eeropai/dolphin-2.9.3-mistral-nemo-12b-gguf overview

{
  "metadata": {},
  "card_data": {
    "license": "apache-2.0",
    "base_model": "mistralai/Mistral-Nemo-Base-2407",
    "tags": [
      "generated_from_trainer",
      "axolotl"
    ],
    "datasets": [
      "cognitivecomputations/Dolphin-2.9",
      "teknium/OpenHermes-2.5",
      "m-a-p/CodeFeedback-Filtered-Instruction",
      "cognitivecomputations/dolphin-coder",
      "cognitivecomputations/samantha-data",
      "microsoft/orca-math-word-problems-200k",
      "Locutusque/function-calling-chatml",
      "internlm/Agent-FLAN"
    ],
    "frontmatter": {
      "license": "apache-2.0",
      "base_model": "mistralai/Mistral-Nemo-Base-2407",
      "tags": [
        "generated_from_trainer",
        "axolotl"
      ],
      "datasets": [
        "cognitivecomputations/Dolphin-2.9",
        "teknium/OpenHermes-2.5",
        "m-a-p/CodeFeedback-Filtered-Instruction",
        "cognitivecomputations/dolphin-coder",
        "cognitivecomputations/samantha-data",
        "microsoft/orca-math-word-problems-200k",
        "Locutusque/function-calling-chatml",
        "internlm/Agent-FLAN"
      ]
    },
    "hero_image_url": "https://cdn-uploads.huggingface.co/production/uploads/63111b2d88942700629f5771/ldkN1J0WIDQwU4vutGYiD.png",
    "summary": "This is the llama.cpp gguf conversion of the original model located here: https://huggingface.co/cognitivecomputations/dolphin-2.9.3-mistral-nemo-12b Curated and trained by Eric Hartford and Cognitive Computations ![Discord](https://discord.gg/h3K4XGj2RH) Discord: https://discord.gg/h3K4XGj2RH  Our appreciation for the sponsors of Dolphin 2.9.3: This model is based on mistralai/Mistral-Nemo-Base-2407, and is governed by the apache 2.0 license. The base model has 128K context, and our finetuning used 8192 sequence length. Dolphin 2.9.3 uses ChatML prompt template format. example: `` system You are Dolphin, a helpful AI assistant. user {prompt} assistant `` Dolphin-2.9.3 has a variety of instruction following, conversational, and coding skills. It also has initial agentic abilities and supports function calling. Dolphin is uncensored. We have filtered the dataset to remove alignment and bias. This makes the model more compliant. You are advised to implement your own alignment layer before exposing the model as a service. It will be highly compliant with any requests, even unethical ones. Please read my blog post about uncensored models. https://erichartford.com/uncensored-models You are responsible for any content you create using this model. Enjoy responsibly. Dolphin is licensed according to apache 2.0 license.  We grant permission for any use, including commercial. Dolphin was trained on data generated from GPT4, among other models.",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nlicense: apache-2.0\nbase_model: mistralai/Mistral-Nemo-Base-2407\ntags:\n- generated_from_trainer\n- axolotl\ndatasets:\n- cognitivecomputations/Dolphin-2.9\n- teknium/OpenHermes-2.5\n- m-a-p/CodeFeedback-Filtered-Instruction\n- cognitivecomputations/dolphin-coder\n- cognitivecomputations/samantha-data\n- microsoft/orca-math-word-problems-200k\n- Locutusque/function-calling-chatml\n- internlm/Agent-FLAN\n---\n\n# Dolphin 2.9.3 Mistral Nemo 12b 🐬\n\nThis is the llama.cpp gguf conversion of the original model located here: \n\nhttps://huggingface.co/cognitivecomputations/dolphin-2.9.3-mistral-nemo-12b\n\nCurated and trained by Eric Hartford and Cognitive