janhq/ichigo

Issues

• epic: Finetune F5-TTS on vietnamese data.

  #150 opened 6 days ago
  0

• RQBottleneckTransformer.load_model("whisper-vq-stoks-v3-7lang-fixed.model") gives error

  #149 opened 13 days ago
  0

• chore: Augmenting current Vietnamese speech dataset

  #148 opened 14 days ago
  1

• chore: Whisper model demo

  #147 opened 10 days ago
  2

• idea: push Ichago inference to upstream llama.cpp

  #142 opened 20 days ago
  0

• feat: More multilingual audio training data

  #143 opened 21 days ago
  1

• training run: tuning ichigo-quantizer

  #144 opened 24 days ago
  15

• task: MLX quantizer to support productionize ichigo

  #141 opened a month ago
  0

• research: Flow Matching for synthetic data generation

  #140 opened a month ago
  6

• task: Train and test text2semantic under decoder only framework for ichigo v0.5

  #145 opened a month ago
  18

• worklog: Ichigo v0.5 development

  #136 opened a month ago
  1

• data: Singlish Dataset

  #135 opened a month ago
  1

• Research: Downloading National Speech Corpus

  #138 opened a month ago
  1

• training run: Lora Ichigo Qwen2.5 32B

  #134 opened 22 days ago
  4

• training run: Qwen 0.5B as Text to semantic model

  #137 opened a month ago
  0

• Some Question about the technology report.

  #132 opened a month ago
  6

• bug: Ichigo Notebook v0.4 not working

  #128 opened a month ago
  0

• idea: Precision scaling research

  #127 opened a month ago
  7

• idea: Training Ichigo on Structured Output

  #126 opened a month ago
  1

• feat: Modify Torchtune to train Ichigo Qwen

  #125 opened a month ago
  4

• task: Use VieTTS if output language is Vietnamese

  #123 opened a month ago
  2

• task: Ichigo LLM v0.5 Training

  #122 opened a month ago
  3

• task: Instruct Dataset Creation for Multilingual Speech (Phase 2)

  #121 opened a month ago
  3

• planning: Train our own Quantizer for multilingual speech to work with Whisper Encoder

  #146 opened a month ago
  13

• data: Dataset Collection & Preprocessing

  #119 opened a month ago
  0

• task: Pretraining dataset for Multilingual Quantizer training (Phase 1)

  #118 opened a month ago
  6

• epic: [DESCRIPTION]

  #117 opened a month ago
  0

• milestone: Ichigo v0.5 Multi-lingual

  #116 opened a month ago
  0

• bug: Running Installation guide from README not working

  #112 opened a month ago
  0

• planning: Ichigo Decoder

  #111 opened 24 days ago
  2

• bug: Google Collab Inference with homebrewltd / Ichigo-llama3.1-s-instruct-v0.4

  #109 opened a month ago
  3

• feat: Ichigo from semantics to sound tokens

  #108 opened a month ago
  4

• Utilizing Ichigo Without Self-Training

  #107 opened a month ago
  1

• bug: The demo is frozen sometimes on multi-turn

  #105 opened a month ago
  2

• bug: [DESCRIPTION] SFT trainer does not use GPU

  #104 opened 24 days ago
  7

• epic: Replace optimizers or code new optimizers?

  #103 opened 2 months ago
  4

• run: ichigo 0.4

  #102 opened 2 months ago
  2

• Can it run on Apple Silicon by any chance?

  #101 opened a month ago
  3

• epic: Introducing naturalness for interuption

  #100 opened 24 days ago
  4

• epic: Reviews of Ichigo paper

  #99 opened 2 months ago
  9

• Reproducing the OpenHermes and Alpaca Audio Evals

  #98 opened 2 months ago
  2

• idea: official docker image / container builds

  #95 opened 2 months ago
  1

• feat: fast open source tts for ichigo

  #94 opened 24 days ago
  12

• planning: Ichigo v0.4 Speech Enhancements

  #93 opened 24 days ago
  6

• planning: Ichigo Encoder

  #92 opened 24 days ago
  1

• planning: Ichigo VAD

  #91 opened 2 months ago
  4

• planning: Ichigo Transcription

  #90 opened 2 months ago
  3

• I am newbie on multimodal. Can you explain sound_tokens types?

  #87 opened a month ago
  0

• License?

  #86 opened 2 months ago
  1

• How can the voice be changed?

  #85 opened a month ago
  1