sparseml

sparseml copied to clipboard

Bumps [tqdm](https://github.com/tqdm/tqdm) from 4.61.1 to 4.66.3. Release notes Sourced from tqdm's releases. tqdm v4.66.3 stable cli: eval safety (fixes CVE-2024-34062, GHSA-g7vv-2v7x-gj9p) tqdm v4.66.2 stable pandas: add DataFrame.progress_map (#1549) notebook: fix...

dependabot[bot]

This PR incorporates changes from @abhinavnmagic's PR https://github.com/neuralmagic/sparseml/pull/2222 into new modifier UX We introduce a new argument `preserve_sparsity_mask` in `SparseGPTModifier` which can be used to extend or ignore the base...

rahul-tuli

Recently a bug was revealed, where if GPTQ modifier was applied consecutively after SparseGPT, the weight sparsity mask was not being respected, this PR fixes that by preserving the mask,...

rahul-tuli

This pull request introduces an integration check to ensure the preservation of mask structure across consecutive runs. The process includes: - **Initial pruning of the model** using a mask structure:...

rahul-tuli

* Update e2e regression tests for channelwise scale and zero-point, added channelwise recipe * Refactored 1.1b test to run on a nightly cadence, 15M will run on commit

Satrat

## Feature Description Now this executes properly: ```python from sparseml.transformers import SparseAutoModelForCausalLM model = SparseAutoModelForCausalLM.from_pretrained("microsoft/Phi-3-mini-128k-instruct",trust_remote_code=True) print(model.__class__.__name__) >> 'Phi3ForCausalLM' ``` The hack was to temporarily rename the class so that the...

dbogunowicz

activation ordering

1

Activation Ordering implementation. Checked lm_eval value with `actorder=True` ``` "metrics":[{"name":"word_perplexity,none","value":10.17568878732032}, ```

horheynm

Hi, while evaluating the performance of the quantized v8 models I realized that the current exporting pipeline does something slightly different from how the models were actually exported for the...

yoloyash

Sparsify custom pytorch models from scratch

7

Hi, I have a model that cannot be traced back to any of the default supported architectures (yolo, llm, trasformers...). I would like to see the benefits of sparsification on...

anberto

Remove QuantizeLinear/DequantizeLinear of ONNX model

4

Hi, I trained YOLOv8 model and exported the model to ONNX format by the quantization_recipe below, I set weight_bits=8 and activation_bits=8 to ensure the full-flow inference of quantized model is...

hoangtv2000