Comments (3)
The language model of the Qwen2 series does not support float16 reasoning. The official document mentioned this. Please refer to https://qwen.readthedocs.io/en/latest/inference/chat.html and check Troubleshooting section
from transformers.
Qwen2系列的语言模型不支持float16推理,官方文档提到了这一点,请参考https://qwen.readthedocs.io/en/latest/inference/chat.html 并查看Troubleshooting部分
For NVIDIA GPUs, being able to perform inference with float32 and bfloat16 but not with float16 is highly unusual and goes against common sense.
I have resolved this issue and will submit a PR for everyone's reference after this PR is merged.
from transformers.
Qwen2系列的语言模型不支持float16推理,官方文档提到了这一点,请参考https://qwen.readthedocs.io/en/latest/inference/chat.html 并查看Troubleshooting部分
Thank you. I've taken a look at the Qwen team's solution, and while it's different from mine, it's quite valuable for reference. I'll test it out tomorrow.
from transformers.
Related Issues (20)
- can't resume lora training due to wandb logging num params
- Phi-3's tokenizer `.convert_ids_to_tokens` returns bytes instead of a string HOT 1
- The forced adoption of low_cpu_mem_usage behavior ruins models that require absolute position embedding. HOT 2
- Phi-3 maximum recursion depth exceeded when execute AutoModelForCausalLM.from_pretrained HOT 2
- for LlaVA output of image_hidden_states not possible? HOT 1
- Bug with finetuning Gemma 2 models HOT 3
- DINOV2 generate different outputs for the same inputs HOT 6
- Can't run docker because of flash attention 2? HOT 1
- Training Resumes with Increased Loss Despite Checkpoint Loading HOT 1
- How do LLMs identify generation start point during fine-tuning? HOT 1
- [Nougat] image after NougatProcessor leads to Traceback HOT 1
- Add "EAT: Self-Supervised Pre-Training with Efficient Audio Transformer"
- How to install transformers==4.45, two or three days I can install successfully, but today cannot. HOT 2
- Accelerate x Trainer issue tracker: HOT 7
- ImportError: attempted relative import with no known parent package HOT 2
- List of Typos in the `transformers[en]` Docs
- bus error on version 4.43.0 with pretrained community CLIP model - MacOS HOT 5
- Can't save quantized models HOT 7
- [Docs] How to build offline HTML or Docset files for other documentation viewers? HOT 1
- Pipelines download the latest revision by default instead of the default/output revision HOT 2
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from transformers.