diff --git a/.gitattributes b/.gitattributes index 720577f6a623cf375ee10b6c1bb7576fecd96b96..a9259ff0147e4ff3cc1677a41c974d0babde90f3 100644 --- a/.gitattributes +++ b/.gitattributes @@ -56,3 +56,7 @@ examples/ASR/Tedlium3-Long-form-Test/sample_0.wav filter=lfs diff=lfs merge=lfs examples/ASR/Tedlium3-Long-form-Test/sample_1.wav filter=lfs diff=lfs merge=lfs -text examples/ASR/Tedlium3-Long-form-Test/sample_2.wav filter=lfs diff=lfs merge=lfs -text examples/SQA/Public-SG-Speech-QA-Test/sample_2.wav filter=lfs diff=lfs merge=lfs -text +examples/SQA/CN-College-Listen-MCQ-Test/sample_0.wav filter=lfs diff=lfs merge=lfs -text +examples/SQA/Spoken-Squad-Test/sample_0.wav filter=lfs diff=lfs merge=lfs -text +examples/SQA/Spoken-Squad-Test/sample_1.wav filter=lfs diff=lfs merge=lfs -text +examples/SQA/Spoken-Squad-Test/sample_2.wav filter=lfs diff=lfs merge=lfs -text diff --git a/examples/AC/AudioCaps-Test/data-00000-of-00001.arrow b/examples/AC/AudioCaps-Test/data-00000-of-00001.arrow index 63b866bb4d0f6f63975f46c9f07b76b28559ee45..58406c24c9292ee43c93960c4478e4d05fb95f80 100644 --- a/examples/AC/AudioCaps-Test/data-00000-of-00001.arrow +++ b/examples/AC/AudioCaps-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:6b86af8e9fe94528cc1e80ab8e1543dcccb04421c9e73d50a03bec19bd5b22f8 -size 972360 +oid sha256:70213ecc31962d6d8bbc0e4d7ae2dd302c851a4af00f12b07735311f5f128288 +size 966216 diff --git a/examples/AC/AudioCaps-Test/sample_0.wav b/examples/AC/AudioCaps-Test/sample_0.wav index 509a39869ae9101b674f191a5887448af94d2664..4b2f8047fa38f9ba3acef7485b26ea02f4ada359 100644 Binary files a/examples/AC/AudioCaps-Test/sample_0.wav and b/examples/AC/AudioCaps-Test/sample_0.wav differ diff --git a/examples/AC/AudioCaps-Test/sample_1.wav b/examples/AC/AudioCaps-Test/sample_1.wav index 1c964ebcc0ac6e615f72aa92421b336272c5e5c2..8724df3f01ccd22778f84a7a851871f1d73434fe 100644 Binary files a/examples/AC/AudioCaps-Test/sample_1.wav and b/examples/AC/AudioCaps-Test/sample_1.wav differ diff --git a/examples/AC/AudioCaps-Test/sample_2.wav b/examples/AC/AudioCaps-Test/sample_2.wav index 1b6418b2410e278e532d31ea339e74f6a29585dc..48d08b19be5f0904ca976c35fbe4ae4d6c19435f 100644 Binary files a/examples/AC/AudioCaps-Test/sample_2.wav and b/examples/AC/AudioCaps-Test/sample_2.wav differ diff --git a/examples/AC/AudioCaps-Test/state.json b/examples/AC/AudioCaps-Test/state.json index 0cd0d4b3978e10a5f85d569ffc028dacc1886b74..fcc438d7c2db870c1b636d8299a9d843d607fbc6 100644 --- a/examples/AC/AudioCaps-Test/state.json +++ b/examples/AC/AudioCaps-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "3729f6c56764c342", + "_fingerprint": "e736bf1821a473f3", "_format_columns": [ "context", "instruction", diff --git a/examples/AC/WavCaps-Test/data-00000-of-00001.arrow b/examples/AC/WavCaps-Test/data-00000-of-00001.arrow index becb1bfe92786e01b7dde8c0122c2549e481780f..ad78b874ff3abccb165ea8ef522d91d8c0f06b48 100644 --- a/examples/AC/WavCaps-Test/data-00000-of-00001.arrow +++ b/examples/AC/WavCaps-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:398df95837d355a7b2329ce2e6805f8a6b88867630bf6650c1f04ca71c9819b9 -size 720624 +oid sha256:9636af636286b1aedad840ccef31ca7d879e824ddc6814fcb7411b7fbdf411aa +size 812352 diff --git a/examples/AC/WavCaps-Test/sample_0.wav b/examples/AC/WavCaps-Test/sample_0.wav index 08a85f269beade4c541e3f49a9b2518f31a95ed9..ad8d45455c35860d7309e0554c6610ba6ddccb68 100644 Binary files a/examples/AC/WavCaps-Test/sample_0.wav and b/examples/AC/WavCaps-Test/sample_0.wav differ diff --git a/examples/AC/WavCaps-Test/sample_1.wav b/examples/AC/WavCaps-Test/sample_1.wav index 462d0c2d0189352555dd97182326599b27096f43..ea10461620e829d47fb78bf4d827b95322791340 100644 Binary files a/examples/AC/WavCaps-Test/sample_1.wav and b/examples/AC/WavCaps-Test/sample_1.wav differ diff --git a/examples/AC/WavCaps-Test/sample_2.wav b/examples/AC/WavCaps-Test/sample_2.wav index 62717178e93e79c98c5b81bc4122f65bc6b52efe..b7fef91fbedf60a5d58f4f9fb93d95c1e205bf67 100644 Binary files a/examples/AC/WavCaps-Test/sample_2.wav and b/examples/AC/WavCaps-Test/sample_2.wav differ diff --git a/examples/AC/WavCaps-Test/state.json b/examples/AC/WavCaps-Test/state.json index a3eb658121b9663c5f4d92aa705da048df185326..0d52b23c8e5b93506a8af809adf9680c9cc7bf86 100644 --- a/examples/AC/WavCaps-Test/state.json +++ b/examples/AC/WavCaps-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "6c3c2a5f2db349d8", + "_fingerprint": "742ab313af054565", "_format_columns": [ "context", "instruction", diff --git a/examples/AQA/AudioCaps-QA-Test/data-00000-of-00001.arrow b/examples/AQA/AudioCaps-QA-Test/data-00000-of-00001.arrow index ea256014e31662a291f0dedd7201b100a43a96aa..91a95941f9251c2e3abe654fa45035c8015d364b 100644 --- a/examples/AQA/AudioCaps-QA-Test/data-00000-of-00001.arrow +++ b/examples/AQA/AudioCaps-QA-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:545f230fd01be33ef8447cf2acacc791d2be64b72311230ee3660dcfc282c609 -size 971952 +oid sha256:92b0b2fe81ee0e3a2690a444bb9b68994d89ca53db6ce174f5802293549256d1 +size 953616 diff --git a/examples/AQA/AudioCaps-QA-Test/sample_0.wav b/examples/AQA/AudioCaps-QA-Test/sample_0.wav index 59fcd56a181073f56c1f0fba45f0ccda8a6337e2..2eec3251fe8dc9acf17f43f66f187a277cf6c6b0 100644 Binary files a/examples/AQA/AudioCaps-QA-Test/sample_0.wav and b/examples/AQA/AudioCaps-QA-Test/sample_0.wav differ diff --git a/examples/AQA/AudioCaps-QA-Test/sample_1.wav b/examples/AQA/AudioCaps-QA-Test/sample_1.wav index 7ca320beb2ea16f32257519116b1f17d456f9bc4..f7e101c5918451111738962b722e47041dd59227 100644 Binary files a/examples/AQA/AudioCaps-QA-Test/sample_1.wav and b/examples/AQA/AudioCaps-QA-Test/sample_1.wav differ diff --git a/examples/AQA/AudioCaps-QA-Test/sample_2.wav b/examples/AQA/AudioCaps-QA-Test/sample_2.wav index 6d062d0a489a8b0334deee75ebe530dbae953dd2..cb15b2ffff83c6ec5541c8b54a8205d58292a2d3 100644 Binary files a/examples/AQA/AudioCaps-QA-Test/sample_2.wav and b/examples/AQA/AudioCaps-QA-Test/sample_2.wav differ diff --git a/examples/AQA/AudioCaps-QA-Test/state.json b/examples/AQA/AudioCaps-QA-Test/state.json index 35ad9ea556daeee418b93d6a77ad56c5cf801dec..267d6cbee3f52f7b8f77f20b959ca9ce159aed16 100644 --- a/examples/AQA/AudioCaps-QA-Test/state.json +++ b/examples/AQA/AudioCaps-QA-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "026dfac674d9ef77", + "_fingerprint": "52bc1dfcaf2a0f4b", "_format_columns": [ "context", "instruction", diff --git a/examples/AQA/Clotho-AQA-Test/data-00000-of-00001.arrow b/examples/AQA/Clotho-AQA-Test/data-00000-of-00001.arrow index 8015d1c105a7dfc1559751da4dfd0d2462fcdd2c..8b76eb63bc6a3adcfad98e20600a07bc5cf84e1b 100644 --- a/examples/AQA/Clotho-AQA-Test/data-00000-of-00001.arrow +++ b/examples/AQA/Clotho-AQA-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:c706ed113af21d49025bd52ba2ebb8fce382ef5c4ecfac32b6173222ae4953b0 -size 1907768 +oid sha256:08b4de88bf163bbfd2097769e6104ba8514380a7a097741d38e1ccc41d5b0f86 +size 2035832 diff --git a/examples/AQA/Clotho-AQA-Test/dataset_info.json b/examples/AQA/Clotho-AQA-Test/dataset_info.json index e584e27299bec6d9aa2eee45871c216c67ac17b0..1b05abb4dce6b496c0a3c6043f27e4ca1f225320 100644 --- a/examples/AQA/Clotho-AQA-Test/dataset_info.json +++ b/examples/AQA/Clotho-AQA-Test/dataset_info.json @@ -32,7 +32,115 @@ "_type": "Value" } }, - "other_attributes": {} + "other_attributes": {}, + "salmonn_7b": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "wavllm_fairseq": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "Qwen2-Audio-7B-Instruct": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "whisper_large_v3_with_llama_3_8b_instruct": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "mowe_audio": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "qwen_audio_chat": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + } }, "homepage": "", "license": "" diff --git a/examples/AQA/Clotho-AQA-Test/sample_0.wav b/examples/AQA/Clotho-AQA-Test/sample_0.wav index c48c65eea211ef53e5929d3744cbd0e73fd166a0..d6a07e6172778d85080c04531658efc7443ae03d 100644 Binary files a/examples/AQA/Clotho-AQA-Test/sample_0.wav and b/examples/AQA/Clotho-AQA-Test/sample_0.wav differ diff --git a/examples/AQA/Clotho-AQA-Test/sample_1.wav b/examples/AQA/Clotho-AQA-Test/sample_1.wav index b2a9524bb08edd5e5ab700fec3d49b610338efd2..139bd3226ca457718b3cdab6d1e7a99dd5e4bd01 100644 Binary files a/examples/AQA/Clotho-AQA-Test/sample_1.wav and b/examples/AQA/Clotho-AQA-Test/sample_1.wav differ diff --git a/examples/AQA/Clotho-AQA-Test/sample_2.wav b/examples/AQA/Clotho-AQA-Test/sample_2.wav index ed8314310c19afb4235badd32f937feeb25387b5..b62a8feb71883e7540f521d41ac6e3eefe3862a3 100644 Binary files a/examples/AQA/Clotho-AQA-Test/sample_2.wav and b/examples/AQA/Clotho-AQA-Test/sample_2.wav differ diff --git a/examples/AQA/Clotho-AQA-Test/state.json b/examples/AQA/Clotho-AQA-Test/state.json index 9cdc1447ea2f55e517f57bddb7f7fefb175a45b3..c58228cce70f0b257254856751c37d68dd8cd64f 100644 --- a/examples/AQA/Clotho-AQA-Test/state.json +++ b/examples/AQA/Clotho-AQA-Test/state.json @@ -4,12 +4,18 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "515a1722077187bd", + "_fingerprint": "e2e76326f448d7c4", "_format_columns": [ "context", "instruction", "answer", - "other_attributes" + "other_attributes", + "salmonn_7b", + "wavllm_fairseq", + "Qwen2-Audio-7B-Instruct", + "whisper_large_v3_with_llama_3_8b_instruct", + "mowe_audio", + "qwen_audio_chat" ], "_format_kwargs": {}, "_format_type": null, diff --git a/examples/AQA/WavCaps-QA-Test/data-00000-of-00001.arrow b/examples/AQA/WavCaps-QA-Test/data-00000-of-00001.arrow index 9083a9e4a5847fc1fd575217bcdaecca6bc6a5f7..5ce3c3745ad16319f0acfb100c443202a55485af 100644 --- a/examples/AQA/WavCaps-QA-Test/data-00000-of-00001.arrow +++ b/examples/AQA/WavCaps-QA-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:c33abcdce0b35d086c28cf0ba814d9edabfb596c0f40b312968d2e32d30f4432 -size 976888 +oid sha256:1736d5bcc9ca0d8d4847d7d000e6c6e63c73f6262177ea0391d180c40649da39 +size 837920 diff --git a/examples/AQA/WavCaps-QA-Test/sample_0.wav b/examples/AQA/WavCaps-QA-Test/sample_0.wav index a7483f6a72f398b0b35db48322f6bcadb048867d..7639cdca2866a648ce90b4f5e385e3e6dc56c04a 100644 Binary files a/examples/AQA/WavCaps-QA-Test/sample_0.wav and b/examples/AQA/WavCaps-QA-Test/sample_0.wav differ diff --git a/examples/AQA/WavCaps-QA-Test/sample_1.wav b/examples/AQA/WavCaps-QA-Test/sample_1.wav index 41cf6d0cc9f8990b3599985156b44550dcbc9dfb..d2cc1a6def6014328e02ea5ea25019414f8960b4 100644 Binary files a/examples/AQA/WavCaps-QA-Test/sample_1.wav and b/examples/AQA/WavCaps-QA-Test/sample_1.wav differ diff --git a/examples/AQA/WavCaps-QA-Test/sample_2.wav b/examples/AQA/WavCaps-QA-Test/sample_2.wav index 