beaulian

can you provide some detailed samples for actual use? Thanks.

Implement `on_<status>` callback functions

8

Hi, I'd like to update the task status in database when a job's status changes. I wonder if these status-change callback functions can be implemented. Thanks

update a non-existent field in instance when updating this instance

1

Hi, According to the following code, when we update an instance, whatever we define other fields (not in the Model) in the Schema, these fields will be inserted to the...

Where is the `run_figure.py` file?

1

Hi @Zuricho , I wonder if you can provide the `run_figure.py` file. There is a related issue two years ago #7 Thank you

I have both used bamtools and htslib to load bam file, I find htslib is 10 times faster than bamtools, so I think bamtools need more optimization. Thanks.

the docs make me confused, is there any project for reference?

2

I have seen the `example_test.go`, but I still feel confused about how to use gorpc. For exmaple, if my implementation of client and server is segregative, how can I use...

Hello, I'm learning your code of web-proxy. I have realized the multithreading and my program should receive all the chunks, but I don't know its correctness.Can you help me with...

## Motivation Remove `column parallel` and `row parallel` comments copied from the previous version in **create_weights** function. It would cause some confusion because it's EP, not TP. ## Modifications Remove...

大佬问一下，按flashattention的理论IO复杂度分心是N^2d^2/M，Memory Access应该是非常低的，下面的代码计算出来的memory access非常大，而且用的是T_r不是T_c，和flashattention的理论分析不太一样，请问如何理解下面的计算？ ```python3 if use_flashattention: name = f"fused_attention" bandwidth, max_OPS, onchip_buffer = self.get_hardware_info() # flashattention-2 https://arxiv.org/pdf/2307.08691.pdf block_size_r = min(math.ceil(onchip_buffer / (kv_byte * head_size)), head_size) n_blocks_r = math.ceil(seqlen / block_size_r)...