[OV] Update optimization docs page with information about VLMs (#1007)

nikita-savelyevv · AlexKoff88 · web-flow · commit 0a9a10be3e5b · 2024-11-26T12:53:12.000+04:00
* Update optimization page

* Update cli usage

* Add more export options to export page

* Update

* Adopt for narrower text fields

* Update docs/source/openvino/export.mdx

* Update docs/source/openvino/export.mdx

Co-authored-by: Alexander Kozlov &lt;alexander.kozlov@intel.com&gt;

* Update docs/source/openvino/optimization.mdx

Co-authored-by: Alexander Kozlov &lt;alexander.kozlov@intel.com&gt;

* Update docs/source/openvino/export.mdx

Co-authored-by: Alexander Kozlov &lt;alexander.kozlov@intel.com&gt;

* Update docs/source/openvino/export.mdx

Co-authored-by: Alexander Kozlov &lt;alexander.kozlov@intel.com&gt;

* Update docs/source/openvino/export.mdx

Co-authored-by: Alexander Kozlov &lt;alexander.kozlov@intel.com&gt;

---------

Co-authored-by: Alexander Kozlov &lt;alexander.kozlov@intel.com&gt;
diff --git a/docs/source/openvino/export.mdx b/docs/source/openvino/export.mdx
@@ -30,70 +30,108 @@ optimum-cli export openvino --model local_llama --task text-generation-with-past
 Check out the help for more options:
 
 ```bash
-optimum-cli export openvino --help
-
-usage: optimum-cli export openvino [-h] -m MODEL [--task TASK] [--framework {pt,tf}] [--trust-remote-code] [--weight-format {fp32,fp16,int8,int4}]
-                                   [--library {transformers,diffusers,timm,sentence_transformers}] [--cache_dir CACHE_DIR] [--pad-token-id PAD_TOKEN_ID] [--ratio RATIO] [--sym]
-                                   [--group-size GROUP_SIZE] [--dataset DATASET] [--all-layers] [--awq] [--scale-estimation] [--sensitivity-metric SENSITIVITY_METRIC] [--num-samples NUM_SAMPLES]
-                                   [--disable-stateful] [--disable-convert-tokenizer]
+usage: optimum-cli export openvino [-h] -m MODEL [--task TASK] [--framework {pt,tf}] [--trust-remote-code]
+                                   [--weight-format {fp32,fp16,int8,int4,mxfp4,nf4}]
+                                   [--library {transformers,diffusers,timm,sentence_transformers,open_clip}]
+                                   [--cache_dir CACHE_DIR] [--pad-token-id PAD_TOKEN_ID] [--ratio RATIO] [--sym]
+                                   [--group-size GROUP_SIZE] [--dataset DATASET] [--all-layers] [--awq]
+                                   [--scale-estimation] [--gptq] [--sensitivity-metric SENSITIVITY_METRIC]
+                                   [--num-samples NUM_SAMPLES] [--disable-stateful] [--disable-convert-tokenizer]
                                    output
 
 optional arguments:
   -h, --help            show this help message and exit
 
 Required arguments:
-  --model MODEL         Model ID on huggingface.co or path on disk to load model from.
-
+  -m MODEL, --model MODEL
+                        Model ID on huggingface.co or path on disk to load model from.
   output                Path indicating the directory where to store the generated OV model.
 
 Optional arguments:
-  --task TASK           The task to export the model for. If not specified, the task will be auto-inferred based on the model. Available tasks depend on the model, but are among: ['image-segmentation',
-                        'feature-extraction', 'mask-generation', 'audio-classification', 'conversational', 'stable-diffusion-xl', 'question-answering', 'sentence-similarity', 'text2text-generation',
-                        'masked-im', 'automatic-speech-recognition', 'fill-mask', 'image-to-text', 'text-generation', 'zero-shot-object-detection', 'multiple-choice', 'object-detection', 'stable-
-                        diffusion', 'audio-xvector', 'text-to-audio', 'zero-shot-image-classification', 'token-classification', 'image-classification', 'depth-estimation', 'image-to-image', 'audio-
-                        frame-classification', 'semantic-segmentation', 'text-classification']. For decoder models, use `xxx-with-past` to export the model using past key values in the decoder.
-  --framework {pt,tf}   The framework to use for the export. If not provided, will attempt to use the local checkpoints original framework or what is available in the environment.
-  --trust-remote-code   Allows to use custom code for the modeling hosted in the model repository. This option should only be set for repositories you trust and in which you have read the code, as it
-                        will execute on your local machine arbitrary code present in the model repository.
