Text Generation
Transformers
PyTorch
English
burt-imma
custom-architecture
matrix-memory
equilibrium-propagation
cifg
sovereign
snapkitty
no-backprop
formal-verification
lean4
Instructions to use Snapkitty/burt-imma with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Snapkitty/burt-imma with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Snapkitty/burt-imma")# Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("Snapkitty/burt-imma", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Snapkitty/burt-imma with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Snapkitty/burt-imma" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Snapkitty/burt-imma", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/Snapkitty/burt-imma
- SGLang
How to use Snapkitty/burt-imma with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "Snapkitty/burt-imma" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Snapkitty/burt-imma", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "Snapkitty/burt-imma" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Snapkitty/burt-imma", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use Snapkitty/burt-imma with Docker Model Runner:
docker model run hf.co/Snapkitty/burt-imma
| // BURT-IMMA Constraint Graph | |
| // Visualize with: dot -Tpng constraint_graph.dot -o constraint_graph.png | |
| // Or: dot -Tsvg constraint_graph.dot -o constraint_graph.svg | |
| digraph BURT_IMMA { | |
| // Global graph attributes | |
| rankdir=TB; | |
| fontname="Helvetica"; | |
| fontsize=12; | |
| label="BURT-IMMA Constraint Graph\nProject: BURT-IMMA | Contact: jessica@collectivekitty.com | License: BSL-1.1"; | |
| labelloc=t; | |
| compound=true; | |
| nodesep=0.8; | |
| ranksep=1.0; | |
| // Node defaults | |
| node [ | |
| shape=box, | |
| style="rounded,filled", | |
| fontname="Helvetica", | |
| fontsize=10, | |
| fillcolor="#E8F4FD", | |
| color="#2C3E50" | |
| ]; | |
| // Edge defaults | |
| edge [ | |
| fontname="Helvetica", | |
| fontsize=8, | |
| color="#34495E" | |
| ]; | |
| // ============================================= | |
| // Subgraph: Feature Extraction | |
| // ============================================= | |
| subgraph cluster_feature_extraction { | |
| label="Feature Extraction"; | |
| style="rounded,dashed"; | |
| color="#3498DB"; | |
| fontcolor="#3498DB"; | |
| fontsize=11; | |
| Input [ | |
| label="Input\n[batch, seq_len]", | |
| shape=ellipse, | |
| fillcolor="#AED6F1" | |
| ]; | |
| SmoothLeaky [ | |
| label="SmoothLeaky\nActivation\n\nk=1.0, alpha=0.01\nSmooth for EP", | |
| fillcolor="#D4EFDF" | |
| ]; | |
| GatesNorm [ | |
| label="GatesNorm\nNormalization\n\ngamma, beta learnable\nPer-expert dim", | |
| fillcolor="#D4EFDF" | |
| ]; | |
| } | |
| // ============================================= | |
| // Subgraph: Memory System | |
| // ============================================= | |
| subgraph cluster_memory_system { | |
| label="Memory System"; | |
| style="rounded,dashed"; | |
| color="#E74C3C"; | |
| fontcolor="#E74C3C"; | |
| fontsize=11; | |
| CIFGMemory [ | |
| label="CIFGMemory\n\nC_global: [d, d]\nC_expert: [K, d_e, d_e]\nf = sigma(W_f * [h,x])\nC_new = f*C + (1-f)*candidate", | |
| fillcolor="#FADBD8", | |
| shape=box3d | |
| ]; | |
| } | |
| // ============================================= | |
| // Subgraph: Routing | |
| // ============================================= | |
| subgraph cluster_routing { | |
| label="Routing (Sparse MoE)"; | |
| style="rounded,dashed"; | |
| color="#F39C12"; | |
| fontcolor="#F39C12"; | |
| fontsize=11; | |
| GatesRouter [ | |
| label="GatesRouter\n\ntop_k=2\nentropy_bound=0.20\nL1 sparsity", | |
| fillcolor="#FEF9E7" | |
| ]; | |
| Expert1 [ | |
| label="Expert 1\n[d_expert=256]", | |
| fillcolor="#FDF2E9" | |
| ]; | |
| Expert2 [ | |
| label="Expert 2\n[d_expert=256]", | |
| fillcolor="#FDF2E9" | |
| ]; | |
| Expert3 [ | |
| label="Expert 3\n[d_expert=256]", | |
| fillcolor="#FDF2E9" | |
| ]; | |
| Expert4 [ | |
| label="Expert 4\n[d_expert=256]", | |
| fillcolor="#FDF2E9" | |
| ]; | |
| } | |
| // ============================================= | |
| // Subgraph: Output Head | |
| // ============================================= | |
| subgraph cluster_output_head { | |
| label="Output Head"; | |
| style="rounded,dashed"; | |
| color="#8E44AD"; | |
| fontcolor="#8E44AD"; | |
| fontsize=11; | |
| InductionHeads [ | |
| label="Superpositioned\nInduction Heads\n\nnum_heads=8\nCopy pattern detection\nMemory superposition", | |
| fillcolor="#E8DAEF" | |
| ]; | |
| QInterference [ | |
| label="Quantum\nInterference\nResolver\n\nAmplitude combination\nConstructive/destructive", | |
