Thiago Crepaldi

'StableDiffusionXLPipeline' object has no attribute 'named_parameters' when using accelerate.load_checkpoint_and_dispatch

4

### Describe the bug How can I make the stable diffusion xl model to work with the `init_empty_weights` API? I've followed https://huggingface.co/docs/accelerate/v0.11.0/en/big_modeling but still got an unexpected error: ### Reproduction...

[Feature request] Mechanism to Disable Accerate in runtime

9

### System Info ```Shell main branch ``` ### Information - [ ] The official example scripts - [ ] My own modified scripts ### Tasks - [ ] One of...

Fix llama_v2_7b_16h for torch.jit.trace

1

Original error: Attention using SDPA can not be traced with torch.jit.trace when no attention_mask is provided. To solve this issue, please either load your model with the argument attn_implementation="eager" or...

[Feature request] Integrate Python-only API to ONNX optimizer proposal

2

Hello @daquexian, I am with Microsoft developing the PyTorch ONNX Exporter project. Currently we are working on a PyTorch Dynamo-based ONNX exporter (`torch.onnx.dynamo_export`) which will replace the existing Torchscript-based ONNX...

Skip ONNX optimization when it fails

7

During ONNX export, model optimization is performed. Currently, if optimization fails, the export also fails This PR allows export to succeed even when the optimization fails. This is useful to...

datettime.now() is not supported by Dynamo

3

### 🐛 Describe the bug I am actually not sure what is the best behavior here. On one hand, when `datetime.now()` is assigned to a variable that is only used...

Fix bfloat16 serialization for ONNXProgram.save

2

Fixes #124973 Blocked by https://github.com/microsoft/onnxscript/issues/1471

### 🚀 Descirbe the improvement or the new tutorial ghstack conveniently submit stacks of diffs to GitHub as separate pull requests. This way, if we are writing a multi-posts tutorial,...

# Bug Report ### Is the issue related to model conversion? No ### Describe the bug Schema Error when creating a new opseet version for an existing op in defs/tradining...

Add Gradient-2 with bfloat16

3

Repro for bug https://github.com/onnx/onnx/issues/6098 Follow-up for https://github.com/onnx/onnx/pull/6079