lyranovaheart/auriaetherwiing_tq3.5-27b-musica-v1-gguf overview

{
  "metadata": {},
  "card_data": {
    "license": "apache-2.0",
    "datasets": [
      "EVA-UNIT-01/Lilith-v0.3",
      "zerofata/Gemini-3.1-Pro-GLM5-Characters",
      "zerofata/Instruct-Anime",
      "zerofata/Anime-AMA-Prose",
      "allura-forge/mimo-v2-pro-claude-distill-hs3",
      "allura-forge/doubao-seed2.0-distill-multiturn-expr-rp",
      "Delta-Vector/Orion-Deepseek-V3-RP-Filtered",
      "Delta-Vector/Orion-Deepseek-R1-RP-Filtered",
      "Gryphe/ChatGPT-4o-Writing-Prompts",
      "Gryphe/Sonnet3.5-Charcard-Roleplay",
      "ToastyPigeon/kimi-stories-instruct",
      "ToastyPigeon/kimi-rp-v3",
      "ToastyPigeon/fujin-filtered-instruct",
      "Dxniz/Novelist-CoT"
    ],
    "language": [
      "en"
    ],
    "base_model": [
      "AuriAetherwiing/TQ3.5-27B-Musica-v1"
    ],
    "base_model_relation": "quantized",
    "pipeline_tag": "image-text-to-text",
    "frontmatter": {
      "license": "apache-2.0",
      "datasets": [
        "EVA-UNIT-01/Lilith-v0.3",
        "zerofata/Gemini-3.1-Pro-GLM5-Characters",
        "zerofata/Instruct-Anime",
        "zerofata/Anime-AMA-Prose",
        "allura-forge/mimo-v2-pro-claude-distill-hs3",
        "allura-forge/doubao-seed2.0-distill-multiturn-expr-rp",
        "Delta-Vector/Orion-Deepseek-V3-RP-Filtered",
        "Delta-Vector/Orion-Deepseek-R1-RP-Filtered",
        "Gryphe/ChatGPT-4o-Writing-Prompts",
        "Gryphe/Sonnet3.5-Charcard-Roleplay",
        "ToastyPigeon/kimi-stories-instruct",
        "ToastyPigeon/kimi-rp-v3",
        "ToastyPigeon/fujin-filtered-instruct",
        "Dxniz/Novelist-CoT"
      ],
      "language": [
        "en"
      ],
      "base_model": [
        "AuriAetherwiing/TQ3.5-27B-Musica-v1"
      ],
      "base_model_relation": "quantized",
      "pipeline_tag": "image-text-to-text"
    },
    "hero_image_url": "",
    "summary": "**GGUF Quants of Qwen 3.5 27B Musica v1** Using llama.cpp release b8533, non iMatrix quants. Original Model: https://huggingface.co/AuriAetherwiing/TQ3.5-27B-Musica-v1 Personal favorite of mine too, I really like this model! Thanks Auri! **Original Model Card Below** **Qwen3.5-27B Musica v1** RP/storygen/conversational tune of Qwen3.5-27B. Stylewise looked pretty nice to me and seems decently steerable, should also reduce refusal rate even further than derestricted ver on which it was based on. Both reasoning and non-reasoning modes are supported, reasoning mode even has several styes of reasoning, reroll to see them (perhaps I should mark them to make them manually evokable on next iter?). Might or might not have slightly better world knowledge than base, lol. This training run was sponsored by ArliAI Wishlist for next iter - more conversational reasoning data (and more reasoning data in general) and perhaps something multiturn for creative writing. Perhaps also train Qwen3.5-9B and Nemotron-3-Super-120B-A12B before iterating on dataset. **Training notes** Rank 64, alpha 64 LoRA on top of ArliAI's Derestricted version, for two epochs with constant scheduler. Training took ~17 hours on OwenArli's 2xRTX Pro 6000 Blackwell. Run's graphs on Comet (DW about that API key in config, I deactivated it before sharing this lol) **Recommended samplers** **Axolotl config** See Axolotl config ``yaml base_model: /home/arli/models/Qwen3.5-27B-Derestricted plugins: cut_cross_entropy: true liger_rope: true liger_rms_norm: true liger_layer_norm: true liger_glu_activation: true load_in_8bit: false load_in_4bit: false shuffle_merged_datasets: true datasets: type: chat_template field_messages: conversations message_property_mappings: role: from content: value type: chat_template field_messages: conversations message_property_mappings: role: from content: value