers
- prepare the weekly report
|
|
|
| Hongcheng Zhang
|
- made zero-shot on clotho and Audiocaps
- result：ROUGE-L（0.2931，0.3998），CIDEr（0.0986，0.2312），SPIDEr（0.0800，0.1762），（clotho，Audiocaps）
|
|
|
| Weiman Sun
|
- Ran a completely opposite ASR task on the same data
- Consistent with last week's result
- Analyze different models
|
|
|
