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
| /- | |
| AnuQuantumInterference | |
| Phase mask operations with constructive/destructive interference | |
| -/ | |
| import Mathlib | |
| noncomputable section | |
| open Real | |
| -- ============================================================ | |
| -- Core Definitions | |
| -- ============================================================ | |
| /-- Apply a phase mask: true = constructive (keep), false = destructive (negate) -/ | |
| def apply_phase_mask (w : β) (constructive : Bool) : β := | |
| if constructive then w else -w | |
| /-- Phase shift values for a layer -/ | |
| def phase_shift (layer_idx : Nat) (n_layers : Nat) : β := | |
| 2 * Real.pi * (layer_idx : β) / (n_layers : β) | |
| /-- Perturbation: small noise added to weights -/ | |
| def perturb (w : β) (epsilon : β) : β := | |
| w + epsilon | |
| /-- Pipeline: sequence of phase-masked operations -/ | |
| def pipeline (weights : List β) (masks : List Bool) : List β := | |
| List.zipWith apply_phase_mask weights masks | |
| -- ============================================================ | |
| -- Theorems (6) | |
| -- ============================================================ | |
| /-- Destructive interference negates the weight -/ | |
| theorem destructive_cancellation (w : β) : | |
| apply_phase_mask w false = -w := by | |
| simp [apply_phase_mask] | |
| /-- Constructive interference preserves the weight -/ | |
| theorem constructive_preservation (w : β) : | |
| apply_phase_mask w true = w := by | |
| simp [apply_phase_mask] | |
| /-- Phase shift values are in [0, 2*pi) for valid layer indices -/ | |
| theorem phase_shift_values (layer_idx n_layers : Nat) | |
| (h_valid : layer_idx < n_layers) | |
| (h_pos : n_layers > 0) : | |
| 0 β€ phase_shift layer_idx n_layers β§ | |
| phase_shift layer_idx n_layers < 2 * Real.pi := by | |
| have hβ : (layer_idx : β) β₯ 0 := by exact_mod_cast Nat.zero_le layer_idx | |
| have hβ : (n_layers : β) > 0 := by exact_mod_cast h_pos | |
| have hβ : (layer_idx : β) < (n_layers : β) := by exact_mod_cast h_valid | |
| have hβ : (layer_idx : β) / (n_layers : β) < 1 := by | |
| rw [div_lt_one hβ]; exact_mod_cast h_valid | |
| constructor | |
| Β· simp [phase_shift] | |
| positivity | |
| Β· simp [phase_shift] | |
| have : (layer_idx : β) / (n_layers : β) < 1 := hβ | |
| have hpi : 0 < Real.pi := Real.pi_pos | |
| nlinarith | |
| /-- Small perturbation preserves sign structure -/ | |
| theorem perturbation_preserves_structure (w epsilon : β) | |
| (h_small : |epsilon| < |w|) | |
| (h_w_pos : w > 0) : | |
| perturb w epsilon > 0 := by | |
| have hβ : |w| = w := abs_of_pos h_w_pos | |
| have hβ : |epsilon| < w := by rwa [hβ] at h_small | |
| have hβ : -w < epsilon := (abs_lt.mp hβ).1 | |
| simp only [perturb] | |
| linarith | |
| /-- Pipeline produces output for every input with a matching mask -/ | |
| theorem pipeline_soundness (weights : List β) (masks : List Bool) | |
| (h_len : weights.length = masks.length) : | |
| (pipeline weights masks).length = weights.length := by | |
| simp only [pipeline, List.length_zipWith, h_len, min_self] | |
| /-- Pipeline output fully determined by inputs (no hidden state) -/ | |
| theorem pipeline_completeness (weights1 weights2 : List β) (masks1 masks2 : List Bool) | |
| (h_w : weights1 = weights2) | |
| (h_m : masks1 = masks2) : | |
| pipeline weights1 masks1 = pipeline weights2 masks2 := by | |
| rw [h_w, h_m] | |
| end | |