commit files to HF hub

README.md

@@ -0,0 +1,25 @@
+---
+language:
+- en
+tags:
+- openvino
+---
+
+# MIT/ast-finetuned-speech-commands-v2
+
+This is the [MIT/ast-finetuned-speech-commands-v2](https://huggingface.co/MIT/ast-finetuned-speech-commands-v2) model converted to [OpenVINO](https://openvino.ai), for accellerated inference.
+
+An example of how to do inference on this model:
+```python
+from optimum.intel.openvino import OVModelForAudioClassification
+from transformers import AutoTokenizer, pipeline
+
+# model_id should be set to either a local directory or a model available on the HuggingFace hub.
+model_id = "helenai/MIT-ast-finetuned-speech-commands-v2-ov"
+feature_extractor = AutoFeatureExtractor.from_pretrained(model_id)
+model = OVModelForAudioClassification.from_pretrained(model_id)
+pipe = pipeline("audio-classification", model=model, feature_extractor=feature_extractor)
+result = pipe("https://datasets-server.huggingface.co/assets/speech_commands/--/v0.01/test/38/audio/audio.mp3")
+print(result)
+```
+

config.json

@@ -0,0 +1,98 @@
+{
+  "_name_or_path": "MIT/ast-finetuned-speech-commands-v2",
+  "architectures": [
+    "ASTForAudioClassification"
+  ],
+  "attention_probs_dropout_prob": 0.0,
+  "frequency_stride": 10,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.0,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "backward",
+    "1": "follow",
+    "2": "five",
+    "3": "bed",
+    "4": "zero",
+    "5": "on",
+    "6": "learn",
+    "7": "two",
+    "8": "house",
+    "9": "tree",
+    "10": "dog",
+    "11": "stop",
+    "12": "seven",
+    "13": "eight",
+    "14": "down",
+    "15": "six",
+    "16": "forward",
+    "17": "cat",
+    "18": "right",
+    "19": "visual",
+    "20": "four",
+    "21": "wow",
+    "22": "no",
+    "23": "nine",
+    "24": "off",
+    "25": "three",
+    "26": "left",
+    "27": "marvin",
+    "28": "yes",
+    "29": "up",
+    "30": "sheila",
+    "31": "happy",
+    "32": "bird",
+    "33": "go",
+    "34": "one"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "backward": 0,
+    "bed": 3,
+    "bird": 32,
+    "cat": 17,
+    "dog": 10,
+    "down": 14,
+    "eight": 13,
+    "five": 2,
+    "follow": 1,
+    "forward": 16,
+    "four": 20,
+    "go": 33,
+    "happy": 31,
+    "house": 8,
+    "learn": 6,
+    "left": 26,
+    "marvin": 27,
+    "nine": 23,
+    "no": 22,
+    "off": 24,
+    "on": 5,
+    "one": 34,
+    "right": 18,
+    "seven": 12,
+    "sheila": 30,
+    "six": 15,
+    "stop": 11,
+    "three": 25,
+    "tree": 9,
+    "two": 7,
+    "up": 29,
+    "visual": 19,
+    "wow": 21,
+    "yes": 28,
+    "zero": 4
+  },
+  "layer_norm_eps": 1e-12,
+  "max_length": 128,
+  "model_type": "audio-spectrogram-transformer",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "num_mel_bins": 128,
+  "patch_size": 16,
+  "qkv_bias": true,
+  "time_stride": 10,
+  "torch_dtype": "float32",
+  "transformers_version": "4.26.1"
+}

inference.py

@@ -0,0 +1,10 @@
+from optimum.intel.openvino import OVModelForAudioClassification
+from transformers import AutoTokenizer, pipeline
+
+# model_id should be set to either a local directory or a model available on the HuggingFace hub.
+model_id = "helenai/MIT-ast-finetuned-speech-commands-v2-ov"
+feature_extractor = AutoFeatureExtractor.from_pretrained(model_id)
+model = OVModelForAudioClassification.from_pretrained(model_id)
+pipe = pipeline("audio-classification", model=model, feature_extractor=feature_extractor)
+result = pipe("https://datasets-server.huggingface.co/assets/speech_commands/--/v0.01/test/38/audio/audio.mp3")
+print(result)

openvino_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89a88fc7b9be32a3c8e60ad6738d24f33ee1dbc4e611a3d1b8bee5a4ef7d4095
+size 170794314

preprocessor_config.json

@@ -0,0 +1,13 @@
+{
+  "do_normalize": true,
+  "feature_extractor_type": "ASTFeatureExtractor",
+  "feature_size": 1,
+  "max_length": 128,
+  "mean": -6.845978,
+  "num_mel_bins": 128,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "return_attention_mask": false,
+  "sampling_rate": 16000,
+  "std": 5.5654526
+}