Text Generation
MLX
Safetensors
English
Chinese
mimo_v2_flash
apple-silicon
mimo-v2
mixture-of-experts
4-bit precision
mtp
conversational
Instructions to use TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP with MLX:
# Make sure mlx-lm is installed # pip install --upgrade mlx-lm # Generate text with mlx-lm from mlx_lm import load, generate model, tokenizer = load("TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP") prompt = "Write a story about Einstein" messages = [{"role": "user", "content": prompt}] prompt = tokenizer.apply_chat_template( messages, add_generation_prompt=True ) text = generate(model, tokenizer, prompt=prompt, verbose=True) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Pi
How to use TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP with Pi:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP"
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "mlx-lm": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP" } ] } } }Run Pi
# Start Pi in your project directory: pi
- MLX LM
How to use TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP with MLX LM:
Generate or start a chat session
# Install MLX LM uv tool install mlx-lm # Interactive chat REPL mlx_lm.chat --model "TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP"
Run an OpenAI-compatible server
# Install MLX LM uv tool install mlx-lm # Start the server mlx_lm.server --model "TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP" # Calling the OpenAI-compatible server with curl curl -X POST "http://localhost:8080/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP", "messages": [ {"role": "user", "content": "Hello"} ] }' - Hermes Agent
How to use TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP with Hermes Agent:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP"
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP with OpenClaw:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP"
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:8080/v1 \ --custom-model-id "TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP" \ --custom-provider-id mlx-lm \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
Standardize bundled DFlash checkpoint filename
Browse files
README.md
CHANGED
|
@@ -40,7 +40,7 @@ The checkpoint includes its native three-layer MTP payload in `mtp/model_mtp.saf
|
|
| 40 |
| Native component | Path | Format |
|
| 41 |
| --- | --- | --- |
|
| 42 |
| MTP predictor | `mtp/model_mtp.safetensors` | MLX 4-bit affine |
|
| 43 |
-
| DFlash drafter | `dflash/
|
| 44 |
| Vision encoder | `omnimodal/vision_encoder.safetensors` | Upstream BF16 |
|
| 45 |
| Audio encoder | `omnimodal/audio_encoder.safetensors` | Upstream BF16 |
|
| 46 |
| Audio tokenizer | `audio_tokenizer/model.safetensors` | Upstream weights |
|
|
|
|
| 40 |
| Native component | Path | Format |
|
| 41 |
| --- | --- | --- |
|
| 42 |
| MTP predictor | `mtp/model_mtp.safetensors` | MLX 4-bit affine |
|
| 43 |
+
| DFlash drafter | `dflash/model.safetensors` | Upstream BF16 |
|
| 44 |
| Vision encoder | `omnimodal/vision_encoder.safetensors` | Upstream BF16 |
|
| 45 |
| Audio encoder | `omnimodal/audio_encoder.safetensors` | Upstream BF16 |
|
| 46 |
| Audio tokenizer | `audio_tokenizer/model.safetensors` | Upstream weights |
|
dflash/{dflash_draft_model.safetensors → model.safetensors}
RENAMED
|
File without changes
|
dflash/model.safetensors.index.json
CHANGED
|
@@ -3,68 +3,68 @@
|
|
| 3 |
"total_size": 2936114304
|
| 4 |
},
|
| 5 |
"weight_map": {
|
| 6 |
-
"fc.weight": "
|
| 7 |
-
"hidden_norm.weight": "
|
| 8 |
-
"norm.weight": "
|
| 9 |
-
"layers.0.input_layernorm.weight": "
