{ "schema": 1, "language": "en", "lang_code": "a", "dialect": "us", "voice": "am_echo", "speed": 0.94, "sample_rate": 24000, "total_duration": 145.54, "media_sha256": "6617b76cc0bc72530e2fe4d369aadb462b6c13e55d361450d25f66c71df561de", "media_bytes": 932525, "render_fingerprint": "1d4afd38300105a854cdf8a3788f9ad0e2ae08b28736288c7f953a705deaee0e", "render_inputs": { "renderer": "render_all_voices-v4", "lesson": "10_classify_cv", "language": "en", "lang_code": "a", "dialect": "us", "voice": "am_echo", "speed": 0.94, "scenes": [ { "narration": "Classify CV trains image models on annotated object crops. The console reports dataset checks, training progress, held-out metrics, inference, warnings, and output paths.", "speech_text": [ "[Classify](/klˈæsəfI/) CV trains image models on annotated object crops.", "The console reports dataset checks, training progress, held-out metrics, inference, warnings, and output paths." ], "authored_hold": 0.7, "resolved_hold": 0.52, "speed_multiplier": 1.0, "effective_speed": 0.94 }, { "narration": "Plate Sources and Workflow selects the measured projects and the stages to run: build a split, train, evaluate, or apply an existing checkpoint.", "speech_text": [ "Plate Sources and Workflow selects the measured projects and the stages to run: build a split, train, evaluate, or apply an existing checkpoint." ], "authored_hold": 0.7, "resolved_hold": 0.52, "speed_multiplier": 1.0, "effective_speed": 0.94 }, { "narration": "Start with Essentials, use Modified to audit deliberate changes, and open All only for controls required by the experiment. Save a Recipe when the setup will be repeated.", "speech_text": [ "Start with Essentials, use Modified to audit deliberate changes, and open All only for controls required by the experiment.", "Save a Recipe when the setup will be repeated." ], "authored_hold": 0.7, "resolved_hold": 0.52, "speed_multiplier": 1.0, "effective_speed": 0.94 }, { "narration": "Labels and Classes selects the annotation or metadata field used as ground truth and maps its saved values to documented class names. Verify that mapping before training.", "speech_text": [ "Labels and Classes selects the annotation or metadata field used as ground truth and maps its saved values to documented class names.", "Verify that mapping before training." ], "authored_hold": 0.7, "resolved_hold": 0.52, "speed_multiplier": 1.0, "effective_speed": 0.94 }, { "narration": "Images and Cropping chooses the object images, channels, balancing, and held-out fraction. Group related images by plate or well so near-duplicates cannot leak between training and evaluation.", "speech_text": [ "Images and Cropping chooses the object images, channels, balancing, and held-out fraction.", "Group related images by plate or well so near-duplicates cannot leak between training and evaluation." ], "authored_hold": 0.7, "resolved_hold": 0.52, "speed_multiplier": 1.0, "effective_speed": 0.94 }, { "narration": "Model and Regularization selects the network, input channels and size, pretrained weights, normalization, and regularization expected by the images. Begin with one reasonable baseline.", "speech_text": [ "Model and Regularization selects the network, input channels and size, pre-trained weights, normalization, and regularization expected by the images.", "Begin with one reasonable baseline." ], "authored_hold": 0.7, "resolved_hold": 0.52, "speed_multiplier": 1.0, "effective_speed": 0.94 }, { "narration": "Training and Loss contains learning rate, augmentation, batches, epochs, and early stopping. Change a small number of controls at a time and judge them on validation behavior.", "speech_text": [ "Training and Loss contains learning rate, augmentation, batches, epochs, and early stopping.", "Change a small number of controls at a time and judge them on validation behavior." ], "authored_hold": 0.7, "resolved_hold": 0.52, "speed_multiplier": 1.0, "effective_speed": 0.94 }, { "narration": "Evaluation and Results adds grouped evaluation, calibration, and strict identity and content-leakage checks. Keep final held-out data separate from model selection.", "speech_text": [ "Evaluation and Results adds grouped evaluation, calibration, and strict identity and content-leakage checks.", "Keep final held-out data separate from model selection." ], "authored_hold": 0.7, "resolved_hold": 0.52, "speed_multiplier": 1.0, "effective_speed": 0.94 }, { "narration": "The same Evaluation and Results section controls full-dataset inference, saved probabilities and class calls, and representative examples. Review those outputs before using predictions biologically.", "speech_text": [ "The same Evaluation and Results section controls full-dataset inference, saved probabilities and class calls, and representative examples.", "Review those outputs before using predictions biologically." ], "authored_hold": 0.7, "resolved_hold": 0.52, "speed_multiplier": 1.0, "effective_speed": 0.94 }, { "narration": "Enable Hyperparameter Search when several plausible configurations need comparison. The mini workbench runs grouped trials and displays each score, fold variation, parameters, and status.", "speech_text": [ "Enable hyper-parameter Search when several plausible configurations need comparison.", "The mini workbench runs grouped trials and displays each score, fold variation, parameters, and status." ], "authored_hold": 0.7, "resolved_hold": 0.52, "speed_multiplier": 1.0, "effective_speed": 0.94 }, { "narration": "Search Settings defines a focused candidate space, ranking metric, grouped folds, and trial budget. Avoid broad searches that spend compute on choices the experiment cannot distinguish.", "speech_text": [ "Search Settings defines a focused candidate space, ranking metric, grouped folds, and trial budget.", "Avoid broad searches that spend compute on choices the experiment cannot distinguish." ], "authored_hold": 0.7, "resolved_hold": 0.52, "speed_multiplier": 1.0, "effective_speed": 0.94 }, { "narration": "Run first validates labels and splits, then trains and evaluates the configured model. Review held-out metrics and examples before applying it broadly; spaCR stores the checkpoint, predictions, settings, and run record.", "speech_text": [ "Run first validates labels and splits, then trains and evaluates the configured model.", "Review held-out metrics and examples before applying it broadly; spacer stores the checkpoint, predictions, settings, and run record." ], "authored_hold": 0.7, "resolved_hold": 0.52, "speed_multiplier": 1.0, "effective_speed": 0.94 } ], "assembly": { "version": "2026-08-11-natural-sentences-v1", "sample_rate": 24000, "lead_silence_seconds": 0.1, "tail_silence_seconds": 0.18, "sentence_pause_seconds": 0.12, "scene_hold_min_seconds": 0.32, "scene_hold_max_seconds": 0.52, "default_scene_hold_seconds": 0.42, "vad_frame_seconds": 0.005, "vad_peak_ratio": 0.002, "vad_absolute_floor": 2e-05 }, "cadence_profile": { "version": "2026-08-11-af-heart-scene-cadence-v6", "sha256": "4d728d67e73eea7e6203878879992f9050ae4a2677d717efdd68545104eb7812", "bytes": 23008, "enabled": false, "targets": { "slow_wpm": 125.0, "fast_wpm": 145.0, "calibration_wpm": 150.0 }, "safe_multiplier_bounds": { "minimum": 0.68, "maximum": 1.18 } }, "mastering": { "filters": [ "loudnorm=I=-16:TP=-3:LRA=11", "loudnorm=I=-16:TP=-5:LRA=11,alimiter=limit=0.7:attack=5:release=50:level=disabled", "loudnorm=I=-16:TP=-5:LRA=11,alimiter=limit=0.5:attack=5:release=50:level=disabled" ], "codec": "aac", "bitrate": "48k", "channels": 1, "maximum_decoded_true_peak_dbfs": -1.0 }, "pronunciation": { "role": "pronunciation_profile", "version": "2026-08-11-en-natural-v6", "sha256": "1fbaa615cfa904094bdef5cba963539f0e8bec5927192a76b0d503e90c4bb83e", "bytes": 13263 }, "synthesis_runtime": { "model": { "repository": "hexgrad/Kokoro-82M", "snapshot": "f3ff3571791e39611d31c381e3a41a3af07b4987", "config": { "role": "kokoro_config", "name": "14a726edd3718279eac426630879ff743955b16a", "bytes": 2351, "sha256": "5abb01e2403b072bf03d04fde160443e209d7a0dad49a423be15196b9b43c17f" }, "weights": { "role": "kokoro_weights", "name": "496dba118d1a58f5f3db2efc88dbdc216e0483fc89fe6e47ee1f2c53f18ad1e4", "bytes": 327212226, "sha256": "496dba118d1a58f5f3db2efc88dbdc216e0483fc89fe6e47ee1f2c53f18ad1e4" } }, "voice_pack": { "role": "kokoro_voice_pack", "name": "8bcfdc852bc985fb45c396c561e571ffb9183930071f962f1b50df5c97b161e8", "bytes": 523420, "sha256": "8bcfdc852bc985fb45c396c561e571ffb9183930071f962f1b50df5c97b161e8", "tensor_sha256": "3968b92c3c4cd1c4416dbded36c13eaa388a90d5788d02a13e4d781f5f8cf3c3", "loaded_tensor_sha256": "3968b92c3c4cd1c4416dbded36c13eaa388a90d5788d02a13e4d781f5f8cf3c3" }, "software": { "kokoro": "0.9.4", "misaki": "0.9.4", "torch": "2.6.0", "soundfile": "0.14.0", "python": "3.12.4", "implementation": "cpython" }, "inference": { "device": "cuda" }, "mastering_tool": { "version": "ffmpeg version 6.1.1-3ubuntu5+esm10 Copyright (c) 2000-2023 the FFmpeg developers", "version_output_sha256": "92dfb2468e6f5f06919b5433c0f14459ac19b81b1bc2d9fdceb83c6ad42e37b5", "executable": { "role": "ffmpeg_executable", "name": "ffmpeg", "bytes": 342488, "sha256": "ad63ef3f2f18fa92e89f0ec2bde6941e81f83724c8d265414cb659c28077392d" } } } }, "scenes": [ { "scene": 1, "speech_start": 0.0, "speech_end": 11.69, "scene_end": 12.209999999999999, "duration": 12.209999999999999, "text": "Classify CV trains image models on annotated object crops. The console reports dataset checks, training progress, held-out metrics, inference, warnings, and output paths.", "speech_text": "[Classify](/klˈæsəfI/) CV trains image models on annotated object crops. The console reports dataset checks, training progress, held-out metrics, inference, warnings, and output paths.", "phonemes": "klˈæsəfI sˌivˈi tɹˈAnz ˈɪmɪʤ mˈɑdᵊlz ˌɔn ˈænətˌATᵻd ˈɑbʤəkt kɹˈɑps. ðə kˈɑnsˌOl ɹəpˈɔɹts dˈATəsˌɛt ʧˈɛks, tɹˈAnɪŋ pɹˈɑɡɹəs, hˈɛldˌWt mˈɛtɹɪks, ˈɪnfəɹəns, wˈɔɹnɪŋz, ænd ˈWtpˌʊt pˈæðz.", "authored_hold_after": 0.7, "hold_after": 0.52, "base_speed": 0.94, "speed_multiplier": 1.0, "effective_speed": 0.94, "sentences": [ { "sentence": 1, "speech_start": 0.0, "speech_end": 4.84, "audible_start": 0.1, "audible_end": 4.66, "duration": 4.84, "gap_from_previous": null, "trimmed_lead": 0.175, "trimmed_tail": 0.435, "text": "Classify CV trains image models on annotated object crops.", "speech_text": "[Classify](/klˈæsəfI/) CV trains image models on annotated object crops.", "phonemes": "klˈæsəfI sˌivˈi tɹˈAnz ˈɪmɪʤ mˈɑdᵊlz ˌɔn ˈænətˌATᵻd ˈɑbʤəkt kɹˈɑps." }, { "sentence": 2, "speech_start": 4.96, "speech_end": 11.69, "audible_start": 5.06, "audible_end": 11.51, "duration": 6.73, "gap_from_previous": 0.4, "trimmed_lead": 0.16, "trimmed_tail": 0.535, "text": "The console reports dataset checks, training progress, held-out metrics, inference, warnings, and output paths.", "speech_text": "The console reports dataset checks, training progress, held-out metrics, inference, warnings, and output paths.", "phonemes": "ðə kˈɑnsˌOl ɹəpˈɔɹts dˈATəsˌɛt ʧˈɛks, tɹˈAnɪŋ pɹˈɑɡɹəs, hˈɛldˌWt mˈɛtɹɪks, ˈɪnfəɹəns, wˈɔɹnɪŋz, ænd ˈWtpˌʊt pˈæðz." } ] }, { "scene": 2, "speech_start": 12.209999999999999, "speech_end": 20.799999999999997, "scene_end": 21.319999999999997, "duration": 9.109999999999998, "text": "Plate Sources and Workflow selects the measured projects and the stages to run: build a split, train, evaluate, or apply an existing checkpoint.", "speech_text": "Plate Sources and Workflow selects the measured projects and the stages to run: build a split, train, evaluate, or apply an existing checkpoint.", "phonemes": "plˈAt sˈɔɹsᵻz ænd wˈɜɹkflˌO səlˈɛkts ðə mˈɛʒəɹd pɹˈɑʤˌɛkts ænd ðə stˈAʤᵻz tə ɹˈʌn: bˈɪld ɐ splˈɪt, tɹˈAn, əvˈæljʊˌAt, ɔɹ əplˈI ɐn ɪɡzˈɪstɪŋ ʧˈɛkpˌYnt.", "authored_hold_after": 0.7, "hold_after": 0.52, "base_speed": 0.94, "speed_multiplier": 1.0, "effective_speed": 0.94, "sentences": [ { "sentence": 1, "speech_start": 12.209999999999999, "speech_end": 20.799999999999997, "audible_start": 12.309999999999999, "audible_end": 20.619999999999997, "duration": 8.59, "gap_from_previous": null, "trimmed_lead": 0.165, "trimmed_tail": 0.52, "text": "Plate Sources and Workflow selects the measured projects and the stages to run: build a split, train, evaluate, or apply an existing checkpoint.", "speech_text": "Plate Sources and Workflow selects the measured projects and the stages to run: build a split, train, evaluate, or apply an existing checkpoint.", "phonemes": "plˈAt sˈɔɹsᵻz ænd wˈɜɹkflˌO səlˈɛkts ðə mˈɛʒəɹd pɹˈɑʤˌɛkts ænd ðə stˈAʤᵻz tə ɹˈʌn: bˈɪld ɐ splˈɪt, tɹˈAn, əvˈæljʊˌAt, ɔɹ əplˈI ɐn ɪɡzˈɪstɪŋ ʧˈɛkpˌYnt." } ] }, { "scene": 3, "speech_start": 21.319999999999997, "speech_end": 32.184999999999995, "scene_end": 32.705, "duration": 11.385000000000002, "text": "Start with Essentials, use Modified to audit deliberate changes, and open All only for controls required by the experiment. Save a Recipe when the setup will be repeated.", "speech_text": "Start with Essentials, use Modified to audit deliberate changes, and open All only for controls required by the experiment. Save a Recipe when the setup will be repeated.", "phonemes": "stˈɑɹt wɪð əsˈɛnʧᵊlz, jˈuz mˈɑdəfˌId tʊ ˈɔdət dəlˈɪbəɹət ʧˈAnʤᵻz, ænd ˈOpᵊn ˈɔl ˈOnli fɔɹ kəntɹˈOlz ɹəkwˈIəɹd bI ði ɪkspˈɛɹəmənt. sˈAv ɐ ɹˈɛsəpˌi wˌɛn ðə sˈɛTˌʌp wɪl bi ɹəpˈiTᵻd.", "authored_hold_after": 0.7, "hold_after": 0.52, "base_speed": 0.94, "speed_multiplier": 1.0, "effective_speed": 0.94, "sentences": [ { "sentence": 1, "speech_start": 21.319999999999997, "speech_end": 29.004999999999995, "audible_start": 21.419999999999998, "audible_end": 28.824999999999996, "duration": 7.685, "gap_from_previous": null, "trimmed_lead": 0.145, "trimmed_tail": 0.545, "text": "Start with Essentials, use Modified to audit deliberate changes, and open All only for controls required by the experiment.", "speech_text": "Start with Essentials, use Modified to audit deliberate changes, and open All only for controls required by the experiment.", "phonemes": "stˈɑɹt wɪð əsˈɛnʧᵊlz, jˈuz mˈɑdəfˌId tʊ ˈɔdət dəlˈɪbəɹət ʧˈAnʤᵻz, ænd ˈOpᵊn ˈɔl ˈOnli fɔɹ kəntɹˈOlz ɹəkwˈIəɹd bI ði ɪkspˈɛɹəmənt." }, { "sentence": 2, "speech_start": 29.124999999999996, "speech_end": 32.184999999999995, "audible_start": 29.224999999999998, "audible_end": 32.004999999999995, "duration": 3.06, "gap_from_previous": 0.4, "trimmed_lead": 0.145, "trimmed_tail": 0.445, "text": "Save a Recipe when the setup will be repeated.", "speech_text": "Save a Recipe when the setup will be repeated.", "phonemes": "sˈAv ɐ ɹˈɛsəpˌi wˌɛn ðə sˈɛTˌʌp wɪl bi ɹəpˈiTᵻd." } ] }, { "scene": 4, "speech_start": 32.705, "speech_end": 44.305, "scene_end": 44.825, "duration": 12.120000000000005, "text": "Labels and Classes selects the annotation or metadata field used as ground truth and maps its saved values to documented class names. Verify that mapping before training.", "speech_text": "Labels and Classes selects the annotation or metadata field used as ground truth and maps its saved values to documented class names. Verify that mapping before training.", "phonemes": "lˈAbəlz ænd klˈæsᵻz səlˈɛkts ði ˌænətˈAʃən ɔɹ mˈɛTədˌATə fˈild jˈuzd æz ɡɹˈWnd tɹˈuθ ænd mˈæps ɪts sˈAvd vˈæljuz tə dˈɑkjəmˌɛntᵻd klˈæs nˈAmz. vˈɛɹəfˌI ðˈæt mˈæpɪŋ bəfˈɔɹ tɹˈAnɪŋ.", "authored_hold_after": 0.7, "hold_after": 0.52, "base_speed": 0.94, "speed_multiplier": 1.0, "effective_speed": 0.94, "sentences": [ { "sentence": 1, "speech_start": 32.705, "speech_end": 41.464999999999996, "audible_start": 32.805, "audible_end": 41.285, "duration": 8.76, "gap_from_previous": null, "trimmed_lead": 0.17, "trimmed_tail": 0.545, "text": "Labels and Classes selects the annotation or metadata field used as ground truth and maps its saved values to documented class names.", "speech_text": "Labels and Classes selects the annotation or metadata field used as ground truth and maps its saved values to documented class names.", "phonemes": "lˈAbəlz ænd klˈæsᵻz səlˈɛkts ði ˌænətˈAʃən ɔɹ mˈɛTədˌATə fˈild jˈuzd æz ɡɹˈWnd tɹˈuθ ænd mˈæps ɪts sˈAvd vˈæljuz tə dˈɑkjəmˌɛntᵻd klˈæs nˈAmz." }, { "sentence": 2, "speech_start": 41.585, "speech_end": 44.305, "audible_start": 41.685, "audible_end": 44.125, "duration": 2.72, "gap_from_previous": 0.4, "trimmed_lead": 0.155, "trimmed_tail": 0.4, "text": "Verify that mapping before training.", "speech_text": "Verify that mapping before training.", "phonemes": "vˈɛɹəfˌI ðˈæt mˈæpɪŋ bəfˈɔɹ tɹˈAnɪŋ." } ] }, { "scene": 5, "speech_start": 44.825, "speech_end": 56.940000000000005, "scene_end": 57.46000000000001, "duration": 12.635000000000005, "text": "Images and Cropping chooses the object images, channels, balancing, and held-out fraction. Group related images by plate or well so near-duplicates cannot leak between training and evaluation.", "speech_text": "Images and Cropping chooses the object images, channels, balancing, and held-out fraction. Group related images by plate or well so near-duplicates cannot leak between training and evaluation.", "phonemes": "ˈɪmɪʤᵻz ænd kɹˈɑpɪŋ ʧˈuzᵻz ði ˈɑbʤəkt ˈɪmɪʤᵻz, ʧˈænᵊlz, bˈælənsɪŋ, ænd hˈɛldˌWt fɹˈækʃən. ɡɹˈup ɹəlˈATᵻd ˈɪmɪʤᵻz bI plˈAt ɔɹ wˈɛl sˌO nˌɪɹdˈupləkəts kənˈɑt lˈik bətwˈin tɹˈAnɪŋ ænd əvˌæljʊˈAʃən.", "authored_hold_after": 0.7, "hold_after": 0.52, "base_speed": 0.94, "speed_multiplier": 1.0, "effective_speed": 0.94, "sentences": [ { "sentence": 1, "speech_start": 44.825, "speech_end": 50.395, "audible_start": 44.925000000000004, "audible_end": 50.215, "duration": 5.57, "gap_from_previous": null, "trimmed_lead": 0.165, "trimmed_tail": 0.49, "text": "Images and Cropping chooses the object images, channels, balancing, and held-out fraction.", "speech_text": "Images and Cropping chooses the object images, channels, balancing, and held-out fraction.", "phonemes": "ˈɪmɪʤᵻz ænd kɹˈɑpɪŋ ʧˈuzᵻz ði ˈɑbʤəkt ˈɪmɪʤᵻz, ʧˈænᵊlz, bˈælənsɪŋ, ænd hˈɛldˌWt fɹˈækʃən." }, { "sentence": 2, "speech_start": 50.515, "speech_end": 56.940000000000005, "audible_start": 50.615, "audible_end": 56.760000000000005, "duration": 6.425, "gap_from_previous": 0.4, "trimmed_lead": 0.175, "trimmed_tail": 0.525, "text": "Group related images by plate or well so near-duplicates cannot leak between training and evaluation.", "speech_text": "Group related images by plate or well so near-duplicates cannot leak between training and evaluation.", "phonemes": "ɡɹˈup ɹəlˈATᵻd ˈɪmɪʤᵻz bI plˈAt ɔɹ wˈɛl sˌO nˌɪɹdˈupləkəts kənˈɑt lˈik bətwˈin tɹˈAnɪŋ ænd əvˌæljʊˈAʃən." } ] }, { "scene": 6, "speech_start": 57.46000000000001, "speech_end": 69.76, "scene_end": 70.28, "duration": 12.819999999999993, "text": "Model and Regularization selects the network, input channels and size, pretrained weights, normalization, and regularization expected by the images. Begin with one reasonable baseline.", "speech_text": "Model and Regularization selects the network, input channels and size, pre-trained weights, normalization, and regularization expected by the images. Begin with one reasonable baseline.", "phonemes": "mˈɑdᵊl ænd ɹˌɛɡjələɹəzˈAʃən səlˈɛkts ðə nˈɛtwˌɜɹk, ˈɪnpˌʊt ʧˈænᵊlz ænd sˈIz, pɹˌitɹˈAnd wˈAts, nˌɔɹmələzˈAʃən, ænd ɹˌɛɡjələɹəzˈAʃən ɪkspˈɛktᵻd bI ði ˈɪmɪʤᵻz. bəɡˈɪn wɪð wˈʌn ɹˈizənəbᵊl bˈAslˌIn.", "authored_hold_after": 0.7, "hold_after": 0.52, "base_speed": 0.94, "speed_multiplier": 1.0, "effective_speed": 0.94, "sentences": [ { "sentence": 1, "speech_start": 57.46000000000001, "speech_end": 66.89000000000001, "audible_start": 57.56000000000001, "audible_end": 66.71000000000001, "duration": 9.43, "gap_from_previous": null, "trimmed_lead": 0.175, "trimmed_tail": 0.545, "text": "Model and Regularization selects the network, input channels and size, pretrained weights, normalization, and regularization expected by the images.", "speech_text": "Model and Regularization selects the network, input channels and size, pre-trained weights, normalization, and regularization expected by the images.", "phonemes": "mˈɑdᵊl ænd ɹˌɛɡjələɹəzˈAʃən səlˈɛkts ðə nˈɛtwˌɜɹk, ˈɪnpˌʊt ʧˈænᵊlz ænd sˈIz, pɹˌitɹˈAnd wˈAts, nˌɔɹmələzˈAʃən, ænd ɹˌɛɡjələɹəzˈAʃən ɪkspˈɛktᵻd bI ði ˈɪmɪʤᵻz." }, { "sentence": 2, "speech_start": 67.01, "speech_end": 69.76, "audible_start": 67.11000000000001, "audible_end": 69.58000000000001, "duration": 2.75, "gap_from_previous": 0.4, "trimmed_lead": 0.195, "trimmed_tail": 0.405, "text": "Begin with one reasonable baseline.", "speech_text": "Begin with one reasonable baseline.", "phonemes": "bəɡˈɪn wɪð wˈʌn ɹˈizənəbᵊl bˈAslˌIn." } ] }, { "scene": 7, "speech_start": 70.28, "speech_end": 81.18, "scene_end": 81.7, "duration": 11.420000000000002, "text": "Training and Loss contains learning rate, augmentation, batches, epochs, and early stopping. Change a small number of controls at a time and judge them on validation behavior.", "speech_text": "Training and Loss contains learning rate, augmentation, batches, epochs, and early stopping. Change a small number of controls at a time and judge them on validation behavior.", "phonemes": "tɹˈAnɪŋ ænd lˈɔs kəntˈAnz lˈɜɹnɪŋ ɹˈAt, ˌɔɡmˌɛntˈAʃən, bˈæʧᵻz, ˈɛpəks, ænd ˈɜɹli stˈɑpɪŋ. ʧˈAnʤ ɐ smˈɔl nˈʌmbəɹ ʌv kəntɹˈOlz æt ɐ tˈIm ænd ʤˈʌʤ ðˌɛm ˌɔn vˌælɪdˈAʃən bəhˈAvjəɹ.", "authored_hold_after": 0.7, "hold_after": 0.52, "base_speed": 0.94, "speed_multiplier": 1.0, "effective_speed": 0.94, "sentences": [ { "sentence": 1, "speech_start": 70.28, "speech_end": 75.845, "audible_start": 70.38, "audible_end": 75.665, "duration": 5.565, "gap_from_previous": null, "trimmed_lead": 0.16, "trimmed_tail": 0.5, "text": "Training and Loss contains learning rate, augmentation, batches, epochs, and early stopping.", "speech_text": "Training and Loss contains learning rate, augmentation, batches, epochs, and early stopping.", "phonemes": "tɹˈAnɪŋ ænd lˈɔs kəntˈAnz lˈɜɹnɪŋ ɹˈAt, ˌɔɡmˌɛntˈAʃən, bˈæʧᵻz, ˈɛpəks, ænd ˈɜɹli stˈɑpɪŋ." }, { "sentence": 2, "speech_start": 75.965, "speech_end": 81.18, "audible_start": 76.065, "audible_end": 81.0, "duration": 5.215, "gap_from_previous": 0.4, "trimmed_lead": 0.16, "trimmed_tail": 0.525, "text": "Change a small number of controls at a time and judge them on validation behavior.", "speech_text": "Change a small number of controls at a time and judge them on validation behavior.", "phonemes": "ʧˈAnʤ ɐ smˈɔl nˈʌmbəɹ ʌv kəntɹˈOlz æt ɐ tˈIm ænd ʤˈʌʤ ðˌɛm ˌɔn vˌælɪdˈAʃən bəhˈAvjəɹ." } ] }, { "scene": 8, "speech_start": 81.7, "speech_end": 92.345, "scene_end": 92.865, "duration": 11.164999999999992, "text": "Evaluation and Results adds grouped evaluation, calibration, and strict identity and content-leakage checks. Keep final held-out data separate from model selection.", "speech_text": "Evaluation and Results adds grouped evaluation, calibration, and strict identity and content-leakage checks. Keep final held-out data separate from model selection.", "phonemes": "əvˌæljʊˈAʃən ænd ɹəzˈʌlts ˈædz ɡɹˈupt əvˌæljʊˈAʃən, kˌæləbɹˈAʃən, ænd stɹˈɪkt IdˈɛntəTi ænd kˈɑntɛntlˌikɪʤ ʧˈɛks. kˈip fˈInᵊl hˈɛldˌWt dˈATə sˈɛpəɹət fɹʌm mˈɑdᵊl səlˈɛkʃən.", "authored_hold_after": 0.7, "hold_after": 0.52, "base_speed": 0.94, "speed_multiplier": 1.0, "effective_speed": 0.94, "sentences": [ { "sentence": 1, "speech_start": 81.7, "speech_end": 88.595, "audible_start": 81.8, "audible_end": 88.415, "duration": 6.895, "gap_from_previous": null, "trimmed_lead": 0.17, "trimmed_tail": 0.485, "text": "Evaluation and Results adds grouped evaluation, calibration, and strict identity and content-leakage checks.", "speech_text": "Evaluation and Results adds grouped evaluation, calibration, and strict identity and content-leakage checks.", "phonemes": "əvˌæljʊˈAʃən ænd ɹəzˈʌlts ˈædz ɡɹˈupt əvˌæljʊˈAʃən, kˌæləbɹˈAʃən, ænd stɹˈɪkt IdˈɛntəTi ænd kˈɑntɛntlˌikɪʤ ʧˈɛks." }, { "sentence": 2, "speech_start": 88.715, "speech_end": 92.345, "audible_start": 88.815, "audible_end": 92.165, "duration": 3.63, "gap_from_previous": 0.4, "trimmed_lead": 0.175, "trimmed_tail": 0.445, "text": "Keep final held-out data separate from model selection.", "speech_text": "Keep final held-out data separate from model selection.", "phonemes": "kˈip fˈInᵊl hˈɛldˌWt dˈATə sˈɛpəɹət fɹʌm mˈɑdᵊl səlˈɛkʃən." } ] }, { "scene": 9, "speech_start": 92.865, "speech_end": 105.46499999999999, "scene_end": 105.98499999999999, "duration": 13.11999999999999, "text": "The same Evaluation and Results section controls full-dataset inference, saved probabilities and class calls, and representative examples. Review those outputs before using predictions biologically.", "speech_text": "The same Evaluation and Results section controls full-dataset inference, saved probabilities and class calls, and representative examples. Review those outputs before using predictions biologically.", "phonemes": "ðə sˈAm əvˌæljʊˈAʃən ænd ɹəzˈʌlts sˈɛkʃən kəntɹˈOlz fˌʊldˈATəsˌɛt ˈɪnfəɹəns, sˈAvd pɹˌɑbəbˈɪləTiz ænd klˈæs kˈɔlz, ænd ɹˌɛpɹəzˈɛntəTɪv ɪɡzˈæmpəlz. ɹəvjˈu ðOz ˈWtpˌʊts bəfˈɔɹ jˈuzɪŋ pɹidˈɪkʃənz bˌIəlˈɑʤəkᵊli.", "authored_hold_after": 0.7, "hold_after": 0.52, "base_speed": 0.94, "speed_multiplier": 1.0, "effective_speed": 0.94, "sentences": [ { "sentence": 1, "speech_start": 92.865, "speech_end": 101.375, "audible_start": 92.96499999999999, "audible_end": 101.195, "duration": 8.51, "gap_from_previous": null, "trimmed_lead": 0.17, "trimmed_tail": 0.545, "text": "The same Evaluation and Results section controls full-dataset inference, saved probabilities and class calls, and representative examples.", "speech_text": "The same Evaluation and Results section controls full-dataset inference, saved probabilities and class calls, and representative examples.", "phonemes": "ðə sˈAm əvˌæljʊˈAʃən ænd ɹəzˈʌlts sˈɛkʃən kəntɹˈOlz fˌʊldˈATəsˌɛt ˈɪnfəɹəns, sˈAvd pɹˌɑbəbˈɪləTiz ænd klˈæs kˈɔlz, ænd ɹˌɛpɹəzˈɛntəTɪv ɪɡzˈæmpəlz." }, { "sentence": 2, "speech_start": 101.49499999999999, "speech_end": 105.46499999999999, "audible_start": 101.595, "audible_end": 105.285, "duration": 3.97, "gap_from_previous": 0.4, "trimmed_lead": 0.165, "trimmed_tail": 0.465, "text": "Review those outputs before using predictions biologically.", "speech_text": "Review those outputs before using predictions biologically.", "phonemes": "ɹəvjˈu ðOz ˈWtpˌʊts bəfˈɔɹ jˈuzɪŋ pɹidˈɪkʃənz bˌIəlˈɑʤəkᵊli." } ] }, { "scene": 10, "speech_start": 105.98499999999999, "speech_end": 118.16999999999999, "scene_end": 118.68999999999998, "duration": 12.704999999999998, "text": "Enable Hyperparameter Search when several plausible configurations need comparison. The mini workbench runs grouped trials and displays each score, fold variation, parameters, and status.", "speech_text": "Enable hyper-parameter Search when several plausible configurations need comparison. The mini workbench runs grouped trials and displays each score, fold variation, parameters, and status.", "phonemes": "ɪnˈAbᵊl hˌIpəɹpəɹˈæməTəɹ sˈɜɹʧ wˌɛn sˈɛvəɹəl plˈɔzəbᵊl kənfˌɪɡjəɹˈAʃənz nˈid kəmpˈɛɹəsᵊn. ðə mˈɪni wˈɜɹkbˌɛnʧ ɹˈʌnz ɡɹˈupt tɹˈIᵊlz ænd dəsplˈAz ˈiʧ skˈɔɹ, fˈOld vˌɛɹiˈAʃən, pəɹˈæməTəɹz, ænd stˈATəs.", "authored_hold_after": 0.7, "hold_after": 0.52, "base_speed": 0.94, "speed_multiplier": 1.0, "effective_speed": 0.94, "sentences": [ { "sentence": 1, "speech_start": 105.98499999999999, "speech_end": 111.66499999999999, "audible_start": 106.08499999999998, "audible_end": 111.48499999999999, "duration": 5.68, "gap_from_previous": null, "trimmed_lead": 0.17, "trimmed_tail": 0.5, "text": "Enable Hyperparameter Search when several plausible configurations need comparison.", "speech_text": "Enable hyper-parameter Search when several plausible configurations need comparison.", "phonemes": "ɪnˈAbᵊl hˌIpəɹpəɹˈæməTəɹ sˈɜɹʧ wˌɛn sˈɛvəɹəl plˈɔzəbᵊl kənfˌɪɡjəɹˈAʃənz nˈid kəmpˈɛɹəsᵊn." }, { "sentence": 2, "speech_start": 111.78499999999998, "speech_end": 118.16999999999999, "audible_start": 111.88499999999999, "audible_end": 117.98999999999998, "duration": 6.385, "gap_from_previous": 0.4, "trimmed_lead": 0.17, "trimmed_tail": 0.52, "text": "The mini workbench runs grouped trials and displays each score, fold variation, parameters, and status.", "speech_text": "The mini workbench runs grouped trials and displays each score, fold variation, parameters, and status.", "phonemes": "ðə mˈɪni wˈɜɹkbˌɛnʧ ɹˈʌnz ɡɹˈupt tɹˈIᵊlz ænd dəsplˈAz ˈiʧ skˈɔɹ, fˈOld vˌɛɹiˈAʃən, pəɹˈæməTəɹz, ænd stˈATəs." } ] }, { "scene": 11, "speech_start": 118.68999999999998, "speech_end": 130.69, "scene_end": 131.21, "duration": 12.520000000000024, "text": "Search Settings defines a focused candidate space, ranking metric, grouped folds, and trial budget. Avoid broad searches that spend compute on choices the experiment cannot distinguish.", "speech_text": "Search Settings defines a focused candidate space, ranking metric, grouped folds, and trial budget. Avoid broad searches that spend compute on choices the experiment cannot distinguish.", "phonemes": "sˈɜɹʧ sˈɛTɪŋz dəfˈInz ɐ fˈOkəst kˈændədˌAt spˈAs, ɹˈæŋkɪŋ mˈɛtɹɪk, ɡɹˈupt fˈOldz, ænd tɹˈIᵊl bˈʌʤət. əvˈYd bɹˈɔd sˈɜɹʧᵻz ðæt spˈɛnd kəmpjˈut ˌɔn ʧˈYsᵻz ði ɪkspˈɛɹəmənt kənˈɑt dəstˈɪŋɡwɪʃ.", "authored_hold_after": 0.7, "hold_after": 0.52, "base_speed": 0.94, "speed_multiplier": 1.0, "effective_speed": 0.94, "sentences": [ { "sentence": 1, "speech_start": 118.68999999999998, "speech_end": 124.98499999999999, "audible_start": 118.78999999999998, "audible_end": 124.80499999999998, "duration": 6.295, "gap_from_previous": null, "trimmed_lead": 0.155, "trimmed_tail": 0.5, "text": "Search Settings defines a focused candidate space, ranking metric, grouped folds, and trial budget.", "speech_text": "Search Settings defines a focused candidate space, ranking metric, grouped folds, and trial budget.", "phonemes": "sˈɜɹʧ sˈɛTɪŋz dəfˈInz ɐ fˈOkəst kˈændədˌAt spˈAs, ɹˈæŋkɪŋ mˈɛtɹɪk, ɡɹˈupt fˈOldz, ænd tɹˈIᵊl bˈʌʤət." }, { "sentence": 2, "speech_start": 125.10499999999999, "speech_end": 130.69, "audible_start": 125.20499999999998, "audible_end": 130.51, "duration": 5.585, "gap_from_previous": 0.4, "trimmed_lead": 0.17, "trimmed_tail": 0.47, "text": "Avoid broad searches that spend compute on choices the experiment cannot distinguish.", "speech_text": "Avoid broad searches that spend compute on choices the experiment cannot distinguish.", "phonemes": "əvˈYd bɹˈɔd sˈɜɹʧᵻz ðæt spˈɛnd kəmpjˈut ˌɔn ʧˈYsᵻz ði ɪkspˈɛɹəmənt kənˈɑt dəstˈɪŋɡwɪʃ." } ] }, { "scene": 12, "speech_start": 131.21, "speech_end": 145.02, "scene_end": 145.54000000000002, "duration": 14.330000000000013, "text": "Run first validates labels and splits, then trains and evaluates the configured model. Review held-out metrics and examples before applying it broadly; spaCR stores the checkpoint, predictions, settings, and run record.", "speech_text": "Run first validates labels and splits, then trains and evaluates the configured model. Review held-out metrics and examples before applying it broadly; spacer stores the checkpoint, predictions, settings, and run record.", "phonemes": "ɹˈʌn fˈɜɹst vˈælɪdˌAts lˈAbəlz ænd splˈɪts, ðˈɛn tɹˈAnz ænd əvˈæljʊˌAts ðə kənfˈɪɡjəɹd mˈɑdᵊl. ɹəvjˈu hˈɛldˌWt mˈɛtɹɪks ænd ɪɡzˈæmpəlz bəfˈɔɹ əplˈIɪŋ ɪt bɹˈɔdli; spˈAsəɹ stˈɔɹz ðə ʧˈɛkpˌYnt, pɹidˈɪkʃənz, sˈɛTɪŋz, ænd ɹˈʌn ɹˈɛkəɹd.", "authored_hold_after": 0.7, "hold_after": 0.52, "base_speed": 0.94, "speed_multiplier": 1.0, "effective_speed": 0.94, "sentences": [ { "sentence": 1, "speech_start": 131.21, "speech_end": 136.92000000000002, "audible_start": 131.31, "audible_end": 136.74, "duration": 5.71, "gap_from_previous": null, "trimmed_lead": 0.165, "trimmed_tail": 0.5, "text": "Run first validates labels and splits, then trains and evaluates the configured model.", "speech_text": "Run first validates labels and splits, then trains and evaluates the configured model.", "phonemes": "ɹˈʌn fˈɜɹst vˈælɪdˌAts lˈAbəlz ænd splˈɪts, ðˈɛn tɹˈAnz ænd əvˈæljʊˌAts ðə kənfˈɪɡjəɹd mˈɑdᵊl." }, { "sentence": 2, "speech_start": 137.04000000000002, "speech_end": 145.02, "audible_start": 137.14000000000001, "audible_end": 144.84, "duration": 7.98, "gap_from_previous": 0.4, "trimmed_lead": 0.165, "trimmed_tail": 0.555, "text": "Review held-out metrics and examples before applying it broadly; spaCR stores the checkpoint, predictions, settings, and run record.", "speech_text": "Review held-out metrics and examples before applying it broadly; spacer stores the checkpoint, predictions, settings, and run record.", "phonemes": "ɹəvjˈu hˈɛldˌWt mˈɛtɹɪks ænd ɪɡzˈæmpəlz bəfˈɔɹ əplˈIɪŋ ɪt bɹˈɔdli; spˈAsəɹ stˈɔɹz ðə ʧˈɛkpˌYnt, pɹidˈɪkʃənz, sˈɛTɪŋz, ænd ɹˈʌn ɹˈɛkəɹd." } ] } ] }