v1.20.1: Patch release

• Fix lora unscaling in diffusion pipelines by @eaidova in #937
• Fix compatibility with diffusers < 0.25.0 by @eaidova in #952
• Allow to use SDPA in clip models by @eaidova in #941
• Updated OVPipelinePart to have separate ov_config by @e-ddykim in #957
• Symbol use in optimum: fix misprint by @jane-intel in #948
• Fix temporary directory saving by @eaidova in #959
• Disable warning about tokenizers version for ov tokenizers >= 2024.5 by @eaidova in #962
• Restore original model_index.json after save_pretrained call by @eaidova in #961
• Add v4.46 transformers support by @echarlaix in #960

v1.20.0: multi-modal and OpenCLIP models support, transformers v4.45

OpenVINO

Multi-modal models support

Adding OVModelForVisionCausalLM by @eaidova in #883

OpenCLIP models support

Adding OpenCLIP models support by @sbalandi in #857

from optimum.intel import OVModelCLIPVisual, OVModelCLIPText

visual_model = OVModelCLIPVisual.from_pretrained(model_name_or_path)
text_model  = OVModelCLIPText.from_pretrained(model_name_or_path)
image = processor(image).unsqueeze(0)
text = tokenizer(["a diagram", "a dog", "a cat"])
image_features = visual_model(image).image_features
text_features = text_model(text).text_features

Diffusion pipeline

Adding OVDiffusionPipeline to simplify diffusers model loading by @IlyasMoutawwakil in #889

  model_id = "stabilityai/stable-diffusion-xl-base-1.0"
- pipeline = OVStableDiffusionXLPipeline.from_pretrained(model_id)
+ pipeline = OVDiffusionPipeline.from_pretrained(model_id)
  image = pipeline("sailing ship in storm by Leonardo da Vinci").images[0]

NNCF GPTQ support

GPTQ support by @nikita-savelyevv in #912

Transformers v4.45

Transformers v4.45 support by @echarlaix in #902

Subfolder

Remove the restriction for the model's config to be in the model's subfolder by @tomaarsen in #933

New Contributors

• @jane-intel made their first contribution in #696
• @andreyanufr made their first contribution in #903
• @MaximProshin made their first contribution in #905
• @tomaarsen made their first contribution in #931

v1.19.0: SentenceTransformers OpenVINO support

• Support SentenceTransformers models inference by @aleksandr-mokrov in #865

from optimum.intel import OVSentenceTransformer

model_id = "sentence-transformers/all-mpnet-base-v2"
model = OVSentenceTransformer.from_pretrained(model_id, export=True)
sentences = ["This is an example sentence", "Each sentence is converted"]
embeddings = model.encode(sentences)

• Infer if the model needs to be exported or not by @echarlaix in #825

  from optimum.intel import OVModelForCausalLM

- model = OVModelForCausalLM.from_pretrained("gpt2", export=True)
+ model = OVModelForCausalLM.from_pretrained("gpt2")

Compatible with transformers>=4.36,<=4.44

Full Changelog: v1.18.0...v1.19.0

v1.18.3: Patch release

Full Changelog: v1.18.2...v1.18.3

v1.18.2: Patch release

• Fix model patching for internlm2 by @eaidova in #814
• Fix loading models from cache by @eaidova in #820
• Disable tpp for un-verified models by @jiqing-feng in #822
• Update default NNCF configurationsby @KodiaqQ in #824
• Fix update causal mask for transformers 4.42 by @eaidova in #852
• Fix bf16 inference accuracy for mistral, phi3, dbrx by @eaidova in #833
• Revert rotary embedding patching for recovering gpu accuracy by @eaidova in #855
• Support transformers 4.43 by @IlyasMoutawwakil in #856

Full Changelog: v1.18.1...v1.18.2

v1.18.1: Patch release

• OV configurations alignment by @KodiaqQ in #787
• Enable transformers v4.42.0 by @echarlaix in #793
• Deprecate onnx/ort model export and quantization by @IlyasMoutawwakil in #795
• Free memory after model export by @eaidova in #800
• Update config import path for neural-compressor v2.6 by @changwangss in #801
• Pin library name to transformers for feature extraction by @IlyasMoutawwakil in #804

Full Changelog: v1.18.0...v1.18.1

v1.18.0: Arctic, Jais, OpenVINO pipelines

OpenVINO

• Enable Arctic, Jais export by @eaidova in #726
• Enable GLM-4 export by @eaidova in #776
• Move data-driven quantization after model export for text-generation models by @nikita-savelyevv in #721
• Create default token_type_ids when needed for inference by @echarlaix #757
• Resolve default int4 config for local models by @eaidova in #760
• Update to NNCF 2.11 by @nikita-savelyevv in #763
• Fix quantization config by @echarlaix in #773
• Expose trust remote code argument when generating calibration dataset for datasets >= v2.20.0 by @echarlaix #767
• Add pipelines by @echarlaix in #740

from optimum.intel.pipelines import pipeline

# Load openvino model
ov_pipe = pipeline("text-generation", "helenai/gpt2-ov", accelerator="openvino")
# Load pytorch model and convert it to openvino before inference
pipe = pipeline("text-generation", "gpt2", accelerator="openvino")

IPEX

• Enable IPEX patching for llama for >= v2.3 by @jiqing-feng in #725
• Refactor llama modeling for IPEX patching by @faaany in #728
• Refactor model loading by @jiqing-feng in #752

v1.17.2: Patch release

• Fix compatibility with transformers < v4.39.0 by @echarlaix in #754

v1.17.1: Patch release

• Add setuptools to fix issue with Python 3.12 by @helena-intel in #747
• Disable warnings by @helena-intel in #748
• Fix Windows TemporaryDirectory issue by @helena-intel in #749
• Fix generation config loading and saving by @eaidova in #750

v1.17.0: ITREX WOQ, IPEX pipeline, extended OpenVINO export

OpenVINO

• Enable Orion, InternLM2 export by @eaidova in #628

• Enable OLMo export by @eaidova in #678

• Enable Phi3 export by @eaidova in #686

• Enable BioGPT, Cohere, Persimmon, XGLM export by @eaidova in #709

• Enable Aquila, InternLM, XVERSE export by @eaidova in #716

• Add OVModelForVision2Seq class by @eaidova in #634

  from optimum.intel import OVModelForVision2Seq
  
  model = OVModelForVision2Seq.from_pretrained("nlpconnect/vit-gpt2-image-captioning", export=True)
  gen_tokens = model.generate(**inputs)

• Introduce OVQuantizationConfig for NNCF quantization by @nikita-savelyevv in #638

• Enable hybrid StableDiffusion models export via optimum-cli by @l-bat in #618

  optimum-cli export openvino --model SimianLuo/LCM_Dreamshaper_v7 --task latent-consistency  --dataset conceptual_captions --weight-format int8 <output_dir>
  

• Convert Tokenizers by default by @apaniukov in #580

• Custom tasks modeling by @IlyasMoutawwakil in #669

• Add dynamic quantization config by @echarlaix in #661

  from optimum.intel import OVModelForCausalLM, OVDynamicQuantizationConfig
  
  model_id = "meta-llama/Meta-Llama-3-8B"
  q_config = OVDynamicQuantizationConfig(bits=8, activations_group_size=32)
  model = OVModelForCausalLM.from_pretrained(model_id, export=True, quantization_config=q_config)

• Transition to a newer NNCF API for PyTorch model quantization by @nikita-savelyevv in #630

ITREX

• Add ITREX weight-only quantization support by @PenghuiCheng in #455

IPEX

• Add IPEX pipeline by @jiqing-feng in #501