nicolas
/
hybrid-llama
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
hybrid-llama/turboquant/examples/diffusion
History
Marvin 1dd0700988 hybrid-llama: merge ik_llama IQK CPU GEMM into TurboQuant fork 
Base: AtomicBot-ai/atomic-llama-cpp-turboquant @ cd5609390. IQK source: ikawrakow/ik_llama.cpp @ fe215a8c (ggml/src/iqk only).

- GGML_IQK_MUL_MAT / GGML_IQK_FLASH_ATTENTION options (default OFF)

- 57 IQK repacked types, blocks, traits; IQK hooks in ggml_compute_forward_mul_mat

- ggml-cpu with IQK ON builds and links; IQK OFF build unaffected

Assisted-by: opencode (Muse Spark)
 		2026-09-05 18:09:49 -03:00
..
CMakeLists.txt hybrid-llama: merge ik_llama IQK CPU GEMM into TurboQuant fork 2026-09-05 18:09:49 -03:00
README.md hybrid-llama: merge ik_llama IQK CPU GEMM into TurboQuant fork 2026-09-05 18:09:49 -03:00
diffusion-cli.cpp hybrid-llama: merge ik_llama IQK CPU GEMM into TurboQuant fork 2026-09-05 18:09:49 -03:00
diffusion.cpp hybrid-llama: merge ik_llama IQK CPU GEMM into TurboQuant fork 2026-09-05 18:09:49 -03:00
diffusion.h hybrid-llama: merge ik_llama IQK CPU GEMM into TurboQuant fork 2026-09-05 18:09:49 -03:00

README.md

Diffusion Text Generation

This directory contains implementations for Diffusion LLMs (DLLMs)

More Info:

Parameters

The diffusion CLI supports various parameters to control the generation process:

Core Diffusion Parameters

  • --diffusion-steps: Number of diffusion steps (default: 256)
  • --diffusion-algorithm: Algorithm for token selection
    • 0: DIFFUSION_ALGORITHM_ORIGIN - Token will be generated in a purely random order from https://arxiv.org/abs/2107.03006.
    • 1: DIFFUSION_ALGORITHM_ENTROPY_BASED - Entropy-based selection
    • 2: DIFFUSION_ALGORITHM_MARGIN_BASED - Margin-based selection
    • 3: DIFFUSION_ALGORITHM_RANDOM - Random selection
    • 4: DIFFUSION_ALGORITHM_CONFIDENCE_BASED - Confidence-based selection (default)
    • More documentation here https://github.com/DreamLM/Dream
  • --diffusion-visual: Enable live visualization during generation

Scheduling Parameters

Choose one of the following scheduling methods:

Timestep-based scheduling:

  • --diffusion-eps: Epsilon value for timestep scheduling (e.g., 0.001)

Block-based scheduling:

  • --diffusion-block-length: Block size for block-based scheduling (e.g., 32)

Sampling Parameters

  • --temp: Temperature for sampling (0.0 = greedy/deterministic, higher = more random)
  • --top-k: Top-k filtering for sampling
  • --top-p: Top-p (nucleus) filtering for sampling
  • --seed: Random seed for reproducibility

Model Parameters

  • -m: Path to the GGUF model file
  • -p: Input prompt text
  • -ub: Maximum sequence length (ubatch size)
  • -c: Context size
  • -b: Batch size

Examples

Dream architecture:

llama-diffusion-cli -m dream7b.gguf -p "write code to train MNIST in pytorch" -ub 512 --diffusion-eps 0.001 --diffusion-algorithm 3 --diffusion-steps 256 --diffusion-visual

LLaDA architecture:

llama-diffusion-cli -m llada-8b.gguf -p "write code to train MNIST in pytorch" -ub 512 --diffusion-block-length 32 --diffusion-steps 256 --diffusion-visual

RND1 architecture:

llama-diffusion-cli -m RND1-Base-0910.gguf -p "write code to train MNIST in pytorch" -ub 512 --diffusion-algorithm 1 --diffusion-steps 256 --diffusion-visual --temp 0.5 --diffusion-eps 0.001