13f24a2d1bfc7e2be0519def176d968885691e74..9629f69dd6d6b8a713b9122b03ee04ec4aae8857 100644 Binary files a/examples/AQA/WavCaps-QA-Test/sample_2.wav and b/examples/AQA/WavCaps-QA-Test/sample_2.wav differ diff --git a/examples/AQA/WavCaps-QA-Test/state.json b/examples/AQA/WavCaps-QA-Test/state.json index 89917a40550177f4b54b4e3fe0885df6d78d0aab..a85966493650dfae84811c91b9f42d7c2811ad9c 100644 --- a/examples/AQA/WavCaps-QA-Test/state.json +++ b/examples/AQA/WavCaps-QA-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "46b38bc22103a7cd", + "_fingerprint": "40995a6cc1fe3dc7", "_format_columns": [ "context", "instruction", diff --git a/examples/AR/VoxCeleb-Accent-Test/data-00000-of-00001.arrow b/examples/AR/VoxCeleb-Accent-Test/data-00000-of-00001.arrow index 5b7e79961738f9a58c839b0c2920f4658152e4b0..4d554c3929f3b9461fca53182975f730f3e8566d 100644 --- a/examples/AR/VoxCeleb-Accent-Test/data-00000-of-00001.arrow +++ b/examples/AR/VoxCeleb-Accent-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:f899c6ad68e95ddacddf1c017f802018d1089d084831b573f5fb81431b0b98a9 -size 1145784 +oid sha256:5307150e5b08f9bfd7b81d56ca7ffbee1d731d002417d20fc3ec0713bc764533 +size 730864 diff --git a/examples/AR/VoxCeleb-Accent-Test/sample_0.wav b/examples/AR/VoxCeleb-Accent-Test/sample_0.wav index 637255b8ce4a76eab5145234259d4d9e27a7449d..ae8061522976216887910263ba9414a7e60685eb 100644 Binary files a/examples/AR/VoxCeleb-Accent-Test/sample_0.wav and b/examples/AR/VoxCeleb-Accent-Test/sample_0.wav differ diff --git a/examples/AR/VoxCeleb-Accent-Test/sample_1.wav b/examples/AR/VoxCeleb-Accent-Test/sample_1.wav index 81f38dc724603587c4c75c351b90f236b23a77d7..209cf78dcde2791b463e32f9a7245514655790aa 100644 Binary files a/examples/AR/VoxCeleb-Accent-Test/sample_1.wav and b/examples/AR/VoxCeleb-Accent-Test/sample_1.wav differ diff --git a/examples/AR/VoxCeleb-Accent-Test/sample_2.wav b/examples/AR/VoxCeleb-Accent-Test/sample_2.wav index f34db19cc48eb83b64600413f61e2e4eff07d2cd..0d44634575f30bfe9c2fa2c2765ae34c192c9c90 100644 Binary files a/examples/AR/VoxCeleb-Accent-Test/sample_2.wav and b/examples/AR/VoxCeleb-Accent-Test/sample_2.wav differ diff --git a/examples/AR/VoxCeleb-Accent-Test/state.json b/examples/AR/VoxCeleb-Accent-Test/state.json index a0e4beb3d20ddc496c9863a6d3809657a69c09ea..5354f99a49cf31f0949fd2359846d46859efd5ff 100644 --- a/examples/AR/VoxCeleb-Accent-Test/state.json +++ b/examples/AR/VoxCeleb-Accent-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "f1df87f5b3ca8c97", + "_fingerprint": "fa91a59f90c22c3c", "_format_columns": [ "context", "instruction", diff --git a/examples/ASR/Common-Voice-15-En-Test/data-00000-of-00001.arrow b/examples/ASR/Common-Voice-15-En-Test/data-00000-of-00001.arrow index ad9848e1ed66f37be73e1db95318ec944d00109a..4560d6338d0bda974170bff980b354867e681e90 100644 --- a/examples/ASR/Common-Voice-15-En-Test/data-00000-of-00001.arrow +++ b/examples/ASR/Common-Voice-15-En-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:d297ee75cd36249355aec83e6526d78c2592e44b1ca8865a31e3df576a0002dc -size 817384 +oid sha256:f9a561ce9aed8ba4c02f86c90883029e1bb566e2b66986b17874f3bb5884d67d +size 489552 diff --git a/examples/ASR/Common-Voice-15-En-Test/sample_0.wav b/examples/ASR/Common-Voice-15-En-Test/sample_0.wav index 8e9932f1a577bd50d53c17860ba30589866112d8..42753b756f05c733803356b486de2df1b1224de4 100644 Binary files a/examples/ASR/Common-Voice-15-En-Test/sample_0.wav and b/examples/ASR/Common-Voice-15-En-Test/sample_0.wav differ diff --git a/examples/ASR/Common-Voice-15-En-Test/sample_1.wav b/examples/ASR/Common-Voice-15-En-Test/sample_1.wav index d17339bbdb7e343da7e8514f03bdbfe8d9ed399f..643e1d9e9e461c2465856a18fbf89bb27f577a18 100644 Binary files a/examples/ASR/Common-Voice-15-En-Test/sample_1.wav and b/examples/ASR/Common-Voice-15-En-Test/sample_1.wav differ diff --git a/examples/ASR/Common-Voice-15-En-Test/sample_2.wav b/examples/ASR/Common-Voice-15-En-Test/sample_2.wav index b760edbde1d6bfe80fcf77aecf7ac87aa21ecd4d..eb0894ce127ebe7c2fadb5b11feea3e5b0ace14f 100644 Binary files a/examples/ASR/Common-Voice-15-En-Test/sample_2.wav and b/examples/ASR/Common-Voice-15-En-Test/sample_2.wav differ diff --git a/examples/ASR/Common-Voice-15-En-Test/state.json b/examples/ASR/Common-Voice-15-En-Test/state.json index 80d695eaf0d326aba5c1c42c08f6620b7711f3fc..1ff74abf94d1cbf9804c3911eac7edf199fb36a2 100644 --- a/examples/ASR/Common-Voice-15-En-Test/state.json +++ b/examples/ASR/Common-Voice-15-En-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "5a02a12eee6eb15a", + "_fingerprint": "468db91ad949e4d4", "_format_columns": [ "context", "instruction", diff --git a/examples/ASR/Earnings21-Test/data-00000-of-00001.arrow b/examples/ASR/Earnings21-Test/data-00000-of-00001.arrow index b2177cd6939d71d10ca2b7b71cecec54dfc7f6d9..543eb95281d82eea0d8930770f2a47b8bd52340c 100644 --- a/examples/ASR/Earnings21-Test/data-00000-of-00001.arrow +++ b/examples/ASR/Earnings21-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:2bada2056866d5e784c649b65add3928a859b35304e9e240e59f94da9f60ccaf -size 235999160 +oid sha256:5fbca261ae8ac8ccc96993ea11f17836bdcfef1070835784f159b79990a5a298 +size 429108160 diff --git a/examples/ASR/Earnings21-Test/sample_0.wav b/examples/ASR/Earnings21-Test/sample_0.wav index c306304487d6f70aee07421b5d8074fda883d82c..877a0f85e77ae3ccb28f14601c2e8765e945c6d8 100644 --- a/examples/ASR/Earnings21-Test/sample_0.wav +++ b/examples/ASR/Earnings21-Test/sample_0.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:db092ee918b48e37b79b079e2de667d127c0008293a0309be1c6a16a4840b820 -size 46016300 +oid sha256:d8fb994964e1b0df9f4675ceaa73d55da55a096f5b94d002d9f7b07c997fc83e +size 97593644 diff --git a/examples/ASR/Earnings21-Test/sample_1.wav b/examples/ASR/Earnings21-Test/sample_1.wav index 338e04286927578da9e1d1ba3ec297bcbf31ae42..e20907a027c9da25fc6544ec87f07185b23c3567 100644 --- a/examples/ASR/Earnings21-Test/sample_1.wav +++ b/examples/ASR/Earnings21-Test/sample_1.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:64031076b7b5a039154ac523bf5a7938d4b1dd8bb30a9a9f55c4cf579347a75a -size 80607654 +oid sha256:bd6ba77731011a6dc02e5854a600a2036713be4c2d71abf63fd6a89b86083c4f +size 178791280 diff --git a/examples/ASR/Earnings21-Test/sample_2.wav b/examples/ASR/Earnings21-Test/sample_2.wav index a9452aa909f581964033a3da8146f4c787f8ef79..0924db55e5560d30768fc9fa8a6e5931f475a642 100644 --- a/examples/ASR/Earnings21-Test/sample_2.wav +++ b/examples/ASR/Earnings21-Test/sample_2.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:33430292a1535095ad33c1b6e428d77d0bdba85a0299eee6a36f45e100c39392 -size 108053804 +oid sha256:1a1d15425069b003730e79f0df467103f4ac7670f87a0539a97c82973a02943e +size 150700076 diff --git a/examples/ASR/Earnings21-Test/state.json b/examples/ASR/Earnings21-Test/state.json index 03ec809e2ddf9f2fe68b495f0237cb499f70be44..ce1987be7235db123fcdddfe4a75272abf7513d4 100644 --- a/examples/ASR/Earnings21-Test/state.json +++ b/examples/ASR/Earnings21-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "0d42a0f2cebd16d8", + "_fingerprint": "8cc0ad99446f1aba", "_format_columns": [ "context", "instruction", diff --git a/examples/ASR/Earnings22-Test/data-00000-of-00001.arrow b/examples/ASR/Earnings22-Test/data-00000-of-00001.arrow index b7b239653af85350422395be4abfbcb9b0e194b0..803b349472ac02701fad19e91bdc02942939196d 100644 --- a/examples/ASR/Earnings22-Test/data-00000-of-00001.arrow +++ b/examples/ASR/Earnings22-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:12c1522382953c79af65ca9df82606ab5aef8aea8c9646e4ecd3de94fc9d8d62 -size 220986632 +oid sha256:497dd6d287df9a8be5194b7875ae88f275127986d3fc538601382e80244bbb7b +size 332277848 diff --git a/examples/ASR/Earnings22-Test/sample_0.wav b/examples/ASR/Earnings22-Test/sample_0.wav index bf1eaeaefa512349520caaecc4885a49c67efdb3..c31c73e0e078edb570b30c9ddcd8679e40ac62dd 100644 --- a/examples/ASR/Earnings22-Test/sample_0.wav +++ b/examples/ASR/Earnings22-Test/sample_0.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:148fc052e32561e0b60d3e3c86b4fd4c2b319ff9aefd1b2b718cf12e2a2a457c -size 64738578 +oid sha256:6530f950b07b4747ca26fa70586ec563cf20c222cab6e53fc2e94e28f7a3d867 +size 167237960 diff --git a/examples/ASR/Earnings22-Test/sample_1.wav b/examples/ASR/Earnings22-Test/sample_1.wav index 44cc571edb96c75656eee8bebe27c76d5d7988a8..e7fc4350e24534cdda9135085b77b08356adf39b 100644 --- a/examples/ASR/Earnings22-Test/sample_1.wav +++ b/examples/ASR/Earnings22-Test/sample_1.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:f3ed5fea7d5d300ae1dabe9391dddee347c54c82fa4356bc12e9ba5a28238edc -size 101514284 +oid sha256:7c4034503f1974ecfdbd050df7e75d97cbae4933ea62af829eefe6e17295c6f4 +size 38863630 diff --git a/examples/ASR/Earnings22-Test/sample_2.wav b/examples/ASR/Earnings22-Test/sample_2.wav index 6f842c17be2dfe11a92689b2a74ff3c49d33ff59..12c42f881e73d8b6f1905855b844c1acc816553f 100644 --- a/examples/ASR/Earnings22-Test/sample_2.wav +++ b/examples/ASR/Earnings22-Test/sample_2.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:ec1e241971074b8965b7069596c5fa1c7cf42573e3a341eec76857239088a0a5 -size 53823852 +oid sha256:ba099990e90f3db2901b585ce601b71960ab54ef76f5bf143a74d816ea615f07 +size 124561076 diff --git a/examples/ASR/Earnings22-Test/state.json b/examples/ASR/Earnings22-Test/state.json index fb5db0657dff508744e282d20abdbfcf783797eb..ac26bdba683e345445fdeb81908f03626f2ced73 100644 --- a/examples/ASR/Earnings22-Test/state.json +++ b/examples/ASR/Earnings22-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "1427a3866fe2cb1a", + "_fingerprint": "331c061bce6e651c", "_format_columns": [ "context", "instruction", diff --git a/examples/ASR/GigaSpeech-Test/data-00000-of-00001.arrow b/examples/ASR/GigaSpeech-Test/data-00000-of-00001.arrow index a0d06c6021eab93248fba649dfde9c9762d86cbc..9f584469430bb4f4aecca47ac52ea723be454e83 100644 --- a/examples/ASR/GigaSpeech-Test/data-00000-of-00001.arrow +++ b/examples/ASR/GigaSpeech-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:5a5d409a5b065382a0e6391eb9d5e726f614216b5809034fa04a34a4ee5d3eda -size 847536 +oid sha256:e46b3463f6df894fa055f08fa9ea1f44b4cea43a1be43af2a2bd8c5e8fee3ec5 +size 573048 diff --git a/examples/ASR/GigaSpeech-Test/sample_0.wav b/examples/ASR/GigaSpeech-Test/sample_0.wav index d5b1a0190ab705e242079ac6552d760aa86442e3..0d232b1996ee07bc47f24d06fe8b860ee1b63b11 100644 Binary files a/examples/ASR/GigaSpeech-Test/sample_0.wav and b/examples/ASR/GigaSpeech-Test/sample_0.wav differ diff --git a/examples/ASR/GigaSpeech-Test/sample_1.wav b/examples/ASR/GigaSpeech-Test/sample_1.wav index 8d4715d3ee2255ff632248568ef394a3cdf11417..ee5fced84fd0691c2b4b288be9a5ad73ac67bc6e 100644 Binary files a/examples/ASR/GigaSpeech-Test/sample_1.wav and