-  --weight-format {fp32,fp16,int8,int4}
+  --task TASK           The task to export the model for. If not specified, the task will be auto-inferred based on
+                        the model. Available tasks depend on the model, but are among: ['fill-mask', 'masked-im',
+                        'audio-classification', 'automatic-speech-recognition', 'text-to-audio', 'image-text-to-text',
+                        'depth-estimation', 'image-to-image', 'text-generation', 'text-to-image', 'mask-generation',
+                        'audio-frame-classification', 'sentence-similarity', 'image-classification', 'multiple-
+                        choice', 'text-classification', 'text2text-generation', 'token-classification', 'feature-
+                        extraction', 'zero-shot-image-classification', 'zero-shot-object-detection', 'object-
+                        detection', 'inpainting', 'question-answering', 'semantic-segmentation', 'image-segmentation',
+                        'audio-xvector', 'image-to-text']. For decoder models, use `xxx-with-past` to export the model
+                        using past key values in the decoder.
+  --framework {pt,tf}   The framework to use for the export. If not provided, will attempt to use the local
+                        checkpoint's original framework or what is available in the environment.
+  --trust-remote-code   Allows to use custom code for the modeling hosted in the model repository. This option should
+                        only be set for repositories you trust and in which you have read the code, as it will execute
+                        on your local machine arbitrary code present in the model repository.
+  --weight-format {fp32,fp16,int8,int4,mxfp4,nf4}
                         The weight format of the exported model.
-  --library {transformers,diffusers,timm,sentence_transformers}
-                        The library used to load the model before export. If not provided, will attempt to infer the local checkpoints library.
+  --library {transformers,diffusers,timm,sentence_transformers,open_clip}
+                        The library used to load the model before export. If not provided, will attempt to infer the
+                        local checkpoint's library
   --cache_dir CACHE_DIR
-                        The path to a directory in which the downloaded model should be cached if the standard cache should not be used.
+                        The path to a directory in which the downloaded model should be cached if the standard cache
+                        should not be used.
   --pad-token-id PAD_TOKEN_ID
-                        This is needed by some models, for some tasks. If not provided, will attempt to use the tokenizer to guess it.
-  --ratio RATIO         A parameter used when applying 4-bit quantization to control the ratio between 4-bit and 8-bit quantization. If set to 0.8, 80% of the layers will be quantized to int4 while
-                        20% will be quantized to int8. This helps to achieve better accuracy at the sacrifice of the model size and inference latency. Default value is 1.0.
+                        This is needed by some models, for some tasks. If not provided, will attempt to use the
+                        tokenizer to guess it.
+  --ratio RATIO         A parameter used when applying 4-bit quantization to control the ratio between 4-bit and 8-bit
+                        quantization. If set to 0.8, 80% of the layers will be quantized to int4 while 20% will be
+                        quantized to int8. This helps to achieve better accuracy at the sacrifice of the model size
+                        and inference latency. Default value is 1.0.
   --sym                 Whether to apply symmetric quantization
   --group-size GROUP_SIZE
-                        The group size to use for int4 quantization. Recommended value is 128 and -1 will results in per-column quantization.
-  --dataset DATASET     The dataset used for data-aware compression or quantization with NNCF. You can use the one from the list ['wikitext2','c4','c4-new'] for language models or
-                        ['conceptual_captions','laion/220k-GPT4Vision-captions-from-LIVIS','laion/filtered-wit'] for diffusion models.
-  --all-layers          Whether embeddings and last MatMul layers should be compressed to INT4. If not provided an weight compression is applied, they are compressed to INT8.
-  --awq                 Whether to apply AWQ algorithm. AWQ improves generation quality of INT4-compressed LLMs, but requires additional time for tuning weights on a calibration dataset. To run AWQ,
