Cg Lai

collection_set feature request

3

**Is your feature request related to a problem? Please describe.** Hello guys, I have a SQL sample running on the Blzaingsql. I see the collection is not supported in 0.20...

Write csv without scientific notation[QST]

1

**What is your question?** Hello, I met a problem to write data into csv file with a scientific notation. Do we have a workaround to remove the scientific notation? Appreciate...

OOM for OPT 30B model using GeminiDDP [BUG]:

4

### 🐛 Describe the bug Hello, I am using OPT code to try 30B model on 8 A-100 GPUs. I got OOM when BS is lager than 8 in using...

[BUG] cuda OOM on loading opt-66B

4

**Describe the bug** Hello, I got OOM to load the `facebook/opt-66b` to GPUs (upto 96 A100-80) using zero3. I suspect that the model is not divided correctly. I use other...

Small memory saving on NVME[BUG]:

6

### 🐛 Describe the bug Hello, I am using NVME to try 66B opt model with Zero3-offload cpu mode, but it seems to only saves 10% CPU memory. It takes...

### 🐛 Describe the bug Hello, I am training OPT model on the A100 GPUs. I found it used 76GB GPU memory when I use `auto` mode and set `gpu_margin_mem_ratio`...

numeric::decimal128 not supported in device code

2

Hello, I am doing an integration between C++ and python code. I try building `shareable-dataframes` on my container with rapids 22.02 but got an error. Appreciate it if anyone has...

Hello, I see there was a PR for VLLM support but it was not active since Feb. I wonder if Flashinfer has a roadmap for vllm support. Many thanks. @yzh119