b/examples/ASR/GigaSpeech-Test/sample_1.wav differ diff --git a/examples/ASR/GigaSpeech-Test/sample_2.wav b/examples/ASR/GigaSpeech-Test/sample_2.wav index ce79e6d621393311dc585ed610534909eca05aa8..2a9edb2ea3e8a96c7ab70b232249205e765f02d0 100644 Binary files a/examples/ASR/GigaSpeech-Test/sample_2.wav and b/examples/ASR/GigaSpeech-Test/sample_2.wav differ diff --git a/examples/ASR/GigaSpeech-Test/state.json b/examples/ASR/GigaSpeech-Test/state.json index 34243fb62eea510a1aba5545b6b66d326f73dc0a..8bd5fd3d45201fc6807cb2364c48d5ba722bb4bf 100644 --- a/examples/ASR/GigaSpeech-Test/state.json +++ b/examples/ASR/GigaSpeech-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "9527d9b9b39b34c3", + "_fingerprint": "68d371cc267ff1d2", "_format_columns": [ "context", "instruction", diff --git a/examples/ASR/IMDA-Part1-ASR-Test/data-00000-of-00001.arrow b/examples/ASR/IMDA-Part1-ASR-Test/data-00000-of-00001.arrow index 9bbaec293c6a0bc3b06ba03fbcec88d4554dfc77..d290d6e2f73409b2f0ba18473a22149cf17268a6 100644 --- a/examples/ASR/IMDA-Part1-ASR-Test/data-00000-of-00001.arrow +++ b/examples/ASR/IMDA-Part1-ASR-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:eb0bffb3a27a05b4de88e853c7472ddccb1e86699b4f215151df94ab5f6624b7 -size 443200 +oid sha256:dd9b1d6729d5ef43f4d811def9a4796bfb383083727e3c4a0c79303089afb6ed +size 423624 diff --git a/examples/ASR/IMDA-Part1-ASR-Test/dataset_info.json b/examples/ASR/IMDA-Part1-ASR-Test/dataset_info.json index 5649fa6ac42fc9645c7aea4be9d03cc30d774647..af0260da818c604e190e3b198aebef1a391419ec 100644 --- a/examples/ASR/IMDA-Part1-ASR-Test/dataset_info.json +++ b/examples/ASR/IMDA-Part1-ASR-Test/dataset_info.json @@ -85,78 +85,6 @@ "_type": "Value" } } - }, - "salmonn_7b": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "wavllm_fairseq": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "whisper_large_v3_with_llama_3_8b_instruct": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "qwen_audio_chat": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } } }, "homepage": "", diff --git a/examples/ASR/IMDA-Part1-ASR-Test/sample_0.wav b/examples/ASR/IMDA-Part1-ASR-Test/sample_0.wav index bc8a46dd654a6ba711d10f8c3d4eda7c84bb20ce..2d696049b68e6f64944055f3bf5b7a01c8bcdad6 100644 Binary files a/examples/ASR/IMDA-Part1-ASR-Test/sample_0.wav and b/examples/ASR/IMDA-Part1-ASR-Test/sample_0.wav differ diff --git a/examples/ASR/IMDA-Part1-ASR-Test/sample_1.wav b/examples/ASR/IMDA-Part1-ASR-Test/sample_1.wav index 12fed4125c25fc3c4377b72a77b37fdd7f4fc4a4..6f926f6a97132a8df675f05d2f14f1dec232a704 100644 Binary files a/examples/ASR/IMDA-Part1-ASR-Test/sample_1.wav and b/examples/ASR/IMDA-Part1-ASR-Test/sample_1.wav differ diff --git a/examples/ASR/IMDA-Part1-ASR-Test/sample_2.wav b/examples/ASR/IMDA-Part1-ASR-Test/sample_2.wav index c4e2909a95a0a3090cab61cd24e9f534075fc8aa..f03f65739284f1757c34e88313998578f6dd67d5 100644 Binary files a/examples/ASR/IMDA-Part1-ASR-Test/sample_2.wav and b/examples/ASR/IMDA-Part1-ASR-Test/sample_2.wav differ diff --git a/examples/ASR/IMDA-Part1-ASR-Test/state.json b/examples/ASR/IMDA-Part1-ASR-Test/state.json index 6105b023680c0feaa933a66b55356fd3dd70a71b..f1cdd75238904ec8645cf1ef5a1a7e20981862a2 100644 --- a/examples/ASR/IMDA-Part1-ASR-Test/state.json +++ b/examples/ASR/IMDA-Part1-ASR-Test/state.json @@ -4,16 +4,12 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "e4d91fe35602a9e0", + "_fingerprint": "9c67b4f2f347692a", "_format_columns": [ "context", "instruction", "answer", - "other_attributes", - "salmonn_7b", - "wavllm_fairseq", - "whisper_large_v3_with_llama_3_8b_instruct", - "qwen_audio_chat" + "other_attributes" ], "_format_kwargs": {}, "_format_type": null, diff --git a/examples/ASR/IMDA-Part2-ASR-Test/data-00000-of-00001.arrow b/examples/ASR/IMDA-Part2-ASR-Test/data-00000-of-00001.arrow index a875f489bec97a4ac987c6efe85a2b74830e01e8..896c4fc75e607933d5155b06f9d4c67812c71847 100644 --- a/examples/ASR/IMDA-Part2-ASR-Test/data-00000-of-00001.arrow +++ b/examples/ASR/IMDA-Part2-ASR-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:7f5847c3e07bca2e318e9668982c04de51132580392439f0ca0db0d1b0f3353a -size 422448 +oid sha256:6203852fc63ee5b00dd73f6857d2a22a3f8a8dc3b87917bb99d1cef8448150a5 +size 438120 diff --git a/examples/ASR/IMDA-Part2-ASR-Test/dataset_info.json b/examples/ASR/IMDA-Part2-ASR-Test/dataset_info.json index 5649fa6ac42fc9645c7aea4be9d03cc30d774647..af0260da818c604e190e3b198aebef1a391419ec 100644 --- a/examples/ASR/IMDA-Part2-ASR-Test/dataset_info.json +++ b/examples/ASR/IMDA-Part2-ASR-Test/dataset_info.json @@ -85,78 +85,6 @@ "_type": "Value" } } - }, - "salmonn_7b": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "wavllm_fairseq": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "whisper_large_v3_with_llama_3_8b_instruct": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "qwen_audio_chat": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } } }, "homepage": "", diff --git a/examples/ASR/IMDA-Part2-ASR-Test/sample_0.wav b/examples/ASR/IMDA-Part2-ASR-Test/sample_0.wav index d5c33dee5bf21565cf381e88d3c4a1ca6dbebe78..d79fde19cec7b066b9f3f546d6ff6366dfa4daef 100644 Binary files a/examples/ASR/IMDA-Part2-ASR-Test/sample_0.wav and b/examples/ASR/IMDA-Part2-ASR-Test/sample_0.wav differ diff --git a/examples/ASR/IMDA-Part2-ASR-Test/sample_1.wav b/examples/ASR/IMDA-Part2-ASR-Test/sample_1.wav index 164c99dcf823a95655abe8aed5a11745da3d55d7..9fac54a02b8b9ad1ee6502ef41ef1d50a7213de5 100644 Binary files a/examples/ASR/IMDA-Part2-ASR-Test/sample_1.wav and b/examples/ASR/IMDA-Part2-ASR-Test/sample_1.wav differ diff --git a/examples/ASR/IMDA-Part2-ASR-Test/sample_2.wav b/examples/ASR/IMDA-Part2-ASR-Test/sample_2.wav index f945ea7be22586a84f70fc5fe138402953d608ce..d866ad7a1ddfc3541ce3adbecf00f688489cdd3f 100644 Binary files a/examples/ASR/IMDA-Part2-ASR-Test/sample_2.wav and b/examples/ASR/IMDA-Part2-ASR-Test/sample_2.wav differ diff --git a/examples/ASR/IMDA-Part2-ASR-Test/state.json b/examples/ASR/IMDA-Part2-ASR-Test/state.json index 8583a6d8b377ebe373993a561db34452f3d28444..3c6b051944e6842913bce863ca9c239da3f8e0be 100644 --- a/examples/ASR/IMDA-Part2-ASR-Test/state.json +++ b/examples/ASR/IMDA-Part2-ASR-Test/state.json @@ -4,16 +4,12 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "0ed051d84878e4e9", + "_fingerprint": "9f3d440792a605d2", "_format_columns": [ "context", "instruction", "answer", - "other_attributes", - "salmonn_7b", - "wavllm_fairseq", - "whisper_large_v3_with_llama_3_8b_instruct", - "qwen_audio_chat" + "other_attributes" ], "_format_kwargs": {}, "_format_type": null, diff --git a/examples/ASR/LibriSpeech-Test-Clean/data-00000-of-00001.arrow b/examples/ASR/LibriSpeech-Test-Clean/data-00000-of-00001.arrow index 0a60b6032f7a395312b8874b927f6c6a3aec1595..ba9c5364de7c43d55140acfff92587f682f7c260 100644 --- a/examples/ASR/LibriSpeech-Test-Clean/data-00000-of-00001.arrow +++ b/examples/ASR/LibriSpeech-Test-Clean/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:e93782657d117940f18822e813788c26fdd0bfd6b77a2f9881419d7638d8989b -size 596184 +oid sha256:e050bcd936107ea58a083a64fc36c85afc55b1f570d8285e5eb93b4244ca3702 +size 491728 diff --git a/examples/ASR/LibriSpeech-Test-Clean/sample_0.wav b/examples/ASR/LibriSpeech-Test-Clean/sample_0.wav index 7f30ab438a700b7e52038787753e5e93d1b4de39..051a14c05f7c270da3d842024c5936075cb5c2e6 100644 Binary files a/examples/ASR/LibriSpeech-Test-Clean/sample_0.wav and b/examples/ASR/LibriSpeech-Test-Clean/sample_0.wav differ diff --git a/examples/ASR/LibriSpeech-Test-Clean/sample_1.wav b/examples/ASR/LibriSpeech-Test-Clean/sample_1.wav index e040ea7d8261366c44b0e9a7ddaca8f8c8044621..30816d1c205dd136109c6abfca19abf249813c68 100644 Binary files a/examples/ASR/LibriSpeech-Test-Clean/sample_1.wav and b/examples/ASR/LibriSpeech-Test-Clean/sample_1.wav differ diff --git a/examples/ASR/LibriSpeech-Test-Clean/sample_2.wav b/examples/ASR/LibriSpeech-Test-Clean/sample_2.wav index 23229886de2e3f6fb8823fe797326027011f3f09..a228ce116181b6b19b741cd9ffc2e1853704adc5 100644 Binary files a/examples/ASR/LibriSpeech-Test-Clean/sample_2.wav and b/examples/ASR/LibriSpeech-Test-Clean/sample_2.wav differ diff --git a/examples/ASR/LibriSpeech-Test-Clean/state.json b/examples/ASR/LibriSpeech-Test-Clean/state.json index e31a1168b4cff86848674d4197e432b9899e0739..dda0f24f40bffbb5dbb1f236bc44f6e715655e67 100644 --- a/examples/ASR/LibriSpeech-Test-Clean/state.json +++ b/examples/ASR/LibriSpeech-Test-Clean/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "db499491d573fb1e", + "_fingerprint": "d9f5d173c305ae96", "_format_columns": [ "context", "instruction", diff --git a/examples/ASR/LibriSpeech-Test-Other/data-00000-of-00001.arrow b/examples/ASR/LibriSpeech-Test-Other/data-00000-of-00001.arrow index 22348d85c6dc04b7ae2c7d9644f901fab3fac0da..f962e9f689dbb890c217156e0c6cbaddb2780490 100644 --- a/examples/ASR/LibriSpeech-Test-Other/data-00000-of-00001.arrow +++ b/examples/ASR/LibriSpeech-Test-Other/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:8b71c4a29c49853f34ca0421302b6375976523b4e9d8d79d682ecb887c3a15c5 -size 370296 +oid sha256:2375da90470a7502211721ccf088ed76808ef698133028c7a74e27606d6f3949 +size 1025488 diff --git a/examples/ASR/LibriSpeech-Test-Other/sample_0.wav b/examples/ASR/LibriSpeech-Test-Other/sample_0.wav index 1343a8bf6091855e6ca7569ef96e2064b3cb69c8..c274d02bd3392db09ea1a95bb050a248627b91cc 100644 Binary files a/examples/ASR/LibriSpeech-Test-Other/sample_0.wav and b/examples/ASR/LibriSpeech-Test-Other/sample_0.wav differ diff --git a/examples/ASR/LibriSpeech-Test-Other/sample_1.wav b/examples/ASR/LibriSpeech-Test-Other/sample_1.wav index 108ec3d17493f7321e6ba6553c94dc62fa96d24b..eea20cd08457ea3ae2d55e91c7240602ae30436f 100644 Binary files a/examples/ASR/LibriSpeech-Test-Other/sample_1.wav and b/examples/ASR/LibriSpeech-Test-Other/sample_1.wav differ diff --git a/examples/ASR/LibriSpeech-Test-Other/sample_2.wav b/examples/ASR/LibriSpeech-Test-Other/sample_2.wav index c5737246f32031af2551b79359c9e5704e49df1f..b7cc7a7ab0cb2cb02a5253dcecede16c688acf86 100644 Binary files a/examples/ASR/LibriSpeech-Test-Other/sample_2.wav and b/examples/ASR/LibriSpeech-Test-Other/sample_2.wav differ diff --git a/examples/ASR/LibriSpeech-Test-Other/state.json b/examples/ASR/LibriSpeech-Test-Other/state.json index c73a8c0f165fb6c789d2533084a09b301b9e0d1a..952838bee518f052b1de767f99c58c1282b17596 100644 --- a/examples/ASR/LibriSpeech-Test-Other/state.json +++ b/examples/ASR/LibriSpeech-Test-Other/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "e751a89ce4227535", + "_fingerprint": "8bd0648dc412be04", "_format_columns": [ "context", "instruction", diff --git a/examples/ASR/Peoples-Speech-Test/data-00000-of-00001.arrow b/examples/ASR/Peoples-Speech-Test/data-00000-of-00001.arrow