-                        please also provide a dataset argument. Note: it is possible that there will be no matching patterns in the model to apply AWQ, in such case it will be skipped.
-  --scale-estimation    Indicates whether to apply a scale estimation algorithm that minimizes the L2 error between the original and compressed layers. Providing a dataset is required to run scale
-                        estimation. Please note, that applying scale estimation takes additional memory and time.
+                        The group size to use for quantization. Recommended value is 128 and -1 uses per-column
+                        quantization.
+  --dataset DATASET     The dataset used for data-aware compression or quantization with NNCF. You can use the one
+                        from the list ['wikitext2','c4','c4-new'] for language models or
+                        ['conceptual_captions','laion/220k-GPT4Vision-captions-from-LIVIS','laion/filtered-wit'] for
+                        diffusion models.
+  --all-layers          Whether embeddings and last MatMul layers should be compressed to INT4. If not provided an
+                        weight compression is applied, they are compressed to INT8.
+  --awq                 Whether to apply AWQ algorithm. AWQ improves generation quality of INT4-compressed LLMs, but
+                        requires additional time for tuning weights on a calibration dataset. To run AWQ, please also
+                        provide a dataset argument. Note: it is possible that there will be no matching patterns in the
+                        model to apply AWQ, in such case it will be skipped.
+  --scale-estimation    Indicates whether to apply a scale estimation algorithm that minimizes the L2 error between
+                        the original and compressed layers. Providing a dataset is required to run scale estimation.
+                        Please note, that applying scale estimation takes additional memory and time.
+  --gptq                Indicates whether to apply GPTQ algorithm that optimizes compressed weights in a layer-wise
+                        fashion to minimize the difference between activations of a compressed and original layer.
+                        Please note, that applying GPTQ takes additional memory and time.
   --sensitivity-metric SENSITIVITY_METRIC
-                        The sensitivity metric for assigning quantization precision to layers. Can be one of the following: ['weight_quantization_error', 'hessian_input_activation',
+                        The sensitivity metric for assigning quantization precision to layers. It can be one of the
+                        following: ['weight_quantization_error', 'hessian_input_activation',
                         'mean_activation_variance', 'max_activation_variance', 'mean_activation_magnitude'].
   --num-samples NUM_SAMPLES
                         The maximum number of samples to take from the dataset for quantization.
-  --disable-stateful    Disable stateful converted models, stateless models will be generated instead. Stateful models are produced by default when this key is not used. In stateful models all kv-cache
-                        inputs and outputs are hidden in the model and are not exposed as model inputs and outputs. If --disable-stateful option is used, it may result in sub-optimal inference
-                        performance. Use it when you intentionally want to use a stateless model, for example, to be compatible with existing OpenVINO native inference code that expects kv-cache inputs
-                        and outputs in the model.
+  --disable-stateful    Disable stateful converted models, stateless models will be generated instead. Stateful models
+                        are produced by default when this key is not used. In stateful models all kv-cache inputs and
+                        outputs are hidden in the model and are not exposed as model inputs and outputs. If --disable-
+                        stateful option is used, it may result in sub-optimal inference performance. Use it when you
+                        intentionally want to use a stateless model, for example, to be compatible with existing
+                        OpenVINO native inference code that expects KV-cache inputs and outputs in the model.
   --disable-convert-tokenizer
                         Do not add converted tokenizer and detokenizer OpenVINO models.
 ```
 