Computations\n\n[![Discord](https://img.shields.io/discord/1156064224225808488?logo=Discord&logoColor=%23ffffff&label=Discord&link=https%3A%2F%2Fdiscord.gg%2FtCMkMDDHwm)](https://discord.gg/h3K4XGj2RH)\nDiscord: https://discord.gg/h3K4XGj2RH\n\n<img src=\"https://cdn-uploads.huggingface.co/production/uploads/63111b2d88942700629f5771/ldkN1J0WIDQwU4vutGYiD.png\" width=\"600\" />\n\nOur appreciation for the sponsors of Dolphin 2.9.3:\n- [Crusoe Cloud](https://crusoe.ai/) - provided excellent on-demand 8xL40S node\n\nThis model is based on mistralai/Mistral-Nemo-Base-2407, and is governed by the apache 2.0 license.\n\nThe base model has 128K context, and our finetuning used 8192 sequence length.\n\nDolphin 2.9.3 uses ChatML prompt template format.\n\nexample:\n\n```\n<|im_start|>system\nYou are Dolphin, a helpful AI assistant.<|im_end|>\n<|im_start|>user\n{prompt}<|im_end|>\n<|im_start|>assistant\n\n```\n\nDolphin-2.9.3 has a variety of instruction following, conversational, and coding skills. It also has initial agentic abilities and supports function calling.\n\nDolphin is uncensored. We have filtered the dataset to remove alignment and bias. This makes the model more compliant. You are advised to implement your own alignment layer before exposing the model as a service. It will be highly compliant with any requests, even unethical ones. Please read my blog post about uncensored models. https://erichartford.com/uncensored-models You are responsible for any content you create using this model. Enjoy responsibly.\n\nDolphin is licensed according to apache 2.0 license.  We grant permission for any use, including commercial. Dolphin was trained on data generated from GPT4, among other models.\n\n## Evals\n\nTBD\n\n## Training\n\n<!-- This model card has been generated automatically according to the information the Trainer had access to. You\nshould probably proofread and complete it, then remove this comment. -->\n\n[<img src=\"https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png\" alt=\"Built with Axolotl\" width=\"200\" height=\"32\"/>](https://github.com/axolotl-ai-cloud/axolotl)\n<details><summary>See axolotl config</summary>\n\naxolotl version: `0.4.1`\n```yaml\nbase_model: /workspace/models/Mistral-Nemo-Base-2407\nmodel_type: AutoModelForCausalLM\ntokenizer_type: AutoTokenizer\n\nload_in_8bit: false\n# load_in_4bit: true\nstrict: false\n\ndatasets:\n  - path: /workspace/datasets/dolphin-2.9.3/dolphin201-sharegpt2.jsonl\n    type: sharegpt\n    conversation: chatml\n  - path: /workspace/datasets/dolphin-2.9.3/SystemChat_filtered_sharegpt.jsonl\n    type: sharegpt\n    conversation: chatml\n  - path: /workspace/datasets/dolphin-2.9.3/SystemChat_multilingual_sharegpt.jsonl\n    type: sharegpt\n    conversation: chatml\n  - path: /workspace/datasets/dolphin-2.9.3/dolphin-coder-translate-sharegpt2.jsonl\n    type: sharegpt\n    conversation: chatml\n  - path: /workspace/datasets/dolphin-2.9.3/dolphin-coder-codegen-sharegpt2.jsonl\n    type: sharegpt\n    conversation: chatml\n  - path: /workspace/datasets/dolphin-2.9.3/m-a-p_Code-Feedback-sharegpt-unfiltered.jsonl\n    type: sharegpt\n    conversation: chatml\n  - path: /workspace/datasets/dolphin-2.9.3/m-a-p_CodeFeedback-Filtered-Instruction-sharegpt-unfiltered.jsonl\n    type: sharegpt\n    conversation: chatml\n  - path: /workspace/datasets/dolphin-2.9.3/not_samantha_norefusals.jsonl\n    type: sharegpt\n    conversation: chatml\n  - path: /workspace/datasets/dolphin-2.9.3/Orca-Math-resort-unfiltered.jsonl\n    