#4003: Adding whisper functional model

models/experimental/functional_whisper/reference/torch_baseline_whisper.py

@@ -0,0 +1,24 @@
+# SPDX-FileCopyrightText: © 2023 Tenstorrent Inc.
+
+# SPDX-License-Identifier: Apache-2.0
+
+from transformers import AutoFeatureExtractor, WhisperModel
+from datasets import load_dataset
+import torch
+
+
+# Generates the baseline expected results for tests within ttnn
+if __name__ == "__main__":
+    model_name = "openai/whisper-base"
+    model = WhisperModel.from_pretrained(model_name).to(torch.bfloat16).eval()
+    feature_extractor = AutoFeatureExtractor.from_pretrained("openai/whisper-base")
+    ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
+    inputs = feature_extractor(ds[0]["audio"]["array"], sampling_rate=16000, return_tensors="pt")
+    input_features = inputs.input_features.type(torch.bfloat16)
+    decoder_input_ids = torch.ones(1, 32).type(torch.int32) * model.config.decoder_start_token_id
+    parameters = model.state_dict()
+    print(parameters.keys())
+    last_hidden_state = model(input_features, decoder_input_ids=decoder_input_ids).last_hidden_state
+    print(last_hidden_state.shape)
+    last_three = last_hidden_state[0, -1, -3:]
+    print(last_three)