index 3fa30aaaca61262d740c3a85cd1c1c226a13ffac..900558873c787e1b6e8cecf5934a8b61fa5237e4 100644 --- a/examples/ASR/Peoples-Speech-Test/data-00000-of-00001.arrow +++ b/examples/ASR/Peoples-Speech-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:ab8837799c2ce18ad1ed493d66f3a6c6709a62e4f721c66a4797e7b0b18a59d6 -size 771768 +oid sha256:4299dbcf7bfc7b223f4a653fda3176a8d3e24c09581a41a806b86fc360527bb4 +size 439504 diff --git a/examples/ASR/Peoples-Speech-Test/sample_0.wav b/examples/ASR/Peoples-Speech-Test/sample_0.wav index 0534e049487d4a5e44133719598cecc5fe9fc23a..ff25cea2d23ec61f9199873fc67227221c1bebca 100644 Binary files a/examples/ASR/Peoples-Speech-Test/sample_0.wav and b/examples/ASR/Peoples-Speech-Test/sample_0.wav differ diff --git a/examples/ASR/Peoples-Speech-Test/sample_1.wav b/examples/ASR/Peoples-Speech-Test/sample_1.wav index 5b34837973bb7e663e05d6f623e1957baebdc905..4e7ddfc1977129dfb02d6c4424b362bcd60c1a15 100644 Binary files a/examples/ASR/Peoples-Speech-Test/sample_1.wav and b/examples/ASR/Peoples-Speech-Test/sample_1.wav differ diff --git a/examples/ASR/Peoples-Speech-Test/sample_2.wav b/examples/ASR/Peoples-Speech-Test/sample_2.wav index beca31e7418576a9fc43d747be077a95e2353450..9f590a810351e28a60b5da87f6821f06f8916790 100644 Binary files a/examples/ASR/Peoples-Speech-Test/sample_2.wav and b/examples/ASR/Peoples-Speech-Test/sample_2.wav differ diff --git a/examples/ASR/Peoples-Speech-Test/state.json b/examples/ASR/Peoples-Speech-Test/state.json index 632eca89f3c4fea70a453e57a3349323c699797f..60db649d02fbab6497da719d1c536be91f9bda39 100644 --- a/examples/ASR/Peoples-Speech-Test/state.json +++ b/examples/ASR/Peoples-Speech-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "d07f97bfff42b092", + "_fingerprint": "de704174c1b2e1ea", "_format_columns": [ "context", "instruction", diff --git a/examples/ASR/Tedlium3-Long-form-Test/data-00000-of-00001.arrow b/examples/ASR/Tedlium3-Long-form-Test/data-00000-of-00001.arrow index 4324f464b21fbc18be3be8656fdf1ab78ca2c0a0..9ab6d2f53dcf630417c03a739060cdb58f7afdb4 100644 --- a/examples/ASR/Tedlium3-Long-form-Test/data-00000-of-00001.arrow +++ b/examples/ASR/Tedlium3-Long-form-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:3f92226f17083de77f86be2c69ad7bb1adb2b80b861a6c62b9cab904911a2f73 -size 72244960 +oid sha256:c292f812d20458d9589f8a14933e7b031b1b35e5a496449f4913c3af6d066bc8 +size 98923056 diff --git a/examples/ASR/Tedlium3-Long-form-Test/dataset_info.json b/examples/ASR/Tedlium3-Long-form-Test/dataset_info.json index cf9965f85a972806830fe14f38cc250366f94118..55009f0312ac6d6605288017abbf50e0bafefdc3 100644 --- a/examples/ASR/Tedlium3-Long-form-Test/dataset_info.json +++ b/examples/ASR/Tedlium3-Long-form-Test/dataset_info.json @@ -49,114 +49,6 @@ "dtype": "string", "_type": "Value" } - }, - "salmonn_7b": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "wavllm_fairseq": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "Qwen2-Audio-7B-Instruct": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "whisper_large_v3_with_llama_3_8b_instruct": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "mowe_audio": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "qwen_audio_chat": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } } }, "homepage": "", diff --git a/examples/ASR/Tedlium3-Long-form-Test/sample_0.wav b/examples/ASR/Tedlium3-Long-form-Test/sample_0.wav index 458550d1f8ba2439a49bbdd3acce6d28a8050798..41dbac609aefb991db732192675b67164a3106d5 100644 --- a/examples/ASR/Tedlium3-Long-form-Test/sample_0.wav +++ b/examples/ASR/Tedlium3-Long-form-Test/sample_0.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:df373978edb883e7c38cdb99f293ab770f71a63bc44214185751e229d53edc4c -size 31435404 +oid sha256:ac85b08c75fad06d968942b517e53495961ff861c6e794b576ecce3b406bcbf8 +size 51095404 diff --git a/examples/ASR/Tedlium3-Long-form-Test/sample_1.wav b/examples/ASR/Tedlium3-Long-form-Test/sample_1.wav index ab9204d8220ac39f77a16c6caa4a5c2205b5299d..17eccb9b542ec47bf29066074f7435f6b06f42cd 100644 --- a/examples/ASR/Tedlium3-Long-form-Test/sample_1.wav +++ b/examples/ASR/Tedlium3-Long-form-Test/sample_1.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:2e6bb9d4e01b093db26d317deb1578c337d56d5672dd8aa5fcb1e231752d222e -size 10910764 +oid sha256:0361b7dc4a9437a698a1d28fb3e73718810a6d685f7932f2a04997ad0475b182 +size 36730348 diff --git a/examples/ASR/Tedlium3-Long-form-Test/sample_2.wav b/examples/ASR/Tedlium3-Long-form-Test/sample_2.wav index a218d274caefb6e72dcdf1ef14639eca3d92c54c..7aaa0adc7dcdd7bc19c70c6295b943e8be3931f1 100644 --- a/examples/ASR/Tedlium3-Long-form-Test/sample_2.wav +++ b/examples/ASR/Tedlium3-Long-form-Test/sample_2.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:bb2d7db63f67644b5a8883727929c9f2e96631c45900321734aa020589ade4c5 -size 29474026 +oid sha256:b8cb8f81de8eecb106d072fb26bd87e749fd9fcd28e5cc06949ddb2c542eff3c +size 11046766 diff --git a/examples/ASR/Tedlium3-Long-form-Test/state.json b/examples/ASR/Tedlium3-Long-form-Test/state.json index 30eac41b1222be656086486a4bb4030c8b8e7c9a..3a56b9cee5a756794d32d1b5f03c01a501c41606 100644 --- a/examples/ASR/Tedlium3-Long-form-Test/state.json +++ b/examples/ASR/Tedlium3-Long-form-Test/state.json @@ -4,18 +4,12 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "ffcb019ec304c5cd", + "_fingerprint": "f95b9bf4e3dea7c1", "_format_columns": [ "context", "instruction", "answer", - "other_attributes", - "salmonn_7b", - "wavllm_fairseq", - "Qwen2-Audio-7B-Instruct", - "whisper_large_v3_with_llama_3_8b_instruct", - "mowe_audio", - "qwen_audio_chat" + "other_attributes" ], "_format_kwargs": {}, "_format_type": null, diff --git a/examples/ASR/Tedlium3-Test/data-00000-of-00001.arrow b/examples/ASR/Tedlium3-Test/data-00000-of-00001.arrow index fb7519b6ff82eb9c5b6f9f1b6f6b0dc9c8928134..043a6aa5f5647a465c24325fac5bd8ab7056a64c 100644 --- a/examples/ASR/Tedlium3-Test/data-00000-of-00001.arrow +++ b/examples/ASR/Tedlium3-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:b64d4bab0a8dc52391da9c28e8a5dea1a34ec1c9d19bc2bf4a8f385e064e1b34 -size 632992 +oid sha256:335b82cbd0d97571df4e6b52cb0d3b81f1410f019212964143d4a6c20333f3cc +size 907976 diff --git a/examples/ASR/Tedlium3-Test/dataset_info.json b/examples/ASR/Tedlium3-Test/dataset_info.json index cf9965f85a972806830fe14f38cc250366f94118..55009f0312ac6d6605288017abbf50e0bafefdc3 100644 --- a/examples/ASR/Tedlium3-Test/dataset_info.json +++ b/examples/ASR/Tedlium3-Test/dataset_info.json @@ -49,114 +49,6 @@ "dtype": "string", "_type": "Value" } - }, - "salmonn_7b": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "wavllm_fairseq": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "Qwen2-Audio-7B-Instruct": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "whisper_large_v3_with_llama_3_8b_instruct": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "mowe_audio": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } - }, - "qwen_audio_chat": { - "answer": { - "dtype": "string", - "_type": "Value" - }, - "model_prediction": { - "dtype": "string", - "_type": "Value" - }, - "task_type": { - "dtype": "string", - "_type": "Value" - }, - "text": { - "dtype": "string", - "_type": "Value" - } } }, "homepage": "", diff --git a/examples/ASR/Tedlium3-Test/sample_0.wav b/examples/ASR/Tedlium3-Test/sample_0.wav index 7f9c12706b47e8985cfae1d8e3a33bbb2b93351e..a07fc005b1f77a01b066c0ef962b04e634f4c356 100644 Binary files a/examples/ASR/Tedlium3-Test/sample_0.wav and b/examples/ASR/Tedlium3-Test/sample_0.wav differ diff --git a/examples/ASR/Tedlium3-Test/sample_1.wav b/examples/ASR/Tedlium3-Test/sample_1.wav index f5bdad33f96732f6068f5c0ae1ccfd191e658f14..f864baa998ef015b529fc69d8bccca6f284233f1 100644 Binary files a/examples/ASR/Tedlium3-Test/sample_1.wav and b/examples/ASR/Tedlium3-Test/sample_1.wav differ diff --git a/examples/ASR/Tedlium3-Test/sample_2.wav b/examples/ASR/Tedlium3-Test/sample_2.wav index 650c17709acccb1026eb3f3926caddac87575039..66c6482678614fbd0c658553c4c727a50d80c57a 100644 Binary files a/examples/ASR/Tedlium3-Test/sample_2.wav and b/examples/ASR/Tedlium3-Test/sample_2.wav differ diff --git a/examples/ASR/Tedlium3-Test/state.json b/examples/ASR/Tedlium3-Test/state.json index c31686ed45f77e866be12eb356f308537d01d8c5..ffb37e795661eaf0f656a4272372d0919a492fe0 100644 --- a/examples/ASR/Tedlium3-Test/state.json +++ b/examples/ASR/Tedlium3-Test/state.json @@ -4,18 +4,12 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "6fde2d47e7fcba36", + "_fingerprint": "fb20b90d5641df89", "_format_columns": [ "context", "instruction", "answer", - "other_attributes", - "salmonn_7b", - "wavllm_fairseq", - "Qwen2-Audio-7B-Instruct", - "whisper_large_v3_with_llama_3_8b_instruct", - "mowe_audio", - "qwen_audio_chat" + "other_attributes" ], "_format_kwargs": {}, "_format_type": null, diff --git a/examples/CNASR/Aishell-ASR-ZH-Test/data-00000-of-00001.arrow b/examples/CNASR/Aishell-ASR-ZH-Test/data-00000-of-00001.arrow index 6bf936c095bfb1eb3293538d0acefe477feed450..e3c04ad74452482acd4a29e1a7407a5fc23cbf57 100644 --- a/examples/CNASR/Aishell-ASR-ZH-Test/data-00000-of-00001.arrow +++ b/examples/CNASR/Aishell-ASR-ZH-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:b25cccd3dcce75374e2cf606587eb8a52ebcee8ef33372c0f4445635d507716b -size 439888 +oid sha256:66386a59d99f85838ad5d64c7f7b3108c15a22782f61783ec2debb208d7a2f8d +size 569936 diff --git a/examples/CNASR/Aishell-ASR-ZH-Test/sample_0.wav b/examples/CNASR/Aishell-ASR-ZH-Test/sample_0.wav index a9896d7452058a68f5fb36a098fecb80a61f0179..a69d64c8284caa8ca7ef3f5ecaf6ebc0519020ef 100644 Binary files a/examples/CNASR/Aishell-ASR-ZH-Test/sample_0.wav and b/examples/CNASR/Aishell-ASR-ZH-Test/sample_0.wav differ diff --git a/examples/CNASR/Aishell-ASR-ZH-Test/sample_1.wav b/examples/CNASR/Aishell-ASR-ZH-Test/sample_1.wav index 2b463dad5a36bdd326d83dde600f460252e97213..a149a1ab68e19b0029225518d217608c573d37e2 100644 Binary files a/examples/CNASR/Aishell-ASR-ZH-Test/sample_1.wav and b/examples/CNASR/Aishell-ASR-ZH-Test/sample_1.wav differ diff --git a/examples/CNASR/Aishell-ASR-ZH-Test/sample_2.wav b/examples/CNASR/Aishell-ASR-ZH-Test/sample_2.wav index 1e797f2f358af46173549971074f7b393c6ea266..b2816f1cb90418b611227d1e6d044e886712b426 100644 Binary files a/examples/CNASR/Aishell-ASR-ZH-Test/sample_2.wav and b/examples/CNASR/Aishell-ASR-ZH-Test/sample_2.wav differ diff --git a/examples/CNASR/Aishell-ASR-ZH-Test/state.json b/examples/CNASR/Aishell-ASR-ZH-Test/state.json index 928046ee4e58edb3f83df389b509f6c73998aa2f..4495f512dec2e1549747a3cd9e31137e0afb8081 100644 --- a/examples/CNASR/Aishell-ASR-ZH-Test/state.json +++ b/examples/CNASR/Aishell-ASR-ZH-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "2f95a38020869f6f", + "_fingerprint": "f9833c929864587b", "_format_columns": [ "context", "instruction", diff --git a/examples/ER/IEMOCAP-Emotion-Test/data-00000-of-00001.arrow b/examples/ER/IEMOCAP-Emotion-Test/data-00000-of-00001.arrow index 062110ba749c3131163d5bbf8e773c3d87a25cca..c2bcf1e6ce6f0a0047dd85cce0bbf11e0184b823 100644 --- a/examples/ER/IEMOCAP-Emotion-Test/data-00000-of-00001.arrow +++ b/examples/ER/IEMOCAP-Emotion-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:38544c34a727f77602f7b05d60f63d16a8678ce6c26a8293e2ca0e04e874dd60 -size 238576 +oid sha256:f2d440f3076e388bbeeec29027bd4549d005f3a8bbb342bdc9a77802c79f01cf +size 838160 diff --git a/examples/ER/IEMOCAP-Emotion-Test/sample_0.wav b/examples/ER/IEMOCAP-Emotion-Test/sample_0.wav index f7fbcec204ce9cad676607ba5d3ecc1164ff7cd7..3aea288a199cf828777f07051ce17bb65dd122b9 100644 Binary files a/examples/ER/IEMOCAP-Emotion-Test/sample_0.wav and b/examples/ER/IEMOCAP-Emotion-Test/sample_0.wav differ diff --git a/examples/ER/IEMOCAP-Emotion-Test/sample_1.wav b/examples/ER/IEMOCAP-Emotion-Test/sample_1.wav index 1a821ca8f4d90e16681d927d1bad127bbc131232..f20cf5efc4a86d62d733d80fc2cde556ea107245 100644 Binary files a/examples/ER/IEMOCAP-Emotion-Test/sample_1.wav and b/examples/ER/IEMOCAP-Emotion-Test/sample_1.wav differ diff --git a/examples/ER/IEMOCAP-Emotion-Test/sample_2.wav b/examples/ER/IEMOCAP-Emotion-Test/sample_2.wav index 9c332fd914253d28f3a7b4851e34ce672c7fb4da..6085d420403bb54190cde8d1cffef75b35f2fa88 100644 Binary files a/examples/ER/IEMOCAP-Emotion-Test/sample_2.wav and b/examples/ER/IEMOCAP-Emotion-Test/sample_2.wav differ diff --git a/examples/ER/IEMOCAP-Emotion-Test/state.json b/examples/ER/IEMOCAP-Emotion-Test/state.json index 4cd3ac71d16a1ccd9aefa29c3c207cb5e00e0dec..f9e210ccb738232f4a7ce004649cc0811b5622e7 100644 --- a/examples/ER/IEMOCAP-Emotion-Test/state.json +++ b/examples/ER/IEMOCAP-Emotion-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "f660f47d60092a28", + "_fingerprint": "78bf80b897adbddb", "_format_columns": [ "context", "instruction", diff --git a/examples/ER/MELD-Emotion-Test/data-00000-of-00001.arrow b/examples/ER/MELD-Emotion-Test/data-00000-of-00001.arrow index 7477d8b3e73b402f8b49db6155cc748a1e7df4ae..5c264b825c805731d035e34124efed546055316f 100644 --- a/examples/ER/MELD-Emotion-Test/data-00000-of-00001.arrow +++ b/examples/ER/MELD-Emotion-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:bdf055e7afa7a21b2cca5c997a59aaf4fc427c03420eef7d85cdd2fe17ec8c77 -size 198760 +oid sha256:0fe36d4394fb74d9d92909224e84e0de896cabb5fc94c878b23c3380594a65bb +size 352408 diff --git a/examples/ER/MELD-Emotion-Test/sample_0.wav b/examples/ER/MELD-Emotion-Test/sample_0.wav index 66b28d69b04ddcdbf6649355a50b7431e068b6c1..819fccfa77653af1d839db36a4d89d6c5073676d 100644 Binary files a/examples/ER/MELD-Emotion-Test/sample_0.wav and b/examples/ER/MELD-Emotion-Test/sample_0.wav differ diff --git a/examples/ER/MELD-Emotion-Test/sample_1.wav b/examples/ER/MELD-Emotion-Test/sample_1.wav index 8b65e8917f9c613b9ace25b5cdf2213b1cc5885c..a21acc5a4a0831f75c28e76a93e0339f98a5dab9 100644 Binary files a/examples/ER/MELD-Emotion-Test/sample_1.wav and b/examples/ER/MELD-Emotion-Test/sample_1.wav differ diff --git a/examples/ER/MELD-Emotion-Test/sample_2.wav b/examples/ER/MELD-Emotion-Test/sample_2.wav index b5212fef08bd6083b353d3de4ca02e557b0895ff..98c1a4ec43768374663eb5cae1305480110d87e8 100644 Binary files a/examples/ER/MELD-Emotion-Test/sample_2.wav and b/examples/ER/MELD-Emotion-Test/sample_2.wav differ diff --git a/examples/ER/MELD-Emotion-Test/state.json b/examples/ER/MELD-Emotion-Test/state.json index 4cd00e9f41fc5acfa195cacd42f4b6d869bcfdba..9c219f8e85ce34e78d9b81244ed1bd84435ebb4c 100644 --- a/examples/ER/MELD-Emotion-Test/state.json +++ b/examples/ER/MELD-Emotion-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "8ed652b090e9b45d", + "_fingerprint": "a5a596edab97a213", "_format_columns": [ "context", "instruction", diff --git a/examples/ER/MELD-Sentiment-Test/data-00000-of-00001.arrow b/examples/ER/MELD-Sentiment-Test/data-00000-of-00001.arrow index 29705d4c5436cd9b56a8476658b2c7932f246712..0e0726e555e52b2e486741a8cf021542e7c2dd11 100644 --- a/examples/ER/MELD-Sentiment-Test/data-00000-of-00001.arrow +++ b/examples/ER/MELD-Sentiment-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:4b93d1b416da4092bad2429888c6eb00ae22b57501e60b07207dbe65e6982aea -size 340088 +oid sha256:2388eb71f8718170fe81707866ea2fb68a654153cb6c8a4f503de2e4ef13804d +size 232392 diff --git a/examples/ER/MELD-Sentiment-Test/sample_0.wav b/examples/ER/MELD-Sentiment-Test/sample_0.wav index 70c2320d56f8b078431ee6b90da30c838aa0a3b5..025684f18597120ad16569fb77cc1700b855f6c9 100644 Binary files a/examples/ER/MELD-Sentiment-Test/sample_0.wav and b/examples/ER/MELD-Sentiment-Test/sample_0.wav differ diff --git a/examples/ER/MELD-Sentiment-Test/sample_1.wav b/examples/ER/MELD-Sentiment-Test/sample_1.wav index 9030fb4ce73e38966b2ec587b110d811c1cf4adc..8d64fff1bab98e316eec6c61842e48f7b8f333c4 100644 Binary files a/examples/ER/MELD-Sentiment-Test/sample_1.wav and b/examples/ER/MELD-Sentiment-Test/sample_1.wav differ diff --git a/examples/ER/MELD-Sentiment-Test/sample_2.wav b/examples/ER/MELD-Sentiment-Test/sample_2.wav index 0e25c2ff0f76cf0fbd9a24f091b7317d8f577fa2..f639c22e1b1e24a5d6db354da2e8cf424cdb45be 100644 Binary files a/examples/ER/MELD-Sentiment-Test/sample_2.wav and b/examples/ER/MELD-Sentiment-Test/sample_2.wav differ diff --git a/examples/ER/MELD-Sentiment-Test/state.json b/examples/ER/MELD-Sentiment-Test/state.json index 0732040b40ce0a76abeca37ca83769b790adaebd..917ccc849fc69d74df055b821dd46d31b29e2e8d 100644 --- a/examples/ER/MELD-Sentiment-Test/state.json +++ b/examples/ER/MELD-Sentiment-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "12b7f85ef427fcec", + "_fingerprint": "a052e830551840d2", "_format_columns": [ "context", "instruction", diff --git a/examples/GR/IEMOCAP-Gender-Test/data-00000-of-00001.arrow b/examples/GR/IEMOCAP-Gender-Test/data-00000-of-00001.arrow index 5922d2abe38daae0b1abb32ac1fbb64d7b2bad6a..55669f769aac298829eac258c09015ed16c75897 100644 --- a/examples/GR/IEMOCAP-Gender-Test/data-00000-of-00001.arrow +++ b/examples/GR/IEMOCAP-Gender-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:1093f00f63eae3e24ad463b6bf00b89c38856da4357438daf1c7a76664e26241 -size 278488 +oid sha256:89d1237586d95f75b99aaea0bd47680ad91deaaa3a4c20d94571e76bbc1e951f +size 411576 diff --git a/examples/GR/IEMOCAP-Gender-Test/sample_0.wav b/examples/GR/IEMOCAP-Gender-Test/sample_0.wav index b14c47968a74f9c8bfda57605b321677db3a951b..0f29a85f3b63eb74da8cd026aab5aa13498e0125 100644 Binary files a/examples/GR/IEMOCAP-Gender-Test/sample_0.wav and b/examples/GR/IEMOCAP-Gender-Test/sample_0.wav differ diff --git a/examples/GR/IEMOCAP-Gender-Test/sample_1.wav b/examples/GR/IEMOCAP-Gender-Test/sample_1.wav index f14fe7b4e7bebf1ba0ea4671bd54061c9aa690af..56cec7eeb6836d092e76201787aa22b9436c13f2 100644 Binary files a/examples/GR/IEMOCAP-Gender-Test/sample_1.wav and b/examples/GR/IEMOCAP-Gender-Test/sample_1.wav differ diff --git a/examples/GR/IEMOCAP-Gender-Test/sample_2.wav b/examples/GR/IEMOCAP-Gender-Test/sample_2.wav index d4e718fbd94d944b8dedc542c4032539ebcd5bf7..5c68747f7620a99d23ad13f8d2fd7386ed49332c 100644 Binary files a/examples/GR/IEMOCAP-Gender-Test/sample_2.wav and b/examples/GR/IEMOCAP-Gender-Test/sample_2.wav differ diff --git a/examples/GR/IEMOCAP-Gender-Test/state.json b/examples/GR/IEMOCAP-Gender-Test/state.json index b556f362aef4e6bfad3e7f0959553733a304f250..909b8a4a39ef78e0e286d8b51244d38d68e3aa31 100644 --- a/examples/GR/IEMOCAP-Gender-Test/state.json +++ b/examples/GR/IEMOCAP-Gender-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "42f5f6e863c92512", + "_fingerprint": "7a4eb80e3f03a3f4", "_format_columns": [ "context", "instruction", diff --git a/examples/GR/VoxCeleb-Gender-Test/data-00000-of-00001.arrow b/examples/GR/VoxCeleb-Gender-Test/data-00000-of-00001.arrow index 282312f78eeed1e6512a7294cbf3b650494b6daf..c522d812d46f2f810770b551d21077850da7c6a2 100644 --- a/examples/GR/VoxCeleb-Gender-Test/data-00000-of-00001.arrow +++ b/examples/GR/VoxCeleb-Gender-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:9ee13e0e68a7c38a6cce347de459ef20fed0b9bf18244338a5ccf5284d20cb58 -size 974208 +oid sha256:f7b5fb1ac8505719dd9952b830c5d064d837a406bf932641a9bdc5de448d9c6c +size 609480 diff --git a/examples/GR/VoxCeleb-Gender-Test/sample_0.wav b/examples/GR/VoxCeleb-Gender-Test/sample_0.wav index f834b4c7fa0bbc42f9ae00985df35f41be1eba1e..2f6e333859e59c14c99e08c0ed07ec34da06b7cc 100644 Binary files a/examples/GR/VoxCeleb-Gender-Test/sample_0.wav and b/examples/GR/VoxCeleb-Gender-Test/sample_0.wav differ diff --git a/examples/GR/VoxCeleb-Gender-Test/sample_1.wav b/examples/GR/VoxCeleb-Gender-Test/sample_1.wav index 318509e9dd39ef9205be46c24307cffb31a89523..cdc77dfd8a54374fbf21b32128129fde94d44c2c 100644 Binary files a/examples/GR/VoxCeleb-Gender-Test/sample_1.wav and b/examples/GR/VoxCeleb-Gender-Test/sample_1.wav differ diff --git a/examples/GR/VoxCeleb-Gender-Test/sample_2.wav b/examples/GR/VoxCeleb-Gender-Test/sample_2.wav index 887e042ebd793961f567457f41b6a9ed19107ba1..8b3634c13fc5cefe14475fc673f9458aa7c79815 100644 Binary files a/examples/GR/VoxCeleb-Gender-Test/sample_2.wav and b/examples/GR/VoxCeleb-Gender-Test/sample_2.wav differ diff --git a/examples/GR/VoxCeleb-Gender-Test/state.json b/examples/GR/VoxCeleb-Gender-Test/state.json index 3fd1c757fa6496421e73ae77b5da8ed09efa38f7..d0591ec49e94a4c129e72d00c70af869214dafd8 100644 --- a/examples/GR/VoxCeleb-Gender-Test/state.json +++ b/examples/GR/VoxCeleb-Gender-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "7951265b5c594ce6", + "_fingerprint": "29964e6c779e5e22", "_format_columns": [ "context", "instruction", diff --git a/examples/SI/ALPACA-Audio-Test/data-00000-of-00001.arrow b/examples/SI/ALPACA-Audio-Test/data-00000-of-00001.arrow index 380ab512432af740564becea400b7281d875c085..0e9f779ec899f578e1d5291ff7b261efe4f4d32c 100644 --- a/examples/SI/ALPACA-Audio-Test/data-00000-of-00001.arrow +++ b/examples/SI/ALPACA-Audio-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:5439d754dc3456c1f300de2522e24201b43d958307ddfab1847bf4857d3d2426 -size 332672 +oid sha256:ba310d4adba762052395b3be53afa08451034a8845e8f9f244fa92c84f7358c6 +size 540072 diff --git a/examples/SI/ALPACA-Audio-Test/sample_0.wav b/examples/SI/ALPACA-Audio-Test/sample_0.wav index 1d3e06ea09dc44a7e851372270ae926ca09161ca..70248c9fd2ff3b5792099536311d05bed53fba47 100644 Binary files a/examples/SI/ALPACA-Audio-Test/sample_0.wav and b/examples/SI/ALPACA-Audio-Test/sample_0.wav differ diff --git a/examples/SI/ALPACA-Audio-Test/sample_1.wav b/examples/SI/ALPACA-Audio-Test/sample_1.wav index 