type: sharegpt\n    conversation: chatml\n  - path: /workspace/datasets/dolphin-2.9.3/agent_instruct_react_unfiltered.jsonl\n    type: sharegpt  \n    conversation: chatml\n  - path: /workspace/datasets/dolphin-2.9.3/toolbench_instruct_j1s1_3k_unfiltered.jsonl\n    type: sharegpt  \n    conversation: chatml\n  - path: /workspace/datasets/dolphin-2.9.3/toolbench_negative_unfiltered.jsonl\n    type: sharegpt\n    conversation: chatml\n  - path: /workspace/datasets/dolphin-2.9.3/toolbench_react_10p_unfiltered.jsonl\n    type: sharegpt\n    conversation: chatml\n  - path: /workspace/datasets/dolphin-2.9.3/toolbench_tflan_cot_30p_unfiltered.jsonl\n    type: sharegpt\n    conversation: chatml\n  - path: /workspace/datasets/dolphin-2.9.3/openhermes200k_unfiltered.jsonl\n    type: sharegpt \n    conversation: chatml\n\nchat_template: chatml\n# adapter: qlora\n# lora_r: 128\n# lora_alpha: 16\n# lora_modules_to_save: [embed_tokens, lm_head]\n# lora_dropout: 0.05\n# lora_target_linear: true\n\n\nunfrozen_parameters:\n- ^lm_head.weight$\n- ^model.embed_tokens.weight$\n- input_layernorm\n- model.norm\n- post_attention_layernorm\n- self_attn.rotary_emb\n# mlp.down_proj layers\n- model.layers.0.mlp.down_proj\n- model.layers.1.mlp.down_proj\n- model.layers.4.mlp.down_proj\n- model.layers.37.mlp.down_proj\n- model.layers.24.mlp.down_proj\n- model.layers.2.mlp.down_proj\n- model.layers.38.mlp.down_proj\n- model.layers.35.mlp.down_proj\n- model.layers.25.mlp.down_proj\n- model.layers.6.mlp.down_proj\n- model.layers.22.mlp.down_proj\n- model.layers.23.mlp.down_proj\n- model.layers.3.mlp.down_proj\n- model.layers.21.mlp.down_proj\n- model.layers.5.mlp.down_proj\n- model.layers.28.mlp.down_proj\n- model.layers.20.mlp.down_proj\n- model.layers.26.mlp.down_proj\n- model.layers.19.mlp.down_proj\n- model.layers.34.mlp.down_proj\n# mlp.gate_proj layers\n- model.layers.2.mlp.gate_proj\n- model.layers.1.mlp.gate_proj\n- model.layers.3.mlp.gate_proj\n- model.layers.5.mlp.gate_proj\n- model.layers.4.mlp.gate_proj\n- model.layers.35.mlp.gate_proj\n- model.layers.36.mlp.gate_proj\n- model.layers.37.mlp.gate_proj\n- model.layers.38.mlp.gate_proj\n- model.layers.34.mlp.gate_proj\n- model.layers.33.mlp.gate_proj\n- model.layers.8.mlp.gate_proj\n- model.layers.32.mlp.gate_proj\n- model.layers.6.mlp.gate_proj\n- model.layers.28.mlp.gate_proj\n- model.layers.26.mlp.gate_proj\n- model.layers.30.mlp.gate_proj\n- model.layers.23.mlp.gate_proj\n- model.layers.29.mlp.gate_proj\n- model.layers.27.mlp.gate_proj\n# mlp.up_proj layers\n- model.layers.3.mlp.up_proj\n- model.layers.4.mlp.up_proj\n- model.layers.6.mlp.up_proj\n- model.layers.2.mlp.up_proj\n- model.layers.5.mlp.up_proj\n- model.layers.8.mlp.up_proj\n- model.layers.10.mlp.up_proj\n- model.layers.9.mlp.up_proj\n- model.layers.7.mlp.up_proj\n- model.layers.0.mlp.up_proj\n- model.layers.17.mlp.up_proj\n- model.layers.15.mlp.up_proj\n- model.layers.22.mlp.up_proj\n- model.layers.18.mlp.up_proj\n- model.layers.16.mlp.up_proj\n- model.layers.11.mlp.up_proj\n- model.layers.21.mlp.up_proj\n- model.layers.23.mlp.up_proj\n- model.layers.20.mlp.up_proj\n- model.layers.27.mlp.up_proj\n# self_attn.k_proj layers\n- model.layers.30.self_attn.k_proj\n- model.layers.27.self_attn.k_proj\n- model.layers.25.self_attn.k_proj\n- model.layers.33.self_attn.k_proj\n- model.layers.26.self_attn.k_proj\n- model.layers.31.self_attn.k_proj\n- model.layers.35.self_attn.k_proj\n- model.layers.39.self_attn.k_proj\n- model.layers.22.self_attn.k_proj\n- model.layers.24.self_attn.k_proj\n- model.layers.21.self_attn.k_proj\n- model.layers.28.self_attn.k_proj\n- model.layers.23.self_attn.k_proj\n- model.layers.36.self_attn.k_proj\n- model.layers.20.self_attn.k_proj\n- model.layers.37.self_attn.k_proj\n- model.layers.29.self_attn.k_proj\n- model.layers.32.self_attn.k_proj\n- model.layers.16.self_attn.k_proj\n- model.layers.18.self_attn.k_proj\n# self_attn.o_proj layers\n- model.layers.7.self_attn.o_proj\n- model.layers.6.self_attn.o_proj\n- model.layers.9.self_attn.o_proj\n- model.layers.5.self_attn.o_proj\n- model.layers.27.self_attn.o_proj\n- model.layers.26.self_attn.o_proj\n- model.layers.4.self_attn.o_proj\n- model.layers.31.self_attn.o_proj\n- model.layers.8.self_attn.o_proj\n- model.layers.16.self_attn.o_proj\n- model.layers.3.self_attn.o_proj\n- model.layers.10.self_attn.o_proj\n- model.layers.18.self_attn.o_proj\n- model.layers.33.self_attn.o_proj\n- model.layers.17.self_attn.o_proj\n- model.layers.32.self_attn.o_proj\n- model.layers.30.self_attn.o_proj\n- model.layers.2.self_attn.o_proj\n- model.layers.15.self_attn.o_proj\n- model.layers.11.self_attn.o_proj\n# self_attn.q_proj layers\n- model.layers.14.self_attn.q_proj\n- model.layers.11.self_attn.q_proj\n- model.layers.15.self_attn.q_proj\n- model.layers.9.self_attn.q_proj\n- model.layers.8.self_attn.q_proj\n- model.layers.18.self_attn.q_proj\n- model.layers.12.self_attn.q_proj\n- model.layers.13.self_attn.q_proj\n- model.layers.19.self_attn.q_proj\n- model.layers.16.self_attn.q_proj\n- model.layers.10.self_attn.q_proj\n- model.layers.17.self_attn.q_proj\n- model.layers.7.self_attn.q_proj\n- model.layers.5.self_attn.q_proj\n- model.layers.20.self_attn.q_proj\n- model.layers.3.self_attn.q_proj\n- model.layers.26.self_attn.q_proj\n- model.layers.27.self_attn.q_proj\n- model.layers.28.self_attn.q_proj\n- model.layers.33.self_attn.q_proj\n# self_attn.v_proj layers\n- model.layers.27.self_attn.v_proj\n- model.layers.20.self_attn.v_proj\n- model.layers.24.self_attn.v_proj\n- model.layers.25.self_attn.v_proj\n- model.layers.30.self_attn.v_proj\n- model.layers.2.self_attn.v_proj\n- model.layers.23.self_attn.v_proj\n- model.layers.22.self_attn.v_proj\n- model.layers.26.self_attn.v_proj\n- model.layers.33.self_attn.v_proj\n- model.layers.37.self_attn.v_proj\n- model.layers.7.self_attn.v_proj\n- model.layers.4.self_attn.v_proj\n- model.layers.18.self_attn.v_proj\n- model.layers.31.self_attn.v_proj\n- model.layers.17.self_attn.v_proj\n- model.layers.35.self_attn.v_proj\n- model.layers.32.self_attn.v_proj\n- model.layers.21.self_attn.v_proj\n- model.layers.3.self_attn.v_proj\n\n\n\ndataset_prepared_path:  /workspace/axolotl/dolph-2.9.3-nemo-prepared\nval_set_size: 0.01\noutput_dir: /workspace/axolotl/dolphin-2.9.3-mistral-nemo\n\nsequence_len: 8192\nsample_packing: true\npad_to_sequence_len: true\n\nwandb_project: dolphin-2.9.3-Mistral-nemo\nwandb_watch:\nwandb_run_id:\nwandb_log_model:\n\ngradient_accumulation_steps: 16\nmicro_batch_size: 1\nnum_epochs: 3\noptimizer: adamw_torch\nlr_scheduler: cosine\nlearning_rate: 5e-6\ntrain_on_inputs: false\ngroup_by_length: false\nbf16: auto\nfp16:\ntf32:\n\ngradient_checkpointing: true\ngradient_checkpointing_kwargs:\n  use_reentrant: false\nearly_stopping_patience:\nresume_from_checkpoint:\nlogging_steps: 1\nxformers_attention:\nflash_attention: true\n\nwarmup_steps: 100\n# evals_per_epoch: 4\neval_table_size:\nsaves_per_epoch: 1\nsave_total_limit: 2\nsave_steps:\ndebug:\ndeepspeed: deepspeed_configs/zero3_bf16.json\nweight_decay: 0.1\nspecial_tokens:\n  eos_token: \"<|im_end|>\"\n  pad_token: \"<pad>\"\n  bos_token: \"<s>\"\n  unk_token: \"<unk>\"\ntokens:\n  - \"<|im_start|>\"\n\n\n# fsdp:\n#   - full_shard\n#   - auto_wrap\n# fsdp_config:\n#   fsdp_limit_all_gathers: true\n#   fsdp_sync_module_states: true\n#   fsdp_offload_params: true\n#   fsdp_use_orig_params: false\n#   fsdp_cpu_ram_efficient_loading: true\n#   fsdp_transformer_layer_cls_to_wrap: MixtralSparseMoeBlock\n#   fsdp_state_dict_type: FULL_STATE_DICT\n#   fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP\n#   fsdp_sharding_strategy: FULL_SHARD\n#   fsdp_forward_prefetch: false\n#   fsdp_backward_prefetch: BACKWARD_PRE\n```\n\n</details><br>\n\n[<img src=\"https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg\" alt=\"Visualize in Weights & Biases\" width=\"200\" height=\"32\"/>](https://wandb.ai/ehartford/dolphin-2.9.3-Mistral-nemo/runs/c23odyoj)\n# workspace/axolotl/dolphin-2.9.3-mistral-nemo\n\nThis model was trained from scratch on the None dataset.\nIt achieves the following results on the evaluation set:\n- Loss: 0.5605\n\n## Model description\n\nMore information needed\n\n## Intended uses & limitations\n\nMore information needed\n\n## Training and evaluation data\n\nMore information needed\n\n## Training procedure\n\n### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 5e-06\n- train_batch_size: 1\n- eval_batch_size: 1\n- seed: 42\n- distributed_type: multi-GPU\n- num_devices: 8\n- gradient_accumulation_steps: 16\n- total_train_batch_size: 128\n- total_eval_batch_size: 8\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: cosine\n- lr_scheduler_warmup_steps: 100\n- num_epochs: 3\n\n### Training results\n\n| Training Loss | Epoch  | Step | Validation Loss |\n|:-------------:|:------:|:----:|:---------------:|\n| 0.5691        | 1.0162 | 983  | 0.5734          |\n| 0.5335        | 2.0174 | 1968 | 0.5609          |\n| 0.5297        | 2.9639 | 2901 | 0.5605          |\n\n\n### Framework versions\n\n- Transformers 4.43.0.dev0\n- Pytorch 2.2.2+cu121\n- Datasets 2.19.1\n- Tokenizers 0.19.1\n\n### Updated GGUF conversions were provided by KoboldAI\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "generated_from_trainer",
    "axolotl",
    "dataset:cognitivecomputations/Dolphin-2.9",
    "dataset:teknium/OpenHermes-2.5",
    "dataset:m-a-p/CodeFeedback-Filtered-Instruction",
    "dataset:cognitivecomputations/dolphin-coder",
    "dataset:cognitivecomputations/samantha-data",
    "dataset:microsoft/orca-math-word-problems-200k",
    "dataset:Locutusque/function-calling-chatml",
    "dataset:internlm/Agent-FLAN",
    "base_model:mistralai/Mistral-Nemo-Base-2407",
    "base_model:quantized:mistralai/Mistral-Nemo-Base-2407",
    "license:apache-2.0",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 0,
  "downloads": 299,
  "gated": false,
  "private": false,
  "last_modified": "2026-02-23T15:14:08.000Z",
  "created_at": "2026-02-23T15:14:07.000Z",
  "pipeline_tag": "",
  "library_name": ""
}

{
  "_id": "699c6ebf4d48b676036df9f6",
  "id": "eeropai/dolphin-2.9.3-mistral-nemo-12b-gguf",
  "modelId": "eeropai/dolphin-2.9.3-mistral-nemo-12b-gguf",
  "sha": "c1c7bb347fec92dbd9c2115a63211880e437b5ad",
  "createdAt": "2026-02-23T15:14:07.000Z",
  "lastModified": "2026-02-23T15:14:08.000Z",
  "author": "eeropai",
  "downloads": 299,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "",
  "library_name": "",
  "siblings_count": 17
}