dataset_prepared_path: ./last_run_prepared val_set_size: 0 output_dir: ./outputs/v1 adapter: lora save_safetensors: true sequence_len: 16384 sample_packing: true pad_to_sequence_len: true lora_r: 64 lora_alpha: 64 lora_dropout: 0.0 lora_target_modules: # Uncomment below to also target the linear attention projections. # These use separate in_proj_qkv / in_proj_z / out_proj (Qwen3.5-specific). # - linear_attn.in_proj_qkv # - linear_attn.in_proj_z # - linear_attn.out_proj lora_mlp_kernel: false lora_qkv_kernel: false lora_o_kernel: false gradient_accumulation_steps: 8 micro_batch_size: 1 num_epochs: 2 optimizer: adamw_torch_fused lr_scheduler: constant learning_rate: 4e-6 max_grad_norm: 0.5 bf16: auto use_comet: true comet_project_name: musica-27b auto_resume_from_checkpoints: false logging_steps: 1 flash_attention: true warmup_ratio: 0 evals_per_epoch: 0 saves_per_epoch: 4 save_total_limit: 4 gradient_checkpointing: false gradient_checkpointing_kwargs: use_reentrant: false fsdp_config: fsdp_version: 2 offload_params: false cpu_ram_efficient_loading: false auto_wrap_policy: TRANSFORMER_BASED_WRAP transformer_layer_cls_to_wrap: Qwen3_5DecoderLayer state_dict_type: FULL_STATE_DICT sharding_strategy: FULL_SHARD reshard_after_forward: true activation_checkpointing: true ``",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nlicense: apache-2.0\ndatasets:\n- EVA-UNIT-01/Lilith-v0.3\n- zerofata/Gemini-3.1-Pro-GLM5-Characters\n- zerofata/Instruct-Anime\n- zerofata/Anime-AMA-Prose\n- allura-forge/mimo-v2-pro-claude-distill-hs3\n- allura-forge/doubao-seed2.0-distill-multiturn-expr-rp\n- Delta-Vector/Orion-Deepseek-V3-RP-Filtered\n- Delta-Vector/Orion-Deepseek-R1-RP-Filtered\n- Gryphe/ChatGPT-4o-Writing-Prompts\n- Gryphe/Sonnet3.5-Charcard-Roleplay\n- ToastyPigeon/kimi-stories-instruct\n- ToastyPigeon/kimi-rp-v3\n- ToastyPigeon/fujin-filtered-instruct\n- Dxniz/Novelist-CoT\nlanguage:\n- en\nbase_model:\n- AuriAetherwiing/TQ3.5-27B-Musica-v1\nbase_model_relation: quantized\npipeline_tag: image-text-to-text\n---\n\n**GGUF Quants of Qwen 3.5 27B Musica v1**\n\nUsing llama.cpp release b8533, non iMatrix quants.\n\nOriginal Model: https://huggingface.co/AuriAetherwiing/TQ3.5-27B-Musica-v1\n\nPersonal favorite of mine too, I really like this model! Thanks Auri! \n\n**Original Model Card Below**\n\n**Qwen3.5-27B Musica v1**\n\nRP/storygen/conversational tune of Qwen3.5-27B. Stylewise looked pretty nice to me and seems decently steerable, should also reduce refusal rate even further than derestricted ver on which it was based on. Both reasoning and non-reasoning modes are supported, reasoning mode even has several styes of reasoning, reroll to see them (perhaps I should mark them to make them manually evokable on next iter?). Might or might not have slightly better world knowledge than base, lol.\n\nThis training run was sponsored by [ArliAI](https://www.arliai.com/)\n\nWishlist for next iter - more conversational reasoning data (and more reasoning data in general) and perhaps something multiturn for creative writing. Perhaps also train Qwen3.5-9B and Nemotron-3-Super-120B-A12B before iterating on dataset. \n\n**Training notes**\n\nRank 64, alpha 64 LoRA on top of ArliAI's Derestricted version, for two epochs with constant scheduler. Training took ~17 hours on OwenArli's 2xRTX Pro 6000 Blackwell.\n\n[Run's graphs on Comet](https://www.comet.com/aetherwiing/musica-27b/view/new/panels) (DW about that API key in config, I deactivated it before sharing this lol)\n\n**Recommended samplers**\n\n- Temperature: 1\n- NSigma: 2\n- Min-P: 0.02\n\n\n**Axolotl config** \n\n<details><summary>See Axolotl config</summary>\n  \n```yaml\nbase_model: /home/arli/models/Qwen3.5-27B-Derestricted\n\nplugins:\n  - axolotl.integrations.liger.LigerPlugin\n  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin\ncut_cross_entropy: true\nliger_rope: true\nliger_rms_norm: true\nliger_layer_norm: true\nliger_glu_activation: true\n\nload_in_8bit: false\nload_in_4bit: false\n\nshuffle_merged_datasets: true\ndatasets:\n  - path: ./musica-nonreasoning-sft-megafix.jsonl\n    type: chat_template\n    field_messages: conversations\n    message_property_mappings:\n      role: from\n      content: value\n  - path: ./musica-reasoning-sft-fix.jsonl\n    type: chat_template\n    field_messages: conversations\n    message_property_mappings:\n      role: from\n      content: value\n\ndataset_prepared_path: ./last_run_prepared\nval_set_size: 0\noutput_dir: ./outputs/v1\nadapter: lora\nsave_safetensors: true\n\nsequence_len: 16384\nsample_packing: true\npad_to_sequence_len: true\n\nlora_r: 64\nlora_alpha: 64\nlora_dropout: 0.0\nlora_target_modules:\n  - q_proj\n  - k_proj\n  - v_proj\n  - o_proj\n  - down_proj\n  - up_proj\n  # Uncomment below to also target the linear attention projections.\n  # These use separate in_proj_qkv / in_proj_z / out_proj (Qwen3.5-specific).\n  # - linear_attn.in_proj_qkv\n  # - linear_attn.in_proj_z\n  # - linear_attn.out_proj\n\nlora_mlp_kernel: false\nlora_qkv_kernel: false\nlora_o_kernel: false\n\ngradient_accumulation_steps: 8\nmicro_batch_size: 1\nnum_epochs: 2\noptimizer: adamw_torch_fused\nlr_scheduler: constant\nlearning_rate: 4e-6\nmax_grad_norm: 0.5\n\nbf16: auto\n\nuse_comet: true\ncomet_project_name: musica-27b\n\nauto_resume_from_checkpoints: false\nlogging_steps: 1\nflash_attention: true\n\nwarmup_ratio: 0\nevals_per_epoch: 0\nsaves_per_epoch: 4\nsave_total_limit: 4\n\ngradient_checkpointing: false\ngradient_checkpointing_kwargs:\n  use_reentrant: false\n\nfsdp_config:\n  fsdp_version: 2\n  offload_params: false\n  cpu_ram_efficient_loading: false\n  auto_wrap_policy: TRANSFORMER_BASED_WRAP\n  transformer_layer_cls_to_wrap: Qwen3_5DecoderLayer\n  state_dict_type: FULL_STATE_DICT\n  sharding_strategy: FULL_SHARD\n  reshard_after_forward: true\n  activation_checkpointing: true\n\n```\n\n</details>",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "image-text-to-text",
    "en",
    "dataset:EVA-UNIT-01/Lilith-v0.3",
    "dataset:zerofata/Gemini-3.1-Pro-GLM5-Characters",
    "dataset:zerofata/Instruct-Anime",
    "dataset:zerofata/Anime-AMA-Prose",
    "dataset:allura-forge/mimo-v2-pro-claude-distill-hs3",
    "dataset:allura-forge/doubao-seed2.0-distill-multiturn-expr-rp",
    "dataset:Delta-Vector/Orion-Deepseek-V3-RP-Filtered",
    "dataset:Delta-Vector/Orion-Deepseek-R1-RP-Filtered",
    "dataset:Gryphe/ChatGPT-4o-Writing-Prompts",
    "dataset:Gryphe/Sonnet3.5-Charcard-Roleplay",
    "dataset:ToastyPigeon/kimi-stories-instruct",
    "dataset:ToastyPigeon/kimi-rp-v3",
    "dataset:ToastyPigeon/fujin-filtered-instruct",
    "dataset:Dxniz/Novelist-CoT",
    "base_model:AuriAetherwiing/TQ3.5-27B-Musica-v1",
    "base_model:quantized:AuriAetherwiing/TQ3.5-27B-Musica-v1",
    "license:apache-2.0",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 2,
  "downloads": 948,
  "gated": false,
  "private": false,
  "last_modified": "2026-03-27T00:27:40.000Z",
  "created_at": "2026-03-26T23:23:56.000Z",
  "pipeline_tag": "image-text-to-text",
  "library_name": ""
}

{
  "_id": "69c5c00c7a313d6bf2fb9c3b",
  "id": "LyraNovaHeart/AuriAetherwiing_TQ3.5-27B-Musica-v1-GGUF",
  "modelId": "LyraNovaHeart/AuriAetherwiing_TQ3.5-27B-Musica-v1-GGUF",
  "sha": "80df8ab261e2392f7b25c66abdb292f2244bbc87",
  "createdAt": "2026-03-26T23:23:56.000Z",
  "lastModified": "2026-03-27T00:27:40.000Z",
  "author": "LyraNovaHeart",
  "downloads": 948,
  "likes": 2,
  "gated": false,
  "private": false,
  "pipeline_tag": "image-text-to-text",
  "library_name": "",
  "siblings_count": 13
}