Dilshod Tadjibaev

If we handle https://github.com/tracel-ai/burn/issues/1544 fully then this ticket will get resolved too.

I fixed unsqueeze scalar (https://github.com/tracel-ai/burn/pull/1690) issue but it's blocked on converting Gemm node to linear module: ``` [burn-import]% cr ./clip_image_model_vitb32_batchsize1.onnx ./out .... DEBUG burn_import::onnx::from_onnx: output name: /visual/transformer/resblocks/resblocks.0/attn/Reshape_3_output_0 DEBUG burn_import::onnx::proto_conversion: Converting...

We need to review to see if the model changed. If the model does not match the record, you'll have a mismatch. I propose we store a string representation of...

@tiruka, we can use ONNX model since it's more stable but I feel like it changes the nature of the example. We may have a tons of existing references and...

CC: @louisfd, @nathanielsimard, @laggui

If gguf contains the model graph information, then we can use what burn-import ONNX facility. In our burn-import, we convert ONNX graph to IR (intermediate representation) (see [this doc](https://github.com/tracel-ai/burn/blob/main/burn-import/DEVELOPMENT.md)). So,...

From my brief research, GGUF format contains metadata + tensor weights. This aligns with burn-import pytorch route and not burn-import/ONNX. This will mean model needs to be constructed in Burn...

GGUF spec: https://github.com/ggerganov/ggml/pull/302

Parser in Rust: https://github.com/Jimexist/gguf

I will investigate next week.