-
单卡部署qwen-7b推理大概需要9g左右的显存，然后给他放到两张8g的卡上跑推理，0卡的显存占用是2g多，1卡的显存占用是7.8g，输入你好，模型可以正常回答，复杂的问题模型就会显存爆掉，但是这个过程中发现，发现0卡的显存好像没有加，反而是1卡的显存继续往上加爆掉了，这个如何调整？
-
I am trying to finetune 'bigcode/starcoder' using Peft.
I try to set the target_module as follows:
config = LoraConfig(
r=16,
lora_alpha=32,
lora_dropout=0.05,
bias="none",
…
-
I have tried to train stabilityLM via stabilityai through PEFT. Tried multiple configs with streaming data, single file, small batch sizes. But it always run into GPU out of memory issue. Currently us…
-
## Environment
```
Collecting system information...
---------------------------------
System Environment Report
Created: 2023-07-03 11:26:38 CST
---------------------------------
…
-
[data.py](https://github.com/nomic-ai/gpt4all/blob/main/data.py) line 47-50
What are they?
Can I delete them out?
I am trying to train the model by the following and it raise the error.
```pyt…
-
**Have you tried to reset Item Piles' system settings?**
Yes.
**Describe the bug**
When using the "Populate Items" the module does not populate items. It used to do so without a problem but now …
-
### Issue you'd like to raise.
Python 3.10.7
langchain 0.0.301
I started exploring langchain features such as stuff mode, map reduce mode and refine mode.
i would like to know more about the map…
-
Cannot run `python data_prep/convert_dataset_hf.py --dataset c4 --data_subset en --out_root my-copy-c4 --splits train_small val_small --concat_tokens 2048 --tokenizer EleutherAI/gpt-neox-20b --e…
-
## Describe the bug
There seems to be an issue with using multiprocessing with `datasets.Dataset.map` (i.e. setting `num_proc` to a value greater than one) combined with a function that uses `torch` …
-
Hey @turboderp, @aljungberg
Firstly thank you for the awesome repo!!
I am running into issues I think it is related to #160 and #128.
When running LLama2 13b at full context length (4096 toke…