| fillcolor="#E8DAEF" | |
| ]; | |
| Output [ | |
| label="Output\n[batch, seq_len, vocab]\n\nConstrained Softmax\nentropy <= 0.20", | |
| shape=ellipse, | |
| fillcolor="#D2B4DE" | |
| ]; | |
| } | |
| // ============================================= | |
| // Data Flow Edges | |
| // ============================================= | |
| // Feature extraction flow | |
| Input -> SmoothLeaky [ | |
| label="embed + pos_encode" | |
| ]; | |
| SmoothLeaky -> GatesNorm [ | |
| label="activated features" | |
| ]; | |
| // Memory interaction | |
| GatesNorm -> CIFGMemory [ | |
| label="h (hidden state)", | |
| style=bold | |
| ]; | |
| CIFGMemory -> InductionHeads [ | |
| label="memory context\nC_global @ h", | |
| color="#E74C3C" | |
| ]; | |
| // Routing flow | |
| GatesNorm -> GatesRouter [ | |
| label="routing logits" | |
| ]; | |
| GatesRouter -> Expert1 [ | |
| label="alpha_1", | |
| style=dashed, | |
| constraint=false | |
| ]; | |
| GatesRouter -> Expert2 [ | |
| label="alpha_2", | |
| style=dashed, | |
| constraint=false | |
| ]; | |
| GatesRouter -> Expert3 [ | |
| label="alpha_3", | |
| style=dashed, | |
| constraint=false | |
| ]; | |
| GatesRouter -> Expert4 [ | |
| label="alpha_4", | |
| style=dashed, | |
| constraint=false | |
| ]; | |
| // Expert to interference | |
| Expert1 -> QInterference [ | |
| label="out_1" | |
| ]; | |
| Expert2 -> QInterference [ | |
| label="out_2" | |
| ]; | |
| Expert3 -> QInterference [ | |
| label="out_3" | |
| ]; | |
| Expert4 -> QInterference [ | |
| label="out_4" | |
| ]; | |
| // Expert memory update | |
| Expert1 -> CIFGMemory [ | |
| label="update C_expert[1]", | |
| style=dotted, | |
| color="#E74C3C", | |
| constraint=false | |
| ]; | |
| Expert2 -> CIFGMemory [ | |
| label="update C_expert[2]", | |
| style=dotted, | |
| color="#E74C3C", | |
| constraint=false | |
| ]; | |
| // Induction heads | |
| GatesNorm -> InductionHeads [ | |
| label="Q, K, V projections" | |
| ]; | |
| // Output assembly | |
| InductionHeads -> QInterference [ | |
| label="attention output", | |
| style=bold | |
| ]; | |
| QInterference -> Output [ | |
| label="resolved output", | |
| style=bold | |
| ]; | |
| // Feedback (memory write from output) | |
| Output -> CIFGMemory [ | |
| label="memory write\n(CIFG gated)", | |
| style=dotted, | |
| color="#E74C3C", | |
| dir=back, | |
| constraint=false | |
| ]; | |
| // ============================================= | |
| // Constraint Annotations | |
| // ============================================= | |
| // Constraint nodes (diamond shape) | |
| node [ | |
| shape=diamond, | |
| fillcolor="#FDEBD0", | |
| color="#D35400", | |
| fontsize=9, | |
| width=1.5, | |
| height=0.8 | |
| ]; | |
| C_entropy [ | |
| label="ENTROPY\n<= 0.20 nats" | |
| ]; | |
| C_spectral [ | |
| label="SPECTRAL\nsigma_max <= 0.95" | |
| ]; | |
| C_huntington [ | |
| label="HUNTINGTON\nPostulates" | |
| ]; | |
| // Constraint edges | |
| C_entropy -> GatesRouter [ | |
| style=bold, | |
| color="#D35400", | |
| label="enforced", | |
| dir=both | |
| ]; | |
| C_entropy -> Output [ | |
| style=bold, | |
| color="#D35400", | |
| label="enforced", | |
| dir=both | |
| ]; | |
| C_spectral -> CIFGMemory [ | |
| style=bold, | |
| color="#D35400", | |
| label="sigma_max(C)", | |
| dir=both | |
| ]; | |
| C_spectral -> Expert1 [ | |
| style=bold, | |
| color="#D35400", | |
| label="sigma_max(W)", | |
| dir=both | |
| ]; | |
| C_spectral -> InductionHeads [ | |
| style=bold, | |
| color="#D35400", | |
| label="sigma_max(W_QKV)", | |
| dir=both | |
| ]; | |
| C_huntington -> QInterference [ | |
| style=bold, | |
| color="#D35400", | |
| label="Boolean lattice", | |
| dir=both | |
| ]; | |
| // ============================================= | |
| // Legend | |
| // ============================================= | |
| subgraph cluster_legend { | |
| label="Legend"; | |
| style="rounded"; | |
| color="#7F8C8D"; | |
| fontcolor="#7F8C8D"; | |
| fontsize=10; | |
| node [shape=plaintext, fillcolor=white, fontsize=8]; | |
| legend [label=< | |
| <TABLE BORDER="0" CELLBORDER="1" CELLSPACING="0" CELLPADDING="4"> | |
| <TR><TD COLSPAN="2"><B>Edge Styles</B></TD></TR> | |
| <TR><TD>Solid bold</TD><TD>Primary data flow</TD></TR> | |
| <TR><TD>Dashed</TD><TD>Routing decisions</TD></TR> | |
| <TR><TD>Dotted (red)</TD><TD>Memory updates</TD></TR> | |
| <TR><TD>Bold (orange)</TD><TD>Constraint enforcement</TD></TR> | |
| <TR><TD COLSPAN="2"><B>Constraints</B></TD></TR> | |
| <TR><TD>entropy_bound</TD><TD>0.20 nats</TD></TR> | |
| <TR><TD>lambda_max</TD><TD>0.95</TD></TR> | |
| <TR><TD>huntington</TD><TD>true (all postulates)</TD></TR> | |
| </TABLE> | |
| >]; | |
| } | |
| } | |