-You can also apply fp16, 8-bit or 4-bit weight-only quantization on the Linear, Convolutional and Embedding layers when exporting your model by setting `--weight-format` to respectively `fp16`, `int8` or `int4`:
+You can also apply fp16, 8-bit or 4-bit weight-only quantization on the Linear, Convolutional and Embedding layers when exporting your model by setting `--weight-format` to respectively `fp16`, `int8` or `int4`.
 
+Export with INT8 weights compression:
 ```bash
 optimum-cli export openvino --model meta-llama/Meta-Llama-3-8B --weight-format int8 ov_model/
 ```
 
+Export with INT4 weights compression:
+```bash
+optimum-cli export openvino --model meta-llama/Meta-Llama-3-8B --weight-format int4 ov_model/
+```
+
+Export with INT4 weights compression and a data-aware AWQ and Scale Estimation algorithms:
+```bash
+optimum-cli export openvino --model meta-llama/Meta-Llama-3-8B \
+    --weight-format int4 --awq --scale-estimation --dataset wikitext2 ov_model/
+```
+
 For more information on the quantization parameters checkout the [documentation](inference#weight-only-quantization)
 
 
@@ -130,6 +168,14 @@ To export your Stable Diffusion XL model to the OpenVINO IR format with the CLI
 optimum-cli export openvino --model stabilityai/stable-diffusion-xl-base-1.0 ov_sdxl/
 ```
 
+You can also apply hybrid quantization during model export. For example:
+```bash
+optimum-cli export openvino --model stabilityai/stable-diffusion-xl-base-1.0 \
+    --weight-format int8 --dataset conceptual_captions ov_sdxl/
+```
+
+For more information about hybrid quantization, take a look at this jupyter [notebook](https://github.com/huggingface/optimum-intel/blob/main/notebooks/openvino/stable_diffusion_hybrid_quantization.ipynb).
+
 ## When loading your model
 
 You can also load your PyTorch checkpoint and convert it to the OpenVINO format on-the-fly, by setting `export=True` when loading your model.
diff --git a/docs/source/openvino/optimization.mdx b/docs/source/openvino/optimization.mdx
@@ -30,29 +30,37 @@ Quantization can be applied on the model's Linear, Convolutional and Embedding l
 
 #### 8-bit
 
-For the 8-bit weight quantization you can set `load_in_8bit=True` to load your model's weights in 8-bit:
+For the 8-bit weight quantization you can provide `quantization_config` equal to `OVWeightQuantizationConfig(bits=8)` to load your model's weights in 8-bit:
 
 ```python
-from optimum.intel import OVModelForCausalLM
+from optimum.intel import OVModelForCausalLM, OVWeightQuantizationConfig
 
 model_id = "helenai/gpt2-ov"
-model = OVModelForCausalLM.from_pretrained(model_id, load_in_8bit=True)
+quantization_config = OVWeightQuantizationConfig(bits=8)
+model = OVModelForCausalLM.from_pretrained(model_id, quantization_config=quantization_config)
 
 # Saves the int8 model that will be x4 smaller than its fp32 counterpart
 model.save_pretrained(saving_directory)
 ```
 
+Weights of language models inside vision-language pipelines can be quantized in a similar way:
+```python
+model = OVModelForVisualCausalLM.from_pretrained(
+    "llava-hf/llava-v1.6-mistral-7b-hf",
+    quantization_config=quantization_config
+)
+```
+
 <Tip warning={true}>
 
-If not specified, `load_in_8bit` will be set to `True` by default when models larger than 1 billion parameters are exported to the OpenVINO format (with `export=True`). You can disable it with `load_in_8bit=False`.
+If quantization_config is not provided, model will be exported in 8 bits by default when it has more than 1 billion parameters. You can disable it with `load_in_8bit=False`.
 
 </Tip>
 
-You can also provide a `quantization_config` instead to specify additional optimization parameters.
 
 #### 4-bit
 
-For the 4-bit weight quantization, you need a `quantization_config` to define the optimization parameters, for example:
+4-bit weight quantization can be achieved in a similar way:
 