63423226a7fea5c7d3cd5cdf708415493411a2ad..0fa85b0e999023c1118dd960236ae84689438330 100644 Binary files a/examples/SI/ALPACA-Audio-Test/sample_1.wav and b/examples/SI/ALPACA-Audio-Test/sample_1.wav differ diff --git a/examples/SI/ALPACA-Audio-Test/sample_2.wav b/examples/SI/ALPACA-Audio-Test/sample_2.wav index aee526d870dcfea114cbe77a58f2b906db3a1846..b5f187db3678a198b23df8b683f8492ae5ff60ac 100644 Binary files a/examples/SI/ALPACA-Audio-Test/sample_2.wav and b/examples/SI/ALPACA-Audio-Test/sample_2.wav differ diff --git a/examples/SI/ALPACA-Audio-Test/state.json b/examples/SI/ALPACA-Audio-Test/state.json index ec1bcb6f73e7f696b90634ca1416212deb70497c..6ba2848fa1bbfd0cf127c67ce6315770b72d590d 100644 --- a/examples/SI/ALPACA-Audio-Test/state.json +++ b/examples/SI/ALPACA-Audio-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "60ec3d739f1f264e", + "_fingerprint": "f46fe3d489641513", "_format_columns": [ "context", "instruction", diff --git a/examples/SI/OpenHermes-Audio-Test/data-00000-of-00001.arrow b/examples/SI/OpenHermes-Audio-Test/data-00000-of-00001.arrow index f002597651f2e6d341418f0cbe62db53e6b23b5b..f78a87653eaed972a87688964c3465cafe9d1cdd 100644 --- a/examples/SI/OpenHermes-Audio-Test/data-00000-of-00001.arrow +++ b/examples/SI/OpenHermes-Audio-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:289cb61cb341b3f3b6eddc92172d0d9eba6e97185d0be9ef84b528b03f03997f -size 601552 +oid sha256:13ec7bd9b5a3ee4bd5a05803c327343b6e461e8ec08413484fd77c43cd6b26bc +size 435240 diff --git a/examples/SI/OpenHermes-Audio-Test/dataset_info.json b/examples/SI/OpenHermes-Audio-Test/dataset_info.json index 454bcaf8f0186d26601a0107af54dc2e906e4091..ac515de4ff190d0b7bd0ce417dfbe2b34d6c0bb8 100644 --- a/examples/SI/OpenHermes-Audio-Test/dataset_info.json +++ b/examples/SI/OpenHermes-Audio-Test/dataset_info.json @@ -49,6 +49,138 @@ "dtype": "int64", "_type": "Value" } + }, + "salmonn_7b": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "audio_text_instruction": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "wavllm_fairseq": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "audio_text_instruction": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "Qwen2-Audio-7B-Instruct": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "audio_text_instruction": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "whisper_large_v3_with_llama_3_8b_instruct": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "audio_text_instruction": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "mowe_audio": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "audio_text_instruction": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "qwen_audio_chat": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "audio_text_instruction": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } } }, "homepage": "", diff --git a/examples/SI/OpenHermes-Audio-Test/sample_0.wav b/examples/SI/OpenHermes-Audio-Test/sample_0.wav index b848eaa0197df53f0c66a042dd2cf0907a18adb6..add65880600b56384cc5350e64c30be65d0ec19d 100644 Binary files a/examples/SI/OpenHermes-Audio-Test/sample_0.wav and b/examples/SI/OpenHermes-Audio-Test/sample_0.wav differ diff --git a/examples/SI/OpenHermes-Audio-Test/sample_1.wav b/examples/SI/OpenHermes-Audio-Test/sample_1.wav index 8990c2638a876f50eabb711a3c65e07809a7aa48..8fa3cc5ffd5e8304b28e2c00f15edfda78c293b2 100644 Binary files a/examples/SI/OpenHermes-Audio-Test/sample_1.wav and b/examples/SI/OpenHermes-Audio-Test/sample_1.wav differ diff --git a/examples/SI/OpenHermes-Audio-Test/sample_2.wav b/examples/SI/OpenHermes-Audio-Test/sample_2.wav index f2d8572e54eeabe52f6e37c3d285d5b1a39060e6..47dc6714c50f2650cb528c81edf0f18a8f3148e9 100644 Binary files a/examples/SI/OpenHermes-Audio-Test/sample_2.wav and b/examples/SI/OpenHermes-Audio-Test/sample_2.wav differ diff --git a/examples/SI/OpenHermes-Audio-Test/state.json b/examples/SI/OpenHermes-Audio-Test/state.json index 3e62882afeb4c4672ca7b438517834bcc61c2771..9a19e3c159f16d99ee14394ad0aab31b7594a7eb 100644 --- a/examples/SI/OpenHermes-Audio-Test/state.json +++ b/examples/SI/OpenHermes-Audio-Test/state.json @@ -4,12 +4,18 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "19108060d5e74733", + "_fingerprint": "bec0fd435c621121", "_format_columns": [ "context", "instruction", "answer", - "other_attributes" + "other_attributes", + "salmonn_7b", + "wavllm_fairseq", + "Qwen2-Audio-7B-Instruct", + "whisper_large_v3_with_llama_3_8b_instruct", + "mowe_audio", + "qwen_audio_chat" ], "_format_kwargs": {}, "_format_type": null, diff --git a/examples/SQA/CN-College-Listen-MCQ-Test/data-00000-of-00001.arrow b/examples/SQA/CN-College-Listen-MCQ-Test/data-00000-of-00001.arrow index 2fbe265869db5d39cad968dcf77a10c462c5cdaa..5b514d7eea4a68baf4f63ab12cce2774d09fdc8e 100644 --- a/examples/SQA/CN-College-Listen-MCQ-Test/data-00000-of-00001.arrow +++ b/examples/SQA/CN-College-Listen-MCQ-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:58aa77c86eb5cbca6ca3dd14c7042ce9a0dcb136e262ef19a1c37e64fb81d0d8 -size 2071440 +oid sha256:4e3578b2ed58ee0fa7527ef31e35664a28c27505b0f9ff79754680b1d7330c70 +size 3518760 diff --git a/examples/SQA/CN-College-Listen-MCQ-Test/dataset_info.json b/examples/SQA/CN-College-Listen-MCQ-Test/dataset_info.json index 81903da2dc3bf5b849827427d5a61f932fe485bb..623b82f8d7d18ff8870b56c298783f1b5f32dc57 100644 --- a/examples/SQA/CN-College-Listen-MCQ-Test/dataset_info.json +++ b/examples/SQA/CN-College-Listen-MCQ-Test/dataset_info.json @@ -45,6 +45,114 @@ "dtype": "string", "_type": "Value" } + }, + "salmonn_7b": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "wavllm_fairseq": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "Qwen2-Audio-7B-Instruct": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "whisper_large_v3_with_llama_3_8b_instruct": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "mowe_audio": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "qwen_audio_chat": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } } }, "homepage": "", diff --git a/examples/SQA/CN-College-Listen-MCQ-Test/sample_0.wav b/examples/SQA/CN-College-Listen-MCQ-Test/sample_0.wav index 2c71842e0c47f529cde47eadbdc21dac098eca8d..cf00703fe4fb0275a04b27b4ab0abf44a8e39cae 100644 Binary files a/examples/SQA/CN-College-Listen-MCQ-Test/sample_0.wav and b/examples/SQA/CN-College-Listen-MCQ-Test/sample_0.wav differ diff --git a/examples/SQA/CN-College-Listen-MCQ-Test/sample_1.wav b/examples/SQA/CN-College-Listen-MCQ-Test/sample_1.wav index 02faddd4fba70d832d7b726a181b922e974bd1ac..8aea2bff27874d78fbee581ad8849d0ab0ac9fc8 100644 Binary files a/examples/SQA/CN-College-Listen-MCQ-Test/sample_1.wav and b/examples/SQA/CN-College-Listen-MCQ-Test/sample_1.wav differ diff --git a/examples/SQA/CN-College-Listen-MCQ-Test/sample_2.wav b/examples/SQA/CN-College-Listen-MCQ-Test/sample_2.wav index 86ba0d6fedd50d874900f973f86967cdf9c54ab4..750f1b8414060167651ac33128408b510b1545ca 100644 Binary files a/examples/SQA/CN-College-Listen-MCQ-Test/sample_2.wav and b/examples/SQA/CN-College-Listen-MCQ-Test/sample_2.wav differ diff --git a/examples/SQA/CN-College-Listen-MCQ-Test/state.json b/examples/SQA/CN-College-Listen-MCQ-Test/state.json index 023c37fec316be4e2d15a4c442c4e4ec31bc090a..1afd2605e58a9e5304f9f324f51af5e7d535d636 100644 --- a/examples/SQA/CN-College-Listen-MCQ-Test/state.json +++ b/examples/SQA/CN-College-Listen-MCQ-Test/state.json @@ -4,12 +4,18 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "ea9d9086266315b3", + "_fingerprint": "b4fb19374756e22d", "_format_columns": [ "context", "instruction", "answer", - "other_attributes" + "other_attributes", + "salmonn_7b", + "wavllm_fairseq", + "Qwen2-Audio-7B-Instruct", + "whisper_large_v3_with_llama_3_8b_instruct", + "mowe_audio", + "qwen_audio_chat" ], "_format_kwargs": {}, "_format_type": null, diff --git a/examples/SQA/DREAM-TTS-MCQ-Test/data-00000-of-00001.arrow b/examples/SQA/DREAM-TTS-MCQ-Test/data-00000-of-00001.arrow index 49410d0803542cc92ea6de8afdba67e464132b99..236abb453583027bafdc652c39704a6870afaec8 100644 --- a/examples/SQA/DREAM-TTS-MCQ-Test/data-00000-of-00001.arrow +++ b/examples/SQA/DREAM-TTS-MCQ-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:bb30219e949747e062d84726985ef1eefac4816306b5605480b48605216cd989 -size 861032 +oid sha256:683f5edfae8193f93c22fc1bf2514edee75cc54f565b0e997d4addbfc3376a4d +size 1272504 diff --git a/examples/SQA/DREAM-TTS-MCQ-Test/dataset_info.json b/examples/SQA/DREAM-TTS-MCQ-Test/dataset_info.json index a16938258fabe5c46b36019db75cd53e5132c21f..b12863f3596549cd265ea212a5964a91c271a7cf 100644 --- a/examples/SQA/DREAM-TTS-MCQ-Test/dataset_info.json +++ b/examples/SQA/DREAM-TTS-MCQ-Test/dataset_info.json @@ -59,6 +59,96 @@ "dtype": "string", "_type": "Value" } + }, + "salmonn_7b": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "wavllm_fairseq": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "Qwen2-Audio-7B-Instruct": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "whisper_large_v3_with_llama_3_8b_instruct": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "qwen_audio_chat": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } } }, "homepage": "", diff --git a/examples/SQA/DREAM-TTS-MCQ-Test/sample_0.wav b/examples/SQA/DREAM-TTS-MCQ-Test/sample_0.wav index ea2575be156853eb2a649d3ef6fdec18caf6dd40..a7133d2c60825df4c133cf3578b84a0252c4a091 100644 --- a/examples/SQA/DREAM-TTS-MCQ-Test/sample_0.wav +++ b/examples/SQA/DREAM-TTS-MCQ-Test/sample_0.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:8c3e4e7fdd462c0af4725901df139b31aae58db1a39d77a13a2063834b1ecf40 -size 145176 +oid sha256:140e86d457811ef438fbbb13e457224fcd9f98a45451965b7183c335042b2003 +size 220930 diff --git a/examples/SQA/DREAM-TTS-MCQ-Test/sample_1.wav b/examples/SQA/DREAM-TTS-MCQ-Test/sample_1.wav index 701eaadff036dbe4397ac3d3e5cf953cdafe2492..15f7e2c92aa5c8e199277474dfbb81d27c9ef002 100644 Binary files a/examples/SQA/DREAM-TTS-MCQ-Test/sample_1.wav and b/examples/SQA/DREAM-TTS-MCQ-Test/sample_1.wav differ diff --git a/examples/SQA/DREAM-TTS-MCQ-Test/sample_2.wav b/examples/SQA/DREAM-TTS-MCQ-Test/sample_2.wav index 4313f84ddf1d93ecdf915b9c48103757d653a138..c30502c27386e06b63b6cf6e6297088178a63535 100644 --- a/examples/SQA/DREAM-TTS-MCQ-Test/sample_2.wav +++ b/examples/SQA/DREAM-TTS-MCQ-Test/sample_2.