-
examples=(Fire and ice dragon or lighting and ice or fire and lightning breed dragons)
-
### 🐛 Describe the bug
when use the newest version 0.4.1，when use dpo_online to finetune mistral0.3-7b model, it has a bug: AttributeError: 'MistralRMSNorm' object has no attribute 'in_place'.
ea…
-
### 🐛 Describe the bug
https://github.com/linkedin/Liger-Kernel/pull/354
1. revert patching causes flce not taking effect (comment out revert patching for now, and only test float32)
2. qwen2 vl …
-
I'm assuming it only works on Ampere, Hopper, Lovelace. Is that correct? It might be nice to specify in the readme, if it is limited to certain GPU types.
-
### System Info
WARNING:root:A model is loaded from './saved_models/fp-meta-llama3', and no v_head weight is found.
### Information
- [ ] The official example scripts
- [X] My own modified scripts…
-
Liger (Linkedin GPU Efficient Runtime) Kernel is a collection of Triton kernels designed specifically for LLM training. It can effectively increase multi-GPU training throughput by 20% and reduces mem…
-
**Description**
The output of the "show platform syseeprom" is not same with the outout get with the eeprom.py.
In test_show_platform.py:test_show_platform_syseeprom, when compare the outputs…
-
### 🐛 Describe the bug
To fix GA bug, latest transformers no longer imports CrossEntropyLoss. Instead, it's wrapped in [`self.loss_function`](https://github.com/huggingface/transformers/blob/a06a0d12…
-
I am using the STREAM python package, and I am having trouble running the detect_transition_genes() function on my data. I can process everything just fine up to the streamplots and the subwaymaps wit…
-
### 🐛 Describe the bug
logits are not calculated here:
https://github.com/linkedin/Liger-Kernel/blob/6ab3b9febc29f5045e6d2e27ba6bacaa4f041d91/src/liger_kernel/transformers/model/qwen2.py#L99-L110
…