 ```python
 from optimum.intel import OVModelForCausalLM, OVWeightQuantizationConfig
@@ -61,10 +69,24 @@ quantization_config = OVWeightQuantizationConfig(bits=4)
 model = OVModelForCausalLM.from_pretrained(model_id, quantization_config=quantization_config)
 ```
 
+Or for vision-language pipelines:
+```python
+model = OVModelForVisualCausalLM.from_pretrained(
+    "llava-hf/llava-v1.6-mistral-7b-hf",
+    quantization_config=quantization_config
+)
+```
+
 You can tune quantization parameters to achieve a better performance accuracy trade-off as follows:
 
 ```python
-quantization_config = OVWeightQuantizationConfig(bits=4, sym=False, ratio=0.8, dataset="wikitext2")
+quantization_config = OVWeightQuantizationConfig(
+    bits=4,
+    sym=False,
+    ratio=0.8,
+    quant_method="awq",
+    dataset="wikitext2"
+)
 ```
 
 By default the quantization scheme will be [asymmetric](https://github.com/openvinotoolkit/nncf/blob/develop/docs/usage/training_time_compression/other_algorithms/LegacyQuantization.md#asymmetric-quantization), to make it [symmetric](https://github.com/openvinotoolkit/nncf/blob/develop/docs/usage/training_time_compression/other_algorithms/LegacyQuantization.md#symmetric-quantization) you can add `sym=True`.
@@ -76,12 +98,21 @@ For 4-bit quantization you can also specify the following arguments in the quant
 Smaller `group_size` and `ratio` values usually improve accuracy at the sacrifice of the model size and inference latency.
 
 Quality of 4-bit weight compressed model can further be improved by employing one of the following data-dependent methods:
-* AWQ which stands for Activation Aware Quantization is an algorithm that tunes model weights for more accurate 4-bit compression. It slightly improves generation quality of compressed LLMs, but requires significant additional time and memory for tuning weights on a calibration dataset. Please note that it is possible that there will be no matching patterns in the model to apply AWQ, in such case it will be skipped.
-* Scale Estimation is a method that tunes quantization scales to minimize the `L2` error between the original and compressed layers. Providing a dataset is required to run scale estimation. Using this method also incurs additional time and memory overhead.
+* **AWQ** which stands for Activation Aware Quantization is an algorithm that tunes model weights for more accurate 4-bit compression. It slightly improves generation quality of compressed LLMs, but requires significant additional time and memory for tuning weights on a calibration dataset. Please note that it is possible that there will be no matching patterns in the model to apply AWQ, in such case it will be skipped.
+* **Scale Estimation** is a method that tunes quantization scales to minimize the `L2` error between the original and compressed layers. Providing a dataset is required to run scale estimation. Using this method also incurs additional time and memory overhead.
+* **GPTQ** optimizes compressed weights in a layer-wise fashion to minimize the difference between activations of a compressed and original layer.
 
-AWQ and Scale Estimation algorithms can be applied together or separately. For that, provide corresponding arguments to the 4-bit `OVWeightQuantizationConfig` together with a dataset. For example:
+Data-aware algorithms can be applied together or separately. For that, provide corresponding arguments to the 4-bit `OVWeightQuantizationConfig` together with a dataset. For example:
 ```python
-quantization_config = OVWeightQuantizationConfig(bits=4, sym=False, ratio=0.8, quant_method="awq", scale_estimation=True, dataset="wikitext2")
+quantization_config = OVWeightQuantizationConfig(
+    bits=4,
+    sym=False,
+    ratio=0.8,
+    quant_method="awq",
+    scale_estimation=True,
+    gptq=True,
+    dataset="wikitext2"
+)
 ```
 
 ### Static quantization
