Mirror Underdog Bark 3.0 Plus / Clean / Pro (CC-BY-4.0)

NOTICE

@@ -0,0 +1,24 @@
+AEmotionStudio/underdog-bark-models — a mirror of the Underdog Bark 3.0 family
+
+  plus/   ConwayResearch/Underdog-Bark-3.0-Plus   @ d4fab2f8b95681e5ad3cfb4e8c5f4e0c009b3deb
+  clean/  ConwayResearch/Underdog-Bark-3.0-Clean  @ 6d34ed8639b2b9a33e764acf06748519bbb950cf
+  pro/    ConwayResearch/Underdog-Bark-3.0-Pro    @ 4b5c51e04aa8d16dda9ecdff71bc29f7360149e4
+  processor/  nvidia/parakeet-tdt-0.6b-v3         @ 541d1f99c6b0c3cd0b11a95167540bb8edefd82b
+
+Underdog Bark 3.0, (c) Conway Research, CC-BY-4.0. It derives from:
+
+  nvidia/parakeet-tdt-0.6b-v3, (c) NVIDIA Corporation, CC-BY-4.0
+    https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3
+  moondream/parakeet-ultra, (c) Moondream, CC-BY-4.0
+    https://huggingface.co/moondream/parakeet-ultra
+
+License: https://creativecommons.org/licenses/by/4.0/
+
+Changes by Conway Research: post-trained and compressed. The FastConformer
+encoder architecture, TDT decoder, tokenizer and feature-extractor
+configuration derive from the models above and remain under CC-BY-4.0.
+
+Changes by Æmotion Studio in this mirror: plus/ and clean/ were decoded
+from Conway's entropy-coded model.bark into safetensors (exact decode;
+tensors stored as fp16 where that is bit-identical, fp32 otherwise). pro/
+and processor/ are byte-identical copies. No weights were otherwise altered.

README.md

@@ -0,0 +1,61 @@
+---
+license: cc-by-4.0
+language:
+  - en
+library_name: pytorch
+pipeline_tag: automatic-speech-recognition
+tags:
+  - automatic-speech-recognition
+  - parakeet
+  - tdt
+  - safetensors
+base_model:
+  - ConwayResearch/Underdog-Bark-3.0-Plus
+  - ConwayResearch/Underdog-Bark-3.0-Clean
+  - ConwayResearch/Underdog-Bark-3.0-Pro
+---
+
+# Underdog Bark 3.0 — MAESTRO mirror
+
+Mirror of the three [Underdog Bark 3.0](https://huggingface.co/ConwayResearch) English speech-recognition
+checkpoints that MAESTRO ships, with the NVIDIA feature-extractor and tokenizer files they share, so the
+in-app downloader has one stable, self-contained, offline-capable source.
+
+| folder | source | what it is |
+|---|---|---|
+| `plus/` | [ConwayResearch/Underdog-Bark-3.0-Plus](https://huggingface.co/ConwayResearch/Underdog-Bark-3.0-Plus) | verbatim — every word as spoken, fillers kept |
+| `clean/` | [ConwayResearch/Underdog-Bark-3.0-Clean](https://huggingface.co/ConwayResearch/Underdog-Bark-3.0-Clean) | fillers, false starts and repeats removed |
+| `pro/` | [ConwayResearch/Underdog-Bark-3.0-Pro](https://huggingface.co/ConwayResearch/Underdog-Bark-3.0-Pro) | the larger-download verbatim checkpoint |
+| `processor/` | [nvidia/parakeet-tdt-0.6b-v3](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3) | feature extractor + tokenizer (shared) |
+
+Each variant folder is a standard `ParakeetForTDT` checkpoint (`config.json`, `generation_config.json`,
+`model.safetensors`). Exact source revisions and the sha256 of every file are in `sha256.json`.
+
+## Attribution (CC-BY-4.0)
+
+Underdog Bark 3.0 is © **Conway Research**, CC BY 4.0. It derives from **nvidia/parakeet-tdt-0.6b-v3**
+(© NVIDIA Corporation, CC BY 4.0) by way of **moondream/parakeet-ultra** (© Moondream, CC BY 4.0).
+License: https://creativecommons.org/licenses/by/4.0/ — full chain in `NOTICE`.
+
+## Changes in this mirror
+
+- `plus/` and `clean/`: Conway publishes these as an entropy-coded `model.bark` that needs the
+  `constriction` ANS decoder and transformers ≥ 5.19 to load. They were decoded once with Conway's own
+  decode routine and written as safetensors. The decode is exact and so is the storage: each tensor is
+  fp16 where that is bit-identical to the decoded value (the level-coded encoder weights, 604M params)
+  and fp32 otherwise (23M params). No values were changed.
+- `pro/` and `processor/`: byte-identical copies.
+
+## Use
+
+MAESTRO loads these with a vendored Parakeet-TDT graph (AI › Transcribe › Underdog Bark). With
+transformers ≥ 5.19 the folders also load directly:
+
+```python
+from transformers import AutoProcessor, ParakeetForTDT
+model = ParakeetForTDT.from_pretrained("AEmotionStudio/underdog-bark-models", subfolder="plus")
+processor = AutoProcessor.from_pretrained("AEmotionStudio/underdog-bark-models", subfolder="processor")
+```
+
+Prepend 0.25 s of silence for Plus and Pro (Conway's evaluated setting; Clean uses none). Prefer fp16 or
+fp32: in MAESTRO's parity runs bf16 changed the transcript on 3 of 10 clips for Pro.

