vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
3447 commits 380 branches 3056 tags 365 MiB
Commit graph

3447 commits

This branch
This branch
All branches
Author SHA1 Message Date
Brian
455c0e53ac Apply suggestions from code review 
Co-authored-by: compilade <git@compilade.net>
 2024-07-16 06:42:38 +10:00
brian khuu
ccff6c7fb2 convert-*.py: remove reference to uuid generation 2024-07-16 06:42:38 +10:00
Brian
8156835d4a constants.py : Revert removal of backward compatibility KEY_GENERAL_SOURCE_URL 2024-07-16 06:42:38 +10:00
Brian
2c060303a6 Update constants.py : spacing correction 2024-07-16 06:42:38 +10:00
Brian
aa4e5892a0 Update convert_hf_to_gguf.py 
Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
 2024-07-16 06:42:38 +10:00
Brian
60278e4f4d Update convert_hf_to_gguf.py 
Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
 2024-07-16 06:42:38 +10:00
brian khuu
ad217d7249 convert-*.py: remove autogenerated uuid 2024-07-16 06:42:38 +10:00
brian khuu
f2b425c59c convert-*.py: import cast from typing and other refactor 2024-07-16 06:42:38 +10:00
brian khuu
04c4fffdcc convert-*.py: prepare_tensors_for_writing() --> prepare_tensors() 
> Especially since it can be used for other purposes than "for writing", like preparing the tensors to then count and sum all their sizes.

Co-authored-by: compilade <git@compilade.net>
 2024-07-16 06:42:38 +10:00
brian khuu
64707b625c convert-*.py: remove redundant gguf_writer.add_name() calls 2024-07-16 06:42:38 +10:00
brian khuu
f8b5931180 convert-*.py: parameter_class_attribute --> size_label 2024-07-16 06:42:38 +10:00
brian khuu
6eb08ac868 convert-*.py: Removing the redundant metadata is not None from all conditions, and indenting them. 
Co-authored-by: compilade <git@compilade.net>
 2024-07-16 06:42:38 +10:00
brian khuu
4c91d077d2 convert-*.py: cast not required if Metadata.load_metadata_override returned a dict[str, Any] instead of a dict[str, object] 
Co-authored-by: compilade <git@compilade.net>
 2024-07-16 06:42:38 +10:00
Brian
74383ba6d2 Apply suggestions from code review 
Co-authored-by: compilade <git@compilade.net>
 2024-07-16 06:42:38 +10:00
brian khuu
dd14b8fdb1 convert-*.py: pyright type fixes 2024-07-16 06:42:38 +10:00
brian khuu
59a01df784 convert-*.py: refactor per model weight count estimation 2024-07-16 06:42:38 +10:00
brian khuu
2a976e1211 convert-*.py: write_tensors() --> prepare_tensors_for_writing() 2024-07-16 06:42:38 +10:00
brian khuu
fdc5a3fc80 convert-*.py: autogenerate general.uuid if missing 2024-07-16 06:42:35 +10:00
brian khuu
7ecb8f00a0 test: remove test_gguf.py and remove test_generate_any_missing_uuid() 2024-07-16 06:38:40 +10:00
brian khuu
007708e32d gguf_writer.py: generate tