Comments (10)
Hi @zitterbewegung
Could you share the stack trace? Has the machine access to the internet?
from trl.
The weird part is that this collab notebook won't crash
https://colab.research.google.com/drive/1GE_riqtg4EiRt7BuAsrzKwTNDUIXCRMZ?usp=sharing
from trl.
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 450.51.05 Driver Version: 450.51.05 CUDA Version: 11.0 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 TITAN RTX On | 00000000:0B:00.0 On | N/A |
| 41% 35C P8 15W / 280W | 361MiB / 24219MiB | 1% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| 0 N/A N/A 2878 G /usr/lib/xorg/Xorg 199MiB |
| 0 N/A N/A 3206 G /usr/bin/gnome-shell 156MiB |
| 0 N/A N/A 102465 G /usr/lib/firefox/firefox 3MiB |
+-----------------------------------------------------------------------------+
from trl.
from trl.
How exactly did the code break? You did not get any error at all? This part of the code is not trl
specific so it might be a transformers
library issue. Have you tried just loading the tokenizer?
from trl.
from trl.
I don't quite get what the segmentation fault
is. Is that a AWS specific error or is it a Python error? If it is a Python error could you share the full error message? Loading pretrained models is possible in transformers==2.6.0
but you could try to be sure. However, there are issues with the trl
libraries in more recent versions (see #9).
from trl.
It could be related to this: huggingface/transformers#4857
from trl.
from trl.
Needed to use requirements.txt in a virtualenv instead of conda
from trl.
Related Issues (20)
- `OnPolicyConfig` - Rename or revise `num_sample_generations` HOT 1
- AttributeError: 'PPOv2Trainer' object has no attribute 'deepspeed' HOT 3
- Incorrect Doc String for `SFTConfig` HOT 2
- plz make GPOTrainer! (Generalized Preference Optimization)
- AttributeError: 'PPOv2Trainer' object has no attribute 'deepspeed' HOT 3
- ConstantLengthDataset should shuffle the order of samples before packing HOT 2
- The ppo_trainer.generate() call results in an error. HOT 2
- DPOTrainer failed on training Custom Mixture of Experts model with config output_router_logits=True
- Disable the dropout by default in Online DPO HOT 1
- RLOOTrainer & PPOv2Trainer - Modify Name for W&B Logged Table HOT 3
- PPOv2Trainer & RLOOTrainer - Add Safety Check that `policy` object != `ref_policy` object HOT 2
- Always allow `ref_model=None` HOT 5
- Reward model HOT 4
- [DPOTrainer] Tokenizer calculation fail during Q+A concat HOT 3
- Model does not generate eos when SFTTrainer with setup_chat_format is used HOT 1
- Deepspeed Zero2 not working when using DPOTrainer HOT 3
- Discrepancy in LLaMA 3.1 performance when using custom trainer and SFTTrainer HOT 7
- why does the ppo calculate advantage reversed the index?
- [Tracking issue] General dataset support
- `GKDTrainer`
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 trl.