clean/config.json

@@ -0,0 +1,50 @@
+{
+  "architectures": [
+    "ParakeetForTDT"
+  ],
+  "blank_token_id": 8192,
+  "decoder_hidden_size": 640,
+  "dtype": "float32",
+  "durations": [
+    0,
+    1,
+    2,
+    3,
+    4
+  ],
+  "encoder_config": {
+    "activation_dropout": 0.1,
+    "attention_bias": false,
+    "attention_dropout": 0.1,
+    "conv_kernel_size": 9,
+    "convolution_bias": false,
+    "dropout": 0.1,
+    "dropout_positions": 0.0,
+    "dtype": "float32",
+    "hidden_act": "silu",
+    "hidden_size": 1024,
+    "initializer_range": 0.02,
+    "intermediate_size": 4096,
+    "layerdrop": 0.1,
+    "max_position_embeddings": 5000,
+    "model_type": "parakeet_encoder",
+    "num_attention_heads": 8,
+    "num_hidden_layers": 24,
+    "num_key_value_heads": 8,
+    "num_mel_bins": 128,
+    "scale_input": false,
+    "subsampling_conv_channels": 256,
+    "subsampling_conv_kernel_size": 3,
+    "subsampling_conv_stride": 2,
+    "subsampling_factor": 8
+  },
+  "hidden_act": "relu",
+  "initializer_range": 0.02,
+  "is_encoder_decoder": true,
+  "max_symbols_per_step": 10,
+  "model_type": "parakeet_tdt",
+  "num_decoder_layers": 2,
+  "pad_token_id": 2,
+  "transformers_version": "5.19.0",
+  "vocab_size": 8193
+}

clean/generation_config.json

@@ -0,0 +1,16 @@
+{
+  "_from_model_config": true,
+  "decoder_start_token_id": 8192,
+  "eos_token_id": 3,
+  "output_attentions": false,
+  "output_hidden_states": false,
+  "pad_token_id": 2,
+  "suppress_tokens": [
+    8193,
+    8194,
+    8195,
+    8196,
+    8197
+  ],
+  "transformers_version": "5.6.0.dev0"
+}

clean/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f56b3239f5afaa23263e04af8d8ca7d3ab83b53058ef24dcc22d6c5bc099023
+size 1299506732