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:e7b53b0b1d82ee825549c32d96d13aa2ebc6a2540fc769affe57ffdfacff5054 -size 271448 +oid sha256:5a339d7324c19ba4228bb9f1de7b947e52cf4591b856a48f08e2c0d2d54a79cd +size 142424 diff --git a/examples/SQA/DREAM-TTS-MCQ-Test/state.json b/examples/SQA/DREAM-TTS-MCQ-Test/state.json index 60f24b2e2e0fc0f619402aedaba80fe1a6102aa6..9429aff4cfdd4360de95ee6eaf6f11a40d6ceefa 100644 --- a/examples/SQA/DREAM-TTS-MCQ-Test/state.json +++ b/examples/SQA/DREAM-TTS-MCQ-Test/state.json @@ -4,12 +4,17 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "8d4dbaf0bbc46c99", + "_fingerprint": "4ae1a389c9652fd2", "_format_columns": [ "answer", "context", "instruction", - "other_attributes" + "other_attributes", + "salmonn_7b", + "wavllm_fairseq", + "Qwen2-Audio-7B-Instruct", + "whisper_large_v3_with_llama_3_8b_instruct", + "qwen_audio_chat" ], "_format_kwargs": {}, "_format_type": null, diff --git a/examples/SQA/Public-SG-Speech-QA-Test/data-00000-of-00001.arrow b/examples/SQA/Public-SG-Speech-QA-Test/data-00000-of-00001.arrow index 3ebcbfbb43d127586bfe91ceaf520a40e423f78e..d1c7960aac65d29f4df237c5245a2aa6c89d0055 100644 --- a/examples/SQA/Public-SG-Speech-QA-Test/data-00000-of-00001.arrow +++ b/examples/SQA/Public-SG-Speech-QA-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:38c3d8b65a56d0e8748b65eb5b187100a0bfea74dda7440d50a99a4153259491 -size 3575960 +oid sha256:41ebee1aeb1c9a28136ca99e86afcdf9f583aba6c003a0a99d94477419dc9f6f +size 3555008 diff --git a/examples/SQA/Public-SG-Speech-QA-Test/sample_0.wav b/examples/SQA/Public-SG-Speech-QA-Test/sample_0.wav index d08a61ee00f5e16fe27d2f62a66936d12515907c..298380997737d4c67d07f69122cc44a28437254d 100644 --- a/examples/SQA/Public-SG-Speech-QA-Test/sample_0.wav +++ b/examples/SQA/Public-SG-Speech-QA-Test/sample_0.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:c54aeaab1fcc652d147417d6f54dab49f7236f7447e801b1f1949143cc4bc4b4 -size 976732 +oid sha256:b62a055ccc0592fb34892fc14c743745bd3ee5ff04475f8231ca064cf5ab292c +size 1599216 diff --git a/examples/SQA/Public-SG-Speech-QA-Test/sample_1.wav b/examples/SQA/Public-SG-Speech-QA-Test/sample_1.wav index 174105b272715751fdbd71ace6ae135f9c114e6d..90ec1058956b790eaba74d366ad8aeb3ebffcfeb 100644 --- a/examples/SQA/Public-SG-Speech-QA-Test/sample_1.wav +++ b/examples/SQA/Public-SG-Speech-QA-Test/sample_1.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:997d15e66b5b9a2d0886412968bd964437a32c952f653ecf0691b66db414dfc3 -size 1527468 +oid sha256:f125d54b780e025d73756918a11f7bbe9badf35feef0a220bdde91c275d154e2 +size 950336 diff --git a/examples/SQA/Public-SG-Speech-QA-Test/sample_2.wav b/examples/SQA/Public-SG-Speech-QA-Test/sample_2.wav index f0b054dabe0e86ff8c6419e192e5c1c0ad3bbcc7..35e65e20490020e270e104069f6937c3909dc0a3 100644 --- a/examples/SQA/Public-SG-Speech-QA-Test/sample_2.wav +++ b/examples/SQA/Public-SG-Speech-QA-Test/sample_2.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:f7862a175d879b81ed5d821a1fb318067f7fbd8bf6a17946433d9399d4f5fe8e -size 1051126 +oid sha256:e682e04d2afa9162c4b0545f05177e076e5a6fc6a09314ced5097c49bf55082e +size 984132 diff --git a/examples/SQA/Public-SG-Speech-QA-Test/state.json b/examples/SQA/Public-SG-Speech-QA-Test/state.json index 966e1f2dfe5fba436d6ddb0873b91c39ee0f9530..0f2668023e6a567ada41c9ea082f99773f2e8f6e 100644 --- a/examples/SQA/Public-SG-Speech-QA-Test/state.json +++ b/examples/SQA/Public-SG-Speech-QA-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "4d4e665c9f359042", + "_fingerprint": "ede505a635b66631", "_format_columns": [ "context", "instruction", diff --git a/examples/SQA/SLUE-P2-SQA5-Test/data-00000-of-00001.arrow b/examples/SQA/SLUE-P2-SQA5-Test/data-00000-of-00001.arrow index d6b18e783b7fde676c103e4b328e9cda9c61f26a..fdc22e998f5d035a4a21db496fc26bb1eb9e7b47 100644 --- a/examples/SQA/SLUE-P2-SQA5-Test/data-00000-of-00001.arrow +++ b/examples/SQA/SLUE-P2-SQA5-Test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:85348a8af5baece22d64c73e700befc3060b01cde643f81b857659f8a7e50f15 -size 5423960 +oid sha256:91fad195c759891c783f8448a7cb42362b53f091aacad15195a7d262da356464 +size 5469256 diff --git a/examples/SQA/SLUE-P2-SQA5-Test/sample_0.wav b/examples/SQA/SLUE-P2-SQA5-Test/sample_0.wav index ff803716c4be62ae71d10056857bab18ab1120d6..7236fa9aac8e0e10818025ab656ddc4ad2bbf812 100644 --- a/examples/SQA/SLUE-P2-SQA5-Test/sample_0.wav +++ b/examples/SQA/SLUE-P2-SQA5-Test/sample_0.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:beb911cb8aa54e5e2ca0f350fef0bb168f9c342be68933e40be88564d30ec4be +oid sha256:398b25e597d18e0e027bb7fda9cc025cff93b716bc683d88a553d34c338285eb size 1280044 diff --git a/examples/SQA/SLUE-P2-SQA5-Test/sample_1.wav b/examples/SQA/SLUE-P2-SQA5-Test/sample_1.wav index 8c354a510ff6872d3626769840ef4477cf60b885..7a80d4c3399ae9256db0ccded497c8fe90aecb5c 100644 --- a/examples/SQA/SLUE-P2-SQA5-Test/sample_1.wav +++ b/examples/SQA/SLUE-P2-SQA5-Test/sample_1.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:3417c4542f84b6c803a2b26bb92f6fc45ca340826de87e6f40ebe65a6026d546 +oid sha256:36769a3e93db8fc6ef00b05552a3c36b08c35c578f93d142a60415d2909f9e7e size 1280044 diff --git a/examples/SQA/SLUE-P2-SQA5-Test/sample_2.wav b/examples/SQA/SLUE-P2-SQA5-Test/sample_2.wav index 31ccc491b64594b509c54c5e58eb5270afacc5e0..92f61a3d6c5de05f4b6070eb892bf22139beeff5 100644 --- a/examples/SQA/SLUE-P2-SQA5-Test/sample_2.wav +++ b/examples/SQA/SLUE-P2-SQA5-Test/sample_2.wav @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:6e9852047dde397382f62b51eab24b14d2ada02089f198d354836580316a7d5a +oid sha256:43e31031a1e0e8cde85470353830b7c4513d4d9f9e5ca394127b80b4938d970a size 1280044 diff --git a/examples/SQA/SLUE-P2-SQA5-Test/state.json b/examples/SQA/SLUE-P2-SQA5-Test/state.json index e79576fcdc5892cb2e18d1920422d90537058f9b..64fcf1074f14e6cc130ef25406cd0718f61798a7 100644 --- a/examples/SQA/SLUE-P2-SQA5-Test/state.json +++ b/examples/SQA/SLUE-P2-SQA5-Test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "0bc180e1898c34b5", + "_fingerprint": "400b504ce3034854", "_format_columns": [ "context", "instruction", diff --git a/examples/SQA/Spoken-Squad-Test/data-00000-of-00001.arrow b/examples/SQA/Spoken-Squad-Test/data-00000-of-00001.arrow new file mode 100644 index 0000000000000000000000000000000000000000..01c3308e636c1ca7b513cc8e863074d83085e4d3 --- /dev/null +++ b/examples/SQA/Spoken-Squad-Test/data-00000-of-00001.arrow @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a8c870f847afa404da100216dd6192919e7e00218c046beecf577d752e958e14 +size 6854040 diff --git a/examples/SQA/Spoken-Squad-Test/dataset_info.json b/examples/SQA/Spoken-Squad-Test/dataset_info.json new file mode 100644 index 0000000000000000000000000000000000000000..5b50e517bf0760f04194c6b7c720d733151674aa --- /dev/null +++ b/examples/SQA/Spoken-Squad-Test/dataset_info.json @@ -0,0 +1,172 @@ +{ + "citation": "", + "description": "", + "features": { + "context": { + "text": { + "dtype": "string", + "_type": "Value" + }, + "audio": { + "sampling_rate": 16000, + "_type": "Audio" + } + }, + "instruction": { + "audio": { + "dtype": "null", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "answer": { + "audio": { + "dtype": "null", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "other_attributes": { + "answer_length": { + "dtype": "int64", + "_type": "Value" + }, + "context_length": { + "dtype": "int64", + "_type": "Value" + }, + "instruction_length": { + "dtype": "int64", + "_type": "Value" + }, + "paragraph_id": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + }, + "topic_id": { + "dtype": "string", + "_type": "Value" + } + }, + "salmonn_7b": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "wavllm_fairseq": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "Qwen2-Audio-7B-Instruct": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "whisper_large_v3_with_llama_3_8b_instruct": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "mowe_audio": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + }, + "qwen_audio_chat": { + "answer": { + "dtype": "string", + "_type": "Value" + }, + "model_prediction": { + "dtype": "string", + "_type": "Value" + }, + "task_type": { + "dtype": "string", + "_type": "Value" + }, + "text": { + "dtype": "string", + "_type": "Value" + } + } + }, + "homepage": "", + "license": "" +} \ No newline at end of file diff --git a/examples/SQA/Spoken-Squad-Test/sample_0.wav b/examples/SQA/Spoken-Squad-Test/sample_0.wav new file mode 100644 index 0000000000000000000000000000000000000000..2b928fafe992bf5c3ecf4b03389d186cff11421b --- /dev/null +++ b/examples/SQA/Spoken-Squad-Test/sample_0.wav @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:304b38f43745d431d607788b839a14409b8e8e2032a093f408302d79287efc3b +size 2143532 diff --git a/examples/SQA/Spoken-Squad-Test/sample_1.wav b/examples/SQA/Spoken-Squad-Test/sample_1.wav new file mode 100644 index 0000000000000000000000000000000000000000..10820a48e3ffd1a5ecb2198457b3891122d161a1 --- /dev/null +++ b/examples/SQA/Spoken-Squad-Test/sample_1.wav @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:97c29b943c6244c1a6dee8bbd871055408ed9735349c3fd57f2d285cd3628041 +size 1789484 diff --git a/examples/SQA/Spoken-Squad-Test/sample_2.wav b/examples/SQA/Spoken-Squad-Test/sample_2.wav new file mode 100644 index 0000000000000000000000000000000000000000..82953be741c78b78f28232b6b13d2392ff533c2e --- /dev/null +++ b/examples/SQA/Spoken-Squad-Test/sample_2.wav @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:26c3bc8688fbd99e8dd6eff774a7994ae34ec6fddfb1242f3bbae500a12b7746 +size 2903852 diff --git a/examples/SQA/Spoken-Squad-Test/state.json b/examples/SQA/Spoken-Squad-Test/state.json new file mode 100644 index 0000000000000000000000000000000000000000..bad03ce21b0c5b24ac005094f5a25236804a54a7 --- /dev/null +++ b/examples/SQA/Spoken-Squad-Test/state.json @@ -0,0 +1,24 @@ +{ + "_data_files": [ + { + "filename": "data-00000-of-00001.arrow" + } + ], + "_fingerprint": "416952584cf805a4", + "_format_columns": [ + "context", + "instruction", + "answer", + "other_attributes", + "salmonn_7b", + "wavllm_fairseq", + "Qwen2-Audio-7B-Instruct", + "whisper_large_v3_with_llama_3_8b_instruct", + "mowe_audio", + "qwen_audio_chat" + ], + "_format_kwargs": {}, + "_format_type": null, + "_output_all_columns": false, + "_split": null +} \ No newline at end of file diff --git a/examples/ST/Covost2-EN-ID-test/data-00000-of-00001.arrow b/examples/ST/Covost2-EN-ID-test/data-00000-of-00001.arrow index c4a4ca3f5d252fbe12cd82446baf89e62495ffd3..35d28390925a96fd109f6c96b0109cbe4c18a228 100644 --- a/examples/ST/Covost2-EN-ID-test/data-00000-of-00001.arrow +++ b/examples/ST/Covost2-EN-ID-test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:1e23b1c4ab3763956724963caa814330f80c2ef29e36823ca5f79a5254d54f2a -size 528648 +oid sha256:e114114e362e4b40cbffbfe8f9857f64de0cac09471062c6f6c4f6e73b235de7 +size 549296 diff --git a/examples/ST/Covost2-EN-ID-test/sample_0.wav b/examples/ST/Covost2-EN-ID-test/sample_0.wav index aaa73891ca80e6daf935a96362b99724bf30e5f2..5e70047fc638745caac9c90c2539cc9f18168f8a 100644 Binary files a/examples/ST/Covost2-EN-ID-test/sample_0.wav and b/examples/ST/Covost2-EN-ID-test/sample_0.wav differ diff --git