|
| 10 |
-
"layers.0.post_attention_layernorm.weight": "
|
| 11 |
-
"layers.0.self_attn.q_norm.weight": "
|
| 12 |
-
"layers.0.self_attn.k_norm.weight": "
|
| 13 |
-
"layers.0.self_attn.q_proj.weight": "
|
| 14 |
-
"layers.0.self_attn.k_proj.weight": "
|
| 15 |
-
"layers.0.self_attn.v_proj.weight": "
|
| 16 |
-
"layers.0.self_attn.o_proj.weight": "
|
| 17 |
-
"layers.0.mlp.gate_proj.weight": "
|
| 18 |
-
"layers.0.mlp.up_proj.weight": "
|
| 19 |
-
"layers.0.mlp.down_proj.weight": "
|
| 20 |
-
"layers.0.self_attn.attention_sink_bias": "
|
| 21 |
-
"layers.1.input_layernorm.weight": "
|
| 22 |
-
"layers.1.post_attention_layernorm.weight": "
|
| 23 |
-
"layers.1.self_attn.q_norm.weight": "
|
| 24 |
-
"layers.1.self_attn.k_norm.weight": "
|
| 25 |
-
"layers.1.self_attn.q_proj.weight": "
|
| 26 |
-
"layers.1.self_attn.k_proj.weight": "
|
| 27 |
-
"layers.1.self_attn.v_proj.weight": "
|
| 28 |
-
"layers.1.self_attn.o_proj.weight": "
|
| 29 |
-
"layers.1.mlp.gate_proj.weight": "
|
| 30 |
-
"layers.1.mlp.up_proj.weight": "
|
| 31 |
-
"layers.1.mlp.down_proj.weight": "
|
| 32 |
-
"layers.1.self_attn.attention_sink_bias": "
|
| 33 |
-
"layers.2.input_layernorm.weight": "
|
| 34 |
-
"layers.2.post_attention_layernorm.weight": "
|
| 35 |
-
"layers.2.self_attn.q_norm.weight": "
|
| 36 |
-
"layers.2.self_attn.k_norm.weight": "
|
| 37 |
-
"layers.2.self_attn.q_proj.weight": "
|
| 38 |
-
"layers.2.self_attn.k_proj.weight": "
|
| 39 |
-
"layers.2.self_attn.v_proj.weight": "
|
| 40 |
-
"layers.2.self_attn.o_proj.weight": "
|
| 41 |
-
"layers.2.mlp.gate_proj.weight": "
|
| 42 |
-
"layers.2.mlp.up_proj.weight": "
|
| 43 |
-
"layers.2.mlp.down_proj.weight": "
|
| 44 |
-
"layers.2.self_attn.attention_sink_bias": "
|
| 45 |
-
"layers.3.input_layernorm.weight": "
|
| 46 |
-
"layers.3.post_attention_layernorm.weight": "
|
| 47 |
-
"layers.3.self_attn.q_norm.weight": "
|
| 48 |
-
"layers.3.self_attn.k_norm.weight": "
|
| 49 |
-
"layers.3.self_attn.q_proj.weight": "
|
| 50 |
-
"layers.3.self_attn.k_proj.weight": "
|
| 51 |
-
"layers.3.self_attn.v_proj.weight": "
|
| 52 |
-
"layers.3.self_attn.o_proj.weight": "
|
| 53 |
-
"layers.3.mlp.gate_proj.weight": "
|
| 54 |
-
"layers.3.mlp.up_proj.weight": "
|
| 55 |
-
"layers.3.mlp.down_proj.weight": "
|
| 56 |
-
"layers.3.self_attn.attention_sink_bias": "
|
| 57 |
-
"layers.4.input_layernorm.weight": "
|
| 58 |
-
"layers.4.post_attention_layernorm.weight": "
|
| 59 |
-
"layers.4.self_attn.q_norm.weight": "
|
| 60 |
-
"layers.4.self_attn.k_norm.weight": "
|
| 61 |
-
"layers.4.self_attn.q_proj.weight": "
|
| 62 |
-
"layers.4.self_attn.k_proj.weight": "
|
| 63 |
-
"layers.4.self_attn.v_proj.weight": "
|
| 64 |
-
"layers.4.self_attn.o_proj.weight": "
|
| 65 |
-
"layers.4.mlp.gate_proj.weight": "
|
| 66 |
-
"layers.4.mlp.up_proj.weight": "
|
| 67 |
-
"layers.4.mlp.down_proj.weight": "
|
| 68 |
-
"layers.4.self_attn.attention_sink_bias": "
|
| 69 |
}
|
| 70 |
}
|
|
|
|
| 3 |
"total_size": 2936114304
|
| 4 |
},
|
| 5 |
"weight_map": {
|
| 6 |
+
"fc.weight": "model.safetensors",
|
| 7 |
+
"hidden_norm.weight": "model.safetensors",
|
| 8 |
+
"norm.weight": "model.safetensors",
|
| 9 |
+
"layers.0.input_layernorm.weight": "model.safetensors",
|
| 10 |
+
"layers.0.post_attention_layernorm.weight": "model.safetensors",
|
| 11 |
+
"layers.0.self_attn.q_norm.weight": "model.safetensors",
|
| 12 |
+
"layers.0.self_attn.k_norm.weight": "model.safetensors",
|
| 13 |
+
"layers.0.self_attn.q_proj.weight": "model.safetensors",
|
| 14 |
+
"layers.0.self_attn.k_proj.weight": "model.safetensors",
|
| 15 |
+
"layers.0.self_attn.v_proj.weight": "model.safetensors",
