feat: auto fit tensors across devices to guarantee optimal load#1736
Merged
Conversation
A --backend module assignment can now list several devices separated by '&'. The module's transformer blocks are partitioned into contiguous ranges sized proportionally to each device's free memory (minus a fixed compute headroom) and registered with the ModelManager with per-tensor compute backends; the existing allocation/staging/LoRA/residency machinery handles the weights unchanged. The module's graphs execute on a ggml_backend_sched spanning the devices, pinning each node to the device of the most recently consumed weight (view ops are never pinned) and splitting each graph exactly once. Supported for the diffusion and te modules; for te the dominant encoder (t5xxl or the LLM) splits while small sub-runners stay on the main device. Graph-cut segmentation and --stream-layers are disabled for split modules. Adds --list-devices to print the ggml device names accepted by the backend specs. Manual placement only; row/tensor split and auto-fit are follow-ups. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…ode row) --split-mode selects how a module assigned multiple runtime devices distributes its weights: layer (default) or row. In row mode the module keeps executing on its main device while its transformer-block matmul weights are allocated in the backend's row-split buffer type (resolved through the "ggml_backend_split_buffer_type" proc, CUDA only for now), which slices each weight's rows across the devices in proportion to free memory and runs the matmuls multi-GPU internally. The ModelManager owns the split buffer types (set_split_buffer_type): params_buffer_type_for returns the split type for eligible tensors when params live on the compute backend, and the staging path groups by (backend, buffer type) and allocates with ggml_backend_alloc_ctx_tensors_from_buft so cpu/disk params residency stages straight into split buffers. Eligibility is limited to contiguous 2D weights of at least 256x256 inside transformer blocks: anything else is consumed by non-matmul ops or sliced into views, which split buffers do not support. Direct LoRA application skips row-split tensors; the automatic LoRA mode selects at_runtime when row split is active. Falls back to a layer split when the backend has no split buffer type or the devices span registries. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
--auto-fit derives the diffusion/te/vae placements from the model metadata and per-device memory budgets, then feeds them through the existing --backend / --params-backend assignment mechanism (the plan and the emitted specs are printed). Budgets reuse --max-vram: positive values cap a device, negative values mean free memory minus that many GiB, unset defaults to free minus a 512 MiB margin. When all components fit resident they are spread across the GPUs; otherwise the plan switches to time-share, giving the heavy components disk params residency (load per phase, free after) and splitting a component too large for any single device across all GPUs via the layer/row split mechanisms (--split-mode still selects which). Components that fit nowhere fall back to the CPU. A VAE decode that still runs out of memory retries once with tiling enabled (temporal for the LTX video VAE, spatial otherwise). Backend initialization now happens after model metadata is loaded so the planner can size components before any backend exists. auto_fit defaults to off. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
# Conflicts: # docs/backend.md # examples/common/common.cpp # examples/common/common.h # include/stable-diffusion.h # src/conditioning/conditioner.hpp # src/core/ggml_extend.hpp # src/core/ggml_extend_backend.cpp # src/core/ggml_extend_backend.h # src/core/util.cpp # src/model_manager.cpp # src/model_manager.h # src/stable-diffusion.cpp
Owner
|
Thank you for your contribution. |
This was referenced Jul 4, 2026
Closed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Part 3 of the split from #1470, contains the auto-fitting logic.
Requires #1734, #1735
Related Issue / Discussion
#1470