feat: add Q8_0 block quantization for SmolVLA weight loading - #1
Open
crustopher-lgtm wants to merge 1 commit into
Open
feat: add Q8_0 block quantization for SmolVLA weight loading#1crustopher-lgtm wants to merge 1 commit into
crustopher-lgtm wants to merge 1 commit into
Conversation
- Add vla_block_q8_0 struct matching GGML block_q8_0 layout (fp16 scale + 32 int8) - Add quantize_row_q8_0 / dequantize_row_q8_0 helpers - Add 'q8_0' dtype support in resolve_weight_dtype() and backend_set_from_f32() - Add 7 unit tests for Q8_0 roundtrip accuracy (random, constant, zeros, small values, large range, negatives, symmetry) - Register test_q8_quant with CTest via CMakeLists.txt
crustopher-lgtm
force-pushed
the
feat/q8_0-weight-quantization
branch
from
July 25, 2026 17:01
27d755c to
9add03c
Compare
3 tasks
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Adds Q8_0 block quantization support to SmolVLA weight loading. Weights can now be stored as int8 quantized matrices (fp16 scale + 32 int8 values per block), reducing memory by 42.5% with 3.56x speedup and R²=0.9995 output correlation vs BF16 baseline.
Changes
Usage
Set env var
VLA_WEIGHT_DTYPE=q8_0before loading the model. Default remains BF16.Verification