mirror of
https://github.com/saymrwulf/onnxruntime.git
synced 2026-06-08 00:23:03 +00:00
914bc409b0
Bumps [transformers](https://github.com/huggingface/transformers) from 4.30.0 to 4.36.0. <details> <summary>Release notes</summary> <p><em>Sourced from <a href="https://github.com/huggingface/transformers/releases">transformers's releases</a>.</em></p> <blockquote> <h2>v4.36: Mixtral, Llava/BakLlava, SeamlessM4T v2, AMD ROCm, F.sdpa wide-spread support</h2> <h2>New model additions</h2> <h3>Mixtral</h3> <p>Mixtral is the new open-source model from Mistral AI announced by the blogpost <a href="https://mistral.ai/news/mixtral-of-experts/">Mixtral of Experts</a>. The model has been proven to have comparable capabilities to Chat-GPT according to the benchmark results shared on the release blogpost.</p> <!-- raw HTML omitted --> <p>The architecture is a sparse Mixture of Experts with Top-2 routing strategy, similar as <code>NllbMoe</code> architecture in transformers. You can use it through <code>AutoModelForCausalLM</code> interface:</p> <pre lang="py"><code>>>> import torch >>> from transformers import AutoModelForCausalLM, AutoTokenizer <p>>>> model = AutoModelForCausalLM.from_pretrained("mistralai/Mixtral-8x7B", torch_dtype=torch.float16, device_map="auto") >>> tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-8x7B")</p> <p>>>> prompt = "My favourite condiment is"</p> <p>>>> model_inputs = tokenizer([prompt], return_tensors="pt").to(device) >>> model.to(device)</p> <p>>>> generated_ids = model.generate(**model_inputs, max_new_tokens=100, do_sample=True) >>> tokenizer.batch_decode(generated_ids)[0] </code></pre></p> <p>The model is compatible with existing optimisation tools such Flash Attention 2, <code>bitsandbytes</code> and PEFT library. The checkpoints are release under <a href="https://huggingface.co/mistralai"><code>mistralai</code></a> organisation on the Hugging Face Hub.</p> <h3>Llava / BakLlava</h3> <p>Llava is an open-source chatbot trained by fine-tuning LlamA/Vicuna on GPT-generated multimodal instruction-following data. It is an auto-regressive language model, based on the transformer architecture. In other words, it is an multi-modal version of LLMs fine-tuned for chat / instructions.</p> <!-- raw HTML omitted --> <p>The Llava model was proposed in <a href="https://arxiv.org/pdf/2310.03744">Improved Baselines with Visual Instruction Tuning</a> by Haotian Liu, Chunyuan Li, Yuheng Li and Yong Jae Lee.</p> <ul> <li>[<code>Llava</code>] Add Llava to transformers by <a href="https://github.com/younesbelkada"><code>@younesbelkada</code></a> in <a href="https://redirect.github.com/huggingface/transformers/issues/27662">#27662</a></li> <li>[LLaVa] Some improvements by <a href="https://github.com/NielsRogge"><code>@NielsRogge</code></a> in <a href="https://redirect.github.com/huggingface/transformers/issues/27895">#27895</a></li> </ul> <p>The integration also includes <a href="https://github.com/SkunkworksAI/BakLLaVA"><code>BakLlava</code></a> which is a Llava model trained with Mistral backbone.</p> <p>The mode is compatible with <code>"image-to-text"</code> pipeline:</p> <pre lang="py"><code>from transformers import pipeline from PIL import Image import requests <p>model_id = "llava-hf/llava-1.5-7b-hf" </tr></table> </code></pre></p> </blockquote> <p>... (truncated)</p> </details> <details> <summary>Commits</summary> <ul> <li><a href=" |
||
|---|---|---|
| .. | ||
| github | ||
| __init__.py | ||
| amd_hipify.py | ||
| build.py | ||
| clean_docker_image_cache.py | ||
| compile_triton.py | ||
| coverage.py | ||
| gen_def.py | ||
| get_docker_image.py | ||
| logger.py | ||
| op_registration_utils.py | ||
| op_registration_validator.py | ||
| patch_manylinux.py | ||
| policheck_exclusions.xml | ||
| reduce_op_kernels.py | ||
| replace_urls_in_deps.py | ||
| requirements.txt | ||
| set-trigger-rules.py | ||
| update_tsaoptions.py | ||
| upload_python_package_to_azure_storage.py | ||