huggingface/optimum-habana
Easy and lightning fast training of 🤗 Transformers on Habana Gaudi processor (HPU)
PythonApache-2.0
Issues
- 7
- 3
[Bug] fp8 measure failed on Llama-2-7b-hf
#890 opened by ClarkChin08 - 2
Update to transformers 4.39.1+
#865 opened by vidyasiv - 6
- 2
After accelerate v0.28.0, the `split_batches` of class Accelerator became a unsettable property method
#872 opened by eraser00 - 6
Bug in Optimum Habana Examples
#741 opened by allenwsh82 - 2
Upgrade DPO and PPO to support trl v0.7.8 or higher
#784 opened by skavulya - 3
- 9
LlamaForCausalLM.forward() got an unexpected keyword argument 'use_flash_attention'
#760 opened by dittops - 12
Error while saving checkpoint with trainer
#744 opened by dittops - 3
Discussion on attention mask handling
#692 opened by vidyasiv - 3
- 2
ENABLE_CONSOLE disrupts version checks in examples
#684 opened by vidyasiv - 2
multi-node distributed training on Ray is failed.
#505 opened by yuanwu2017 - 15
Device Aquire failed
#367 opened by DannyAtal - 1
- 1
run_lora_clm.py support for other datasets
#629 opened by tmabraham - 7
Text Generation runtime error
#614 opened by danielfleischer - 7
Error: Getting size for given data type is not supported while fine tuning starcoder model on optimum-habana
#350 opened by anindya-saha - 5
Default value of ignore_eos
#455 opened by bhargaveede - 2
- 8
text_generation_launcher: Waiting for shard to be ready... rank=1 forever if we pass --num-shard
#376 opened by avinashkarani - 6
- 3
StableDiffusion v2.1 produces incorrect images
#547 opened by mgawarkiewicz - 2
Will Hugging Face support GLM series models (ChatGLM-6B, ChatGLM2-6B ...) in Transformers?
#299 opened by jychen-habana - 1
- 4
FileNotFoundError: Couldn't find a dataset script
#432 opened by OmerBoucris - 4
- 3
Add support for max_length in run_generation
#472 opened by ankurneog - 1
Adaptive output and contextual dialogue capabilities of text-generation-inference
#424 opened by MLikeWater - 1
- 1
- 3
Performance is better in 1.6.1 release compared to 1.7.4 release in many models
#419 opened by vineethanandh - 2
Beach search transformers test cases are failing with KeyError: 'limit_hpu_graphs'
#445 opened by ankurneog - 7
- 4
Does it make sense to also provide an option of max input tokens for text generation ?
#423 opened by puneeshkhanna - 9
- 2
accelerate llama inference in TGI
#339 opened by sywangyi - 1
- 11
dock build fail in https://github.com/huggingface/optimum-habana/tree/main/text-generation-inference
#337 opened by sywangyi - 1
- 3
Where in the directory "/tmp/tst-summarization", is the summarization output stored?
#292 opened by Abhaycnvrg - 9
RuntimeError: Device acquire failed. in /usr/local/lib/python3.8/dist-packages/habana_frameworks/torch/hpu/__init__.py"
#288 opened by Abhaycnvrg - 10
return super().__torch_function__(func, types, new_args, kwargs) RuntimeError: [Rank:0] FATAL ERROR :: MODULE:PT_DEVMEM Allocation failed for size::3288334336 (3136)MB
#289 opened by Abhaycnvrg - 3
Enable beam_search for text-generation
#270 opened by ZhaiFeiyue - 9
- 4
Adding profiling
#245 opened by ZhaiFeiyue - 2
Can not run text-generation with bloom deepspeed?
#240 opened by ZhaiFeiyue - 3
When loading datasets by HuggingFace datasets.load_dataset like cifar10, could it be possible to return the dataset without decoding automatically.
#236 opened by jychen-habana - 10
GPT2 support bf16 for both training and inferecne
#230 opened by ZhaiFeiyue