-
**Describe the bug**
`zero_quantized_nontrainable_weights=True` when using PEFT+DeepSpeed with Mixed-Precision training using BF16 leads to `float != c10::BFloat16` error
**To Reproduce**
Steps …
-
Hey there, I am so interested in this terrific work, and found some questions when I tried to reproduce the results in the paper:
- Q1: Do the checkpoints released on the Hugging Face (3 * FlanT5 m…
-
Hi! I'm having the following issue on the forward pass (only when using an AQLM model) while prompt tuning an AQLM model. I'm using https://huggingface.co/BlackSamorez/Mixtral-8x7b-AQLM-2Bit-1x16-hf-t…
-
Discussion about the ongoing implementation of the compiler.
keean updated
5 years ago
-
Thank you for your excellent work on MultimodalOCR!
When I run the following command:
`GPUS=2 BATCH_SIZE=8 sh shell/minimonkey/minimonkey_finetune_full.sh`
I meet the following issue:
`
+ GP…
-
I can't run nnsight on llama models. I get a runtime error `RuntimeError: User specified an unsupported autocast device_type 'meta'`
MWE:
```py
from nnsight import LanguageModel
model = LanguageMo…
-
**Describe the bug**
I try to use deepspeed ZERO-3 with huggingface Trainer to finetune a galactica 30b model (gpt-2 like), with 4 nodes, each 4 A100 gpu. I get oom error though the model should fit …
-
## Executive Summary
This is a request to add a string literal syntax using paired Unicode delimiters, perhaps ⟪ and ⟫, for use in non-standard string literal macros. This is proposed as an altern…
-
请提供下述完整信息以便快速定位问题/Please provide the following information to quickly locate the problem
- 系统环境/System Environment:Ubuntu 20.04
- 版本号/Version:Paddle:2.6.0 PaddleNLP 2.7.2 : 问题相关组件/Related compone…
-
```
PS C:\Users\tmjj1\Downloads\joj-master\joj-master> docker-compose up
Creating network "joj-master_tjoj-network" with the default driver
Pulling blockchain_tests (node:14.3.0-slim)...
14.3.0-…