|
| 16 |
+
"layers.0.self_attn.o_proj.weight": "model.safetensors",
|
| 17 |
+
"layers.0.mlp.gate_proj.weight": "model.safetensors",
|
| 18 |
+
"layers.0.mlp.up_proj.weight": "model.safetensors",
|
| 19 |
+
"layers.0.mlp.down_proj.weight": "model.safetensors",
|
| 20 |
+
"layers.0.self_attn.attention_sink_bias": "model.safetensors",
|
| 21 |
+
"layers.1.input_layernorm.weight": "model.safetensors",
|
| 22 |
+
"layers.1.post_attention_layernorm.weight": "model.safetensors",
|
| 23 |
+
"layers.1.self_attn.q_norm.weight": "model.safetensors",
|
| 24 |
+
"layers.1.self_attn.k_norm.weight": "model.safetensors",
|
| 25 |
+
"layers.1.self_attn.q_proj.weight": "model.safetensors",
|
| 26 |
+
"layers.1.self_attn.k_proj.weight": "model.safetensors",
|
| 27 |
+
"layers.1.self_attn.v_proj.weight": "model.safetensors",
|
| 28 |
+
"layers.1.self_attn.o_proj.weight": "model.safetensors",
|
| 29 |
+
"layers.1.mlp.gate_proj.weight": "model.safetensors",
|
| 30 |
+
"layers.1.mlp.up_proj.weight": "model.safetensors",
|
| 31 |
+
"layers.1.mlp.down_proj.weight": "model.safetensors",
|
| 32 |
+
"layers.1.self_attn.attention_sink_bias": "model.safetensors",
|
| 33 |
+
"layers.2.input_layernorm.weight": "model.safetensors",
|
| 34 |
+
"layers.2.post_attention_layernorm.weight": "model.safetensors",
|
| 35 |
+
"layers.2.self_attn.q_norm.weight": "model.safetensors",
|
| 36 |
+
"layers.2.self_attn.k_norm.weight": "model.safetensors",
|
| 37 |
+
"layers.2.self_attn.q_proj.weight": "model.safetensors",
|
| 38 |
+
"layers.2.self_attn.k_proj.weight": "model.safetensors",
|
| 39 |
+
"layers.2.self_attn.v_proj.weight": "model.safetensors",
|
| 40 |
+
"layers.2.self_attn.o_proj.weight": "model.safetensors",
|
| 41 |
+
"layers.2.mlp.gate_proj.weight": "model.safetensors",
|
| 42 |
+
"layers.2.mlp.up_proj.weight": "model.safetensors",
|
| 43 |
+
"layers.2.mlp.down_proj.weight": "model.safetensors",
|
| 44 |
+
"layers.2.self_attn.attention_sink_bias": "model.safetensors",
|
| 45 |
+
"layers.3.input_layernorm.weight": "model.safetensors",
|
| 46 |
+
"layers.3.post_attention_layernorm.weight": "model.safetensors",
|
| 47 |
+
"layers.3.self_attn.q_norm.weight": "model.safetensors",
|
| 48 |
+
"layers.3.self_attn.k_norm.weight": "model.safetensors",
|
| 49 |
+
"layers.3.self_attn.q_proj.weight": "model.safetensors",
|
| 50 |
+
"layers.3.self_attn.k_proj.weight": "model.safetensors",
|
| 51 |
+
"layers.3.self_attn.v_proj.weight": "model.safetensors",
|
| 52 |
+
"layers.3.self_attn.o_proj.weight": "model.safetensors",
|
| 53 |
+
"layers.3.mlp.gate_proj.weight": "model.safetensors",
|
| 54 |
+
"layers.3.mlp.up_proj.weight": "model.safetensors",
|
| 55 |
+
"layers.3.mlp.down_proj.weight": "model.safetensors",
|
| 56 |
+
"layers.3.self_attn.attention_sink_bias": "model.safetensors",
|
| 57 |
+
"layers.4.input_layernorm.weight": "model.safetensors",
|
| 58 |
+
"layers.4.post_attention_layernorm.weight": "model.safetensors",
|
| 59 |
+
"layers.4.self_attn.q_norm.weight": "model.safetensors",
|
| 60 |
+
"layers.4.self_attn.k_norm.weight": "model.safetensors",
|
| 61 |
+
"layers.4.self_attn.q_proj.weight": "model.safetensors",
|
| 62 |
+
"layers.4.self_attn.k_proj.weight": "model.safetensors",
|
| 63 |
+
"layers.4.self_attn.v_proj.weight": "model.safetensors",
|
| 64 |
+
"layers.4.self_attn.o_proj.weight": "model.safetensors",
|
| 65 |
+
"layers.4.mlp.gate_proj.weight": "model.safetensors",
|
| 66 |
+
"layers.4.mlp.up_proj.weight": "model.safetensors",
|
| 67 |
+
"layers.4.mlp.down_proj.weight": "model.safetensors",
|
| 68 |
+
"layers.4.self_attn.attention_sink_bias": "model.safetensors"
|
| 69 |
}
|
| 70 |
}
|