commit files to HF hub

README.md

@@ -7,25 +7,19 @@ tags:
 
 # anton-l/wav2vec2-base-superb-sd
 
-This is the [anton-l/wav2vec2-base-superb-sd](https://huggingface.co/anton-l/wav2vec2-base-superb-sd) model, converted
-to OpenVINO. An example of how to do inference on this model:
+This is the [anton-l/wav2vec2-base-superb-sd](https://huggingface.co/anton-l/wav2vec2-base-superb-sd) model converted to [OpenVINO](https://openvino.ai), for accelerated inference.
 
+An example of how to do inference on this model:
 ```python
-from transformers import AutoFeatureExtractor
-from optimum.intel.openvino import OVModelForAudioFrameClassification
-from datasets import load_dataset
-import torch
+from optimum.intel import OVModelForAudioFrameClassification
+from transformers import AutoFeatureExtractor, pipeline
 
-dataset = load_dataset("hf-internal-testing/librispeech_asr_demo", "clean", split="validation")
-dataset = dataset.sort("id")
-sampling_rate = dataset.features["audio"].sampling_rate
-
-feature_extractor = AutoFeatureExtractor.from_pretrained("helenai/anton-l-wav2vec2-base-superb-sd-ov")
-model =  OVModelForAudioFrameClassification.from_pretrained("helenai/anton-l-wav2vec2-base-superb-sd-ov")
-inputs = feature_extractor(dataset[0]["audio"]["array"], return_tensors="np", sampling_rate=sampling_rate)
-logits = model(**inputs).logits
-
-probabilities = torch.sigmoid(torch.as_tensor(logits)[0])
-labels = (probabilities > 0.5).long()
-print(labels[0].tolist())
+# model_id should be set to either a local directory or a model available on the HuggingFace hub.
+model_id = "helenai/anton-l-wav2vec2-base-superb-sd-ov"
+feature_extractor = AutoFeatureExtractor.from_pretrained(model_id, trust_remote_code=True)
+model = OVModelForAudioFrameClassification.from_pretrained(model_id)
+pipe = pipeline("None", model=model, feature_extractor=feature_extractor)
+result = pipe("hello world")
+print(result)
 ```
+

config.json

@@ -1,6 +1,7 @@
 {
   "_name_or_path": "anton-l/wav2vec2-base-superb-sd",
   "activation_dropout": 0.0,
+  "adapter_attn_dim": null,
   "adapter_kernel_size": 3,
   "adapter_stride": 2,
   "add_adapter": false,
@@ -118,7 +119,7 @@
     1
   ],
   "torch_dtype": "float32",
-  "transformers_version": "4.28.1",
+  "transformers_version": "4.39.0",
   "use_weighted_layer_sum": true,
   "vocab_size": 32,
   "xvector_output_dim": 512

inference.py

@@ -0,0 +1,10 @@
+from optimum.intel import OVModelForAudioFrameClassification
+from transformers import AutoFeatureExtractor, pipeline
+
+# model_id should be set to either a local directory or a model available on the HuggingFace hub.
+model_id = "helenai/anton-l-wav2vec2-base-superb-sd-ov"
+feature_extractor = AutoFeatureExtractor.from_pretrained(model_id, trust_remote_code=True)
+model = OVModelForAudioFrameClassification.from_pretrained(model_id)
+pipe = pipeline("None", model=model, feature_extractor=feature_extractor)
+result = pipe("hello world")
+print(result)

openvino_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:561fc066c6afba3c5fa987b2a022ba1a4721856e2a179569cb454a63a528b0f6
-size 377493804
+oid sha256:231a8c2e48c0feddd0480116cb79181c5880537c411edfd3b3ee1110ab703f30
+size 188744922

preprocessor_config.json

@@ -0,0 +1,9 @@
+{
+  "do_normalize": false,
+  "feature_extractor_type": "Wav2Vec2FeatureExtractor",
+  "feature_size": 1,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "return_attention_mask": true,
+  "sampling_rate": 16000
+}