a/examples/ST/Covost2-EN-ID-test/sample_1.wav b/examples/ST/Covost2-EN-ID-test/sample_1.wav index 8bb713755533c9c9c40fd7bc6fb97ceccc71b5c1..832716522492f3cbeb066ed58c45eab2e9036a2f 100644 Binary files a/examples/ST/Covost2-EN-ID-test/sample_1.wav and b/examples/ST/Covost2-EN-ID-test/sample_1.wav differ diff --git a/examples/ST/Covost2-EN-ID-test/sample_2.wav b/examples/ST/Covost2-EN-ID-test/sample_2.wav index e82baf335e4809fb09b4fd7447b6c96cad4e289f..2ccb333e9339d1fda08f7c84fe4fe2cd0b22508c 100644 Binary files a/examples/ST/Covost2-EN-ID-test/sample_2.wav and b/examples/ST/Covost2-EN-ID-test/sample_2.wav differ diff --git a/examples/ST/Covost2-EN-ID-test/state.json b/examples/ST/Covost2-EN-ID-test/state.json index d35329d7cd163f8aeae7496f3cfa3ebdc423220f..d460d81ba05ec305339ae4a50c65c2654e706950 100644 --- a/examples/ST/Covost2-EN-ID-test/state.json +++ b/examples/ST/Covost2-EN-ID-test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "b3bfe1ac88ff72a0", + "_fingerprint": "17a5c97a84a7f33c", "_format_columns": [ "context", "instruction", diff --git a/examples/ST/Covost2-EN-TA-test/data-00000-of-00001.arrow b/examples/ST/Covost2-EN-TA-test/data-00000-of-00001.arrow index 170a3850f19f31f770e1922e4d937e3115186f12..50796f046accb5692f9772e9beeacf58537e9a1d 100644 --- a/examples/ST/Covost2-EN-TA-test/data-00000-of-00001.arrow +++ b/examples/ST/Covost2-EN-TA-test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:b9263a598f39b02d2754fa7aeb160fe0c618d9964065ae55a5728f1e03c8a959 -size 705320 +oid sha256:6a1667eba0de4088fbe20e9f6c9beaacf5beb556a1d817c92af178757da3bc78 +size 561528 diff --git a/examples/ST/Covost2-EN-TA-test/sample_0.wav b/examples/ST/Covost2-EN-TA-test/sample_0.wav index 0c62dd1aab9919d9bcd776efbe484e001b592598..ad0f73bf6dbfe42e17635b063161d6154cfcf28a 100644 Binary files a/examples/ST/Covost2-EN-TA-test/sample_0.wav and b/examples/ST/Covost2-EN-TA-test/sample_0.wav differ diff --git a/examples/ST/Covost2-EN-TA-test/sample_1.wav b/examples/ST/Covost2-EN-TA-test/sample_1.wav index 4246c85fb21814573dbb736d188584a42e63f29f..e5d11995c4bd1e3484f0c15d828c678300d899b2 100644 Binary files a/examples/ST/Covost2-EN-TA-test/sample_1.wav and b/examples/ST/Covost2-EN-TA-test/sample_1.wav differ diff --git a/examples/ST/Covost2-EN-TA-test/sample_2.wav b/examples/ST/Covost2-EN-TA-test/sample_2.wav index d339cd3dc96f925f24a1f7501e4645c8844fb369..8910c6f24aebdc2aab0e2517f304448129282655 100644 Binary files a/examples/ST/Covost2-EN-TA-test/sample_2.wav and b/examples/ST/Covost2-EN-TA-test/sample_2.wav differ diff --git a/examples/ST/Covost2-EN-TA-test/state.json b/examples/ST/Covost2-EN-TA-test/state.json index 8ca1326d120d940e69fbea2e66a9174a5d88f864..de82636ce42828a45d6c8c4ca1edde9874d87cc3 100644 --- a/examples/ST/Covost2-EN-TA-test/state.json +++ b/examples/ST/Covost2-EN-TA-test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "76e8db43e251f03a", + "_fingerprint": "e9d273226522711f", "_format_columns": [ "context", "instruction", diff --git a/examples/ST/Covost2-EN-ZH-test/data-00000-of-00001.arrow b/examples/ST/Covost2-EN-ZH-test/data-00000-of-00001.arrow index 53dc4d701961d58f2a1fc4d4cb722474b171fc6b..02b18146d4be540926d80ef57b9bd914ef51aec2 100644 --- a/examples/ST/Covost2-EN-ZH-test/data-00000-of-00001.arrow +++ b/examples/ST/Covost2-EN-ZH-test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:b458b896008a8b9df6ce5d55e057eb07c9b3527ae949dea86900ca25b13f1999 -size 509640 +oid sha256:f4e2a6c95fe3b8adbc4c2f55bd53fdb56b064d436c49ee20ac2e4333114ceac7 +size 600520 diff --git a/examples/ST/Covost2-EN-ZH-test/sample_0.wav b/examples/ST/Covost2-EN-ZH-test/sample_0.wav index a5ace1d21eb77c249ac8bf9f268b8f1bfec5519a..f428967146cecccb88b318b8388e3b897cd14fca 100644 Binary files a/examples/ST/Covost2-EN-ZH-test/sample_0.wav and b/examples/ST/Covost2-EN-ZH-test/sample_0.wav differ diff --git a/examples/ST/Covost2-EN-ZH-test/sample_1.wav b/examples/ST/Covost2-EN-ZH-test/sample_1.wav index 8c40d7e137c6f005a0b3fd087f5b68aff125adc7..e91b4565eb72ab11f760c9b276daf4f6998f9e5f 100644 Binary files a/examples/ST/Covost2-EN-ZH-test/sample_1.wav and b/examples/ST/Covost2-EN-ZH-test/sample_1.wav differ diff --git a/examples/ST/Covost2-EN-ZH-test/sample_2.wav b/examples/ST/Covost2-EN-ZH-test/sample_2.wav index b1ce2588bbd2cd289ea7b7d3996c95a904eeb06d..38f60afc485334971149ebeb3985573093dd9a96 100644 Binary files a/examples/ST/Covost2-EN-ZH-test/sample_2.wav and b/examples/ST/Covost2-EN-ZH-test/sample_2.wav differ diff --git a/examples/ST/Covost2-EN-ZH-test/state.json b/examples/ST/Covost2-EN-ZH-test/state.json index cf87654cc0e72ceb65b8e71d7428d648a2f1549b..7dafba85d0dbd927e3c9aae50f0d1a3cf3e2c71c 100644 --- a/examples/ST/Covost2-EN-ZH-test/state.json +++ b/examples/ST/Covost2-EN-ZH-test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "d27fe19e989510df", + "_fingerprint": "86243bad639f0cb6", "_format_columns": [ "context", "instruction", diff --git a/examples/ST/Covost2-ID-EN-test/data-00000-of-00001.arrow b/examples/ST/Covost2-ID-EN-test/data-00000-of-00001.arrow index e9a50e6f9d741de06e05189522e6db19dde9d9df..037f35bf9787de62b458d9a89969d0abcd8b02eb 100644 --- a/examples/ST/Covost2-ID-EN-test/data-00000-of-00001.arrow +++ b/examples/ST/Covost2-ID-EN-test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:ede28759781fab3253daed83adb1e2dfe4e004098eff002f29f916f1f663ce8d -size 406960 +oid sha256:1785e6d316adfd52a99ce27e0b14fd68fa410a93f5082fab34e8f35a89563d8c +size 529872 diff --git a/examples/ST/Covost2-ID-EN-test/sample_0.wav b/examples/ST/Covost2-ID-EN-test/sample_0.wav index c5fc387f794ec6644cd2c2a04cbced7d49950f5f..60cf3a26a30e046d87c72a4e4d4f15cf54732039 100644 Binary files a/examples/ST/Covost2-ID-EN-test/sample_0.wav and b/examples/ST/Covost2-ID-EN-test/sample_0.wav differ diff --git a/examples/ST/Covost2-ID-EN-test/sample_1.wav b/examples/ST/Covost2-ID-EN-test/sample_1.wav index 89cea5e89bc92eb93b885643c1e250e7a987a73c..0b37d62b0b132a59e5ce6d2c4551c1d701143efb 100644 Binary files a/examples/ST/Covost2-ID-EN-test/sample_1.wav and b/examples/ST/Covost2-ID-EN-test/sample_1.wav differ diff --git a/examples/ST/Covost2-ID-EN-test/sample_2.wav b/examples/ST/Covost2-ID-EN-test/sample_2.wav index bbcdcfd796d393d9d227a57fe543b8e442b1d94f..5eda0ef756def2b097ee2bef2a92b8b9e99ff5c3 100644 Binary files a/examples/ST/Covost2-ID-EN-test/sample_2.wav and b/examples/ST/Covost2-ID-EN-test/sample_2.wav differ diff --git a/examples/ST/Covost2-ID-EN-test/state.json b/examples/ST/Covost2-ID-EN-test/state.json index 060cb29ce89ace31ec246bf22af9ae83459382e3..592b67b79c5171bdccfcf07d28855f756e7f4f23 100644 --- a/examples/ST/Covost2-ID-EN-test/state.json +++ b/examples/ST/Covost2-ID-EN-test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "1ad122cc3b0e20fe", + "_fingerprint": "69b492f6dd79179e", "_format_columns": [ "context", "instruction", diff --git a/examples/ST/Covost2-TA-EN-test/data-00000-of-00001.arrow b/examples/ST/Covost2-TA-EN-test/data-00000-of-00001.arrow index f6f47478010a025a50fbc67426126876b923eeef..d4966d98c3fec664dfe65f4a2094d48ef716ac78 100644 --- a/examples/ST/Covost2-TA-EN-test/data-00000-of-00001.arrow +++ b/examples/ST/Covost2-TA-EN-test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:570d32be1c6c8a30c171fc39a9a7360956afe2f8902adeafa295a2f460a0f2b7 -size 451680 +oid sha256:2f3b8c1a000b5085ab5af45333fd8c4fc37508c74595f1dba8668f7da20b4d21 +size 477584 diff --git a/examples/ST/Covost2-TA-EN-test/sample_0.wav b/examples/ST/Covost2-TA-EN-test/sample_0.wav index 070bd162c47080814b4afe6d0ff0c1f629926ff7..642a3386437533748a4df9cbf7c45ab6f5bd0b6f 100644 Binary files a/examples/ST/Covost2-TA-EN-test/sample_0.wav and b/examples/ST/Covost2-TA-EN-test/sample_0.wav differ diff --git a/examples/ST/Covost2-TA-EN-test/sample_1.wav b/examples/ST/Covost2-TA-EN-test/sample_1.wav index c5ad72bcc3bdcbdc2b2b9a13b92c0e78ea2f6e95..a9c003707ecdee7e8918014b2f77a69adbe71d93 100644 Binary files a/examples/ST/Covost2-TA-EN-test/sample_1.wav and b/examples/ST/Covost2-TA-EN-test/sample_1.wav differ diff --git a/examples/ST/Covost2-TA-EN-test/sample_2.wav b/examples/ST/Covost2-TA-EN-test/sample_2.wav index ccf5eda02b57d943a995a0504dde21135376ff21..cd0ce42ab1336cc1591caa5cd56acf6db9afbc29 100644 Binary files a/examples/ST/Covost2-TA-EN-test/sample_2.wav and b/examples/ST/Covost2-TA-EN-test/sample_2.wav differ diff --git a/examples/ST/Covost2-TA-EN-test/state.json b/examples/ST/Covost2-TA-EN-test/state.json index 047e39432ccad934ad34d2ebdf861a7031e4ff96..c7aef0119e13ade230e9d5c50ca9b66df7193400 100644 --- a/examples/ST/Covost2-TA-EN-test/state.json +++ b/examples/ST/Covost2-TA-EN-test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "ec8fc1af5b770f60", + "_fingerprint": "6f095ca26fe268ab", "_format_columns": [ "answer", "context", diff --git a/examples/ST/Covost2-ZH-EN-test/data-00000-of-00001.arrow b/examples/ST/Covost2-ZH-EN-test/data-00000-of-00001.arrow index 50f793b9a37fb5df343a0aca8a42d615cc082709..dc342fcd295e1e1f058737f14d28154c8718cab1 100644 --- a/examples/ST/Covost2-ZH-EN-test/data-00000-of-00001.arrow +++ b/examples/ST/Covost2-ZH-EN-test/data-00000-of-00001.arrow @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:1063ece9b69092484d5cec94cc56f768b0fc16ac542bbbcd5d13c9492265a1f3 -size 647064 +oid sha256:acab5090424e3e32cac285d3a346f48308a64629aa5fba171addc37fbf4f5337 +size 554256 diff --git a/examples/ST/Covost2-ZH-EN-test/sample_0.wav b/examples/ST/Covost2-ZH-EN-test/sample_0.wav index 261eb86cee105db49f22654b63ce14980709edec..a0add517f30ee8b82cef1be3aba2d471645bd648 100644 Binary files a/examples/ST/Covost2-ZH-EN-test/sample_0.wav and b/examples/ST/Covost2-ZH-EN-test/sample_0.wav differ diff --git a/examples/ST/Covost2-ZH-EN-test/sample_1.wav b/examples/ST/Covost2-ZH-EN-test/sample_1.wav index 4bfd72806def045cf386666dcd2ca307040049ec..6fad51ce295d62bb0b68826f1d1f3c3f4a2756e9 100644 Binary files a/examples/ST/Covost2-ZH-EN-test/sample_1.wav and b/examples/ST/Covost2-ZH-EN-test/sample_1.wav differ diff --git a/examples/ST/Covost2-ZH-EN-test/sample_2.wav b/examples/ST/Covost2-ZH-EN-test/sample_2.wav index 1fd592e9eccd9f45016e64157bed001a02da4421..e7bebe13c8efed3bb9b8c9b9a46831c8152240b5 100644 Binary files a/examples/ST/Covost2-ZH-EN-test/sample_2.wav and b/examples/ST/Covost2-ZH-EN-test/sample_2.wav differ diff --git a/examples/ST/Covost2-ZH-EN-test/state.json b/examples/ST/Covost2-ZH-EN-test/state.json index e230ae52dd845127cf654fc006737e7883a1dcc9..4199c3b1fc42c18396a4f54e001d032ad37e6d97 100644 --- a/examples/ST/Covost2-ZH-EN-test/state.json +++ b/examples/ST/Covost2-ZH-EN-test/state.json @@ -4,7 +4,7 @@ "filename": "data-00000-of-00001.arrow" } ], - "_fingerprint": "f049563334d9978e", + "_fingerprint": "977bd2807131826b", "_format_columns": [ "context", "instruction",