-
As asked in the title, is there any way to convert a .pkl model into .gguf directly?
_Originally posted by @ElaineWu66 in https://github.com/ggerganov/llama.cpp/discussions/8790_
-
There are a few commonly used transformer-based optimizations using custom-ops in onnxruntime some of which are listed.
**Custom ops to be converted to functions for ONNX Script** (will add more cu…
-
Just one tip. the proposed environment doesn't work for me. So I build newest environment with cu 12 and transformer 4.24.0 clip-anytorch 2.6.0. etc. At the beginning there is something wrong with tor…
-
python export_qwen2_1.5.py -m /media/yanxiao/机械硬盘1/LLM/Qwen2-7B-Instruct -o ./
WARNING:root:*** Note: please apply modications to model before conversion:
modication 1: in Qwen2ForCausalLM.forwar…
-
### Question
Hi there,
### Thanks for all these great sharing.
**We're trying to evaluate TensorRT performance with Salience-DETR (based on `FocalNet`) so far.
Here're the experiments we have …
-
in GRATIS/Graph Data/Graph Classification/nets/superpixels_graph_classification/gated_gcn_net.py line 40:
`kfea = kfea.unsqueeze(1).repeat(1, N, 1) #[B,N,D]`
that means Key Matrix has the same r…
-
Thank you for your amazing work! We are trying to replicate your results. The paper mentions
> "CaCao can generate extra training data D automatically in a low-resource way and flexibly extend the…
-
**Is your feature request related to a problem? Please describe.**
[Rule-based graph transformer](https://github.com/microsoft/onnxruntime/blob/master/include/onnxruntime/core/optimizer/rule_based_gr…
-
LLM extracts nodes, but it misinterprets their meaning.
We would like to have an "Event" node with a "date" property for historical events, but llm-graph-builder generates nodes for the date values…
-
### System Info
```shell
optimum==1.19.0.dev0
torch==2.1.2
onnx==1.16.0
onnxruntime==1.18.0
cuda==11.8
optimum from mht-sharma:add_llava
```
### Who can help?
@mht-sharma @xenova
…