plus/config.json

@@ -0,0 +1,50 @@
+{
+  "architectures": [
+    "ParakeetForTDT"
+  ],
+  "blank_token_id": 8192,
+  "decoder_hidden_size": 640,
+  "dtype": "float32",
+  "durations": [
+    0,
+    1,
+    2,
+    3,
+    4
+  ],
+  "encoder_config": {
+    "activation_dropout": 0.1,
+    "attention_bias": false,
+    "attention_dropout": 0.1,
+    "conv_kernel_size": 9,
+    "convolution_bias": false,
+    "dropout": 0.1,
+    "dropout_positions": 0.0,
+    "dtype": "float32",
+    "hidden_act": "silu",
+    "hidden_size": 1024,
+    "initializer_range": 0.02,
+    "intermediate_size": 4096,
+    "layerdrop": 0.1,
+    "max_position_embeddings": 5000,
+    "model_type": "parakeet_encoder",
+    "num_attention_heads": 8,
+    "num_hidden_layers": 24,
+    "num_key_value_heads": 8,
+    "num_mel_bins": 128,
+    "scale_input": false,
+    "subsampling_conv_channels": 256,
+    "subsampling_conv_kernel_size": 3,
+    "subsampling_conv_stride": 2,
+    "subsampling_factor": 8
+  },
+  "hidden_act": "relu",
+  "initializer_range": 0.02,
+  "is_encoder_decoder": true,
+  "max_symbols_per_step": 10,
+  "model_type": "parakeet_tdt",
+  "num_decoder_layers": 2,
+  "pad_token_id": 2,
+  "transformers_version": "5.19.0",
+  "vocab_size": 8193
+}

plus/generation_config.json

@@ -0,0 +1,16 @@
+{
+  "_from_model_config": true,
+  "decoder_start_token_id": 8192,
+  "eos_token_id": 3,
+  "output_attentions": false,
+  "output_hidden_states": false,
+  "pad_token_id": 2,
+  "suppress_tokens": [
+    8193,
+    8194,
+    8195,
+    8196,
+    8197
+  ],
+  "transformers_version": "5.6.0.dev0"
+}

plus/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4fb681ad5f806a533222b29224d64b5f1cdbbeca1cf6e53b34e442e7372e01c
+size 1299506732

pro/config.json

@@ -0,0 +1,50 @@
+{
+  "architectures": [
+    "ParakeetForTDT"
+  ],
+  "blank_token_id": 8192,
+  "decoder_hidden_size": 640,
+  "dtype": "bfloat16",
+  "durations": [
+    0,
+    1,
+    2,
+    3,
+    4
+  ],
+  "encoder_config": {
+    "activation_dropout": 0.1,
+    "attention_bias": false,
+    "attention_dropout": 0.1,
+    "conv_kernel_size": 9,
+    "convolution_bias": false,
+    "dropout": 0.1,
+    "dropout_positions": 0.0,
+    "dtype": "float32",
+    "hidden_act": "silu",
+    "hidden_size": 1024,
+    "initializer_range": 0.02,
+    "intermediate_size": 4096,
+    "layerdrop": 0.1,
+    "max_position_embeddings": 5000,
+    "model_type": "parakeet_encoder",
+    "num_attention_heads": 8,
+    "num_hidden_layers": 24,
+    "num_key_value_heads": 8,
+    "num_mel_bins": 128,
+    "scale_input": false,
+    "subsampling_conv_channels": 256,
+    "subsampling_conv_kernel_size": 3,
+    "subsampling_conv_stride": 2,
+    "subsampling_factor": 8
+  },
+  "hidden_act": "relu",
+  "initializer_range": 0.02,
+  "is_encoder_decoder": true,
+  "max_symbols_per_step": 10,
+  "model_type": "parakeet_tdt",
+  "num_decoder_layers": 2,
+  "pad_token_id": 2,
+  "transformers_version": "5.19.0",
+  "vocab_size": 8193
+}

pro/generation_config.json

@@ -0,0 +1,16 @@
+{
+  "_from_model_config": true,
+  "decoder_start_token_id": 8192,
+  "eos_token_id": 3,
+  "output_attentions": false,
+  "output_hidden_states": false,
+  "pad_token_id": 2,
+  "suppress_tokens": [
+    8193,
+    8194,
+    8195,
+    8196,
+    8197
+  ],
+  "transformers_version": "5.19.0"
+}

pro/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f84b33e63ebd1851e910e8b2fe299d29f5e0efc6f57e84d92d2f79af464e0e5
+size 1254196636

processor/processor_config.json

@@ -0,0 +1,16 @@
+{
+  "blank_token": "<blank>",
+  "feature_extractor": {
+    "feature_extractor_type": "ParakeetFeatureExtractor",
+    "feature_size": 128,
+    "hop_length": 160,
+    "n_fft": 512,
+    "padding_side": "right",
+    "padding_value": 0.0,
+    "preemphasis": 0.97,
+    "return_attention_mask": true,
+    "sampling_rate": 16000,
+    "win_length": 400
+  },
+  "processor_class": "ParakeetProcessor"
+}

