huggingface alignment-handbook issues

huggingface / alignment-handbook

Robust recipes to align language models with human and AI preferences

https://huggingface.co/HuggingFaceH4

Apache License 2.0

4.2k stars 357 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

SFTConfig type wrong

#178 the-utkarshjain opened 1 day ago
1
oliver.huneck.info@gmail.com

#177 ratterdull78 closed 1 week ago
0
https://github.com/huggingface/alignment-handbook

#176 ratterdull78 closed 1 week ago
1
TRL/Alignment-Handbook torch.dtype Issues

#175 neelsjain opened 2 weeks ago
1
Question about torch_dtype when runnging run_orpo.py

#174 sylee96 opened 2 weeks ago
1
Wrong exception handling when loading dataset from local disk

#173 ganler opened 1 month ago
2
Question on "mlm" in continued pre-training

#172 tanliboy opened 1 month ago
2
Unexpected behavior in apply_chat_template function adding repeated assistant turns

#171 iseesaw closed 1 month ago
1
Question about sft with deepspeed

#170 XXares opened 1 month ago
1
Cannot flatten integer dtype tensors

#169 jaywongs opened 1 month ago
1
Released model weights for ablations of KTO/IPO/DPO cannot be found

#168 ChenDRAG opened 1 month ago
0
[ORPO] system special token is included in chosen/rejected samples after applying chat template

#167 blakechi closed 1 month ago
1
CI failing due to `mistralai/Mistral-7B-Instruct-v0.2` being gated now

#166 alvarobartt opened 2 months ago
0
Add `scripts/run_kto.py`

#165 alvarobartt opened 2 months ago
1
Add Zephyr TinyLlama training recipe

#164 Ritvik19 closed 2 months ago
0
Add recipe for zephyr-tinyllama-sft

#163 Ritvik19 closed 2 months ago
0
Issue Running `run_sft.py` After Configuration Changes in GMAL Folder : (ChildFailedError)

#162 alielfilali01 closed 2 months ago
3
Add 'do_train' check to cpt

#161 BramVanroy opened 2 months ago
1
Add fsdp+qlora support

#160 deep-diver closed 2 months ago
4
FSDP + QDoRA Support

#159 iseesaw opened 2 months ago
6
How to work with local data

#158 pretidav opened 2 months ago
1
Clarification on dataset mixer

#157 deep-diver opened 2 months ago
5
Sync with alignment handbook

#156 scottsuk0306 closed 2 months ago
1
Dependency updates for QLoRA+FSDP

#155 deep-diver opened 2 months ago
0
Add ORPO within `README.md` files

#154 alvarobartt closed 2 months ago
1
Different dtype while saving optimizer with FSDP

#153 heraclex12 closed 2 months ago
2
Update README.md, there is problem if it have to be flash_attn==2.3.6

#152 ZizhengYang closed 2 months ago
1
experiments with llama1

#151 sfc-gh-rsamdani closed 3 months ago
0
Method to disable evaluation

#150 ZhiCheng0326 closed 3 months ago
0
CPT training is giving pretty unstalbe results with the learning rate 2e-5.

#149 shamanez opened 3 months ago
1
Cannot reproduce zephyr-7b-gemma-v0.1

#148 jasonyux closed 3 months ago
3
Multi-GPU Training with DPO Full Parameter Stucks

#147 Taishi-N324 opened 3 months ago
0
Fix an issue with philschmid/gemma-tokenizer-chatml tokenizer in sft

#146 kykim0 closed 3 months ago
1
Constitutional AI models do not achieve MT-Bench scores as reported

#145 JingtongSu opened 3 months ago
0
Can we please add the option to work with a tokenized dataset, escpailly for the CPT task.

#144 shamanez opened 3 months ago
0
Add `run_orpo.py`

#143 alvarobartt closed 3 months ago
3
Efficient dialog data format for KTO training

#142 DavidFarago opened 3 months ago
0
Can any one share the script what params should be passed to run_dpo.py

#141 Oscarjia opened 3 months ago
1
fix trust_remote_code for tokenizer in model_utils.py

#140 csarron closed 3 months ago
1
fix: Zephyr LoRA fine-tuning fixed

#139 Serega6678 closed 3 months ago
1
How to select parts to bp in sft

#138 Fu-Dayuan opened 3 months ago
0
Fix dataloading for cpt

#137 BramVanroy closed 3 months ago
1
Missing config_qlora.yaml

#136 IsraelAbebe closed 3 months ago
2
🌟

#135 lewtun closed 4 months ago
2
Is there a way to freeze some layers of a model ?

#134 shamanez opened 4 months ago
0
Early Stopping Issue when used with ConstantLengthDataset

#133 sankydesai opened 4 months ago
0
Not able to run Zephyr 7B Gemma with 4 80GB A100s

#132 TJ-Solergibert opened 4 months ago
1
Adding continued_pretraining task

#131 BramVanroy closed 3 months ago
3
Downloading latest CUDA version (11.6 or above) for MacOS to use FlashAttention

#130 shubhamcs162 opened 4 months ago
0
🪁

#129 lewtun closed 4 months ago
1