processor/tokenizer_config.json

@@ -0,0 +1,10 @@
+{
+  "backend": "tokenizers",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<pad>",
+  "processor_class": "ParakeetProcessor",
+  "tokenizer_class": "ParakeetTokenizer",
+  "unk_token": "<unk>"
+}

sha256.json

@@ -0,0 +1,54 @@
+{
+  "files": {
+    "NOTICE": "fbab25d9ef392a9e0ec8b58e9792be2ac5de42ebc7b44208c2c000a1362e70f7",
+    "README.md": "f8e814f38be9e6f198fae2a022c35ae75af0da096ab5319ad21f4c1293a07169",
+    "clean/config.json": "3daa160911a59090325b95a91a4c28e2ff191780f61f025b41484e9cc64675cf",
+    "clean/generation_config.json": "b141de6ec6d7f982ece13f98f604e3fe1807ea9c0e839185d0ab7064604209d0",
+    "clean/model.safetensors": "8f56b3239f5afaa23263e04af8d8ca7d3ab83b53058ef24dcc22d6c5bc099023",
+    "plus/config.json": "3daa160911a59090325b95a91a4c28e2ff191780f61f025b41484e9cc64675cf",
+    "plus/generation_config.json": "b141de6ec6d7f982ece13f98f604e3fe1807ea9c0e839185d0ab7064604209d0",
+    "plus/model.safetensors": "d4fb681ad5f806a533222b29224d64b5f1cdbbeca1cf6e53b34e442e7372e01c",
+    "pro/config.json": "269282621d92eb22c3590fe0340530fec8a092d0387169bcd895c65541fa1ce0",
+    "pro/generation_config.json": "96e230a759fa74e65c36f654e5f9c909a426459d105e752f9003ddd7d328f8f3",
+    "pro/model.safetensors": "6f84b33e63ebd1851e910e8b2fe299d29f5e0efc6f57e84d92d2f79af464e0e5",
+    "processor/processor_config.json": "8346a93a3b987fa1dec57a78f045cd0817d21786589a5a096b41a57a446fd1d7",
+    "processor/tokenizer.json": "bd321b096832a3f270bd3b2a88823957920f1a5c5ada71114a26ea729d0cbe91",
+    "processor/tokenizer_config.json": "0b2fe0037599ee335f0b972fa682bf0ece74e4ccfec755cb7daa3405d3d3e874"
+  },
+  "provenance": {
+    "plus": {
+      "repo": "ConwayResearch/Underdog-Bark-3.0-Plus",
+      "revision": "d4fab2f8b95681e5ad3cfb4e8c5f4e0c009b3deb",
+      "source_file": "model.bark",
+      "source_sha256": "fdd9f1052e4fdead07d71e8cd2138813eb0c209fd6ff8b1b6f47737e1e501857",
+      "fp16_params": 604401926,
+      "fp32_params": 22655360
+    },
+    "clean": {
+      "repo": "ConwayResearch/Underdog-Bark-3.0-Clean",
+      "revision": "6d34ed8639b2b9a33e764acf06748519bbb950cf",
+      "source_file": "model.bark",
+      "source_sha256": "b67ca90d94b79a36af4616f1cb3f241884925de7aa9f3e673a8d5c6c4cfd3436",
+      "fp16_params": 604401926,
+      "fp32_params": 22655360
+    },
+    "pro": {
+      "repo": "ConwayResearch/Underdog-Bark-3.0-Pro",
+      "revision": "4b5c51e04aa8d16dda9ecdff71bc29f7360149e4",
+      "copied_verbatim": [
+        "model.safetensors",
+        "config.json",
+        "generation_config.json"
+      ]
+    },
+    "processor": {
+      "repo": "nvidia/parakeet-tdt-0.6b-v3",
+      "revision": "541d1f99c6b0c3cd0b11a95167540bb8edefd82b",
+      "copied_verbatim": [
+        "processor_config.json",
+        "tokenizer.json",
+        "tokenizer_config.json"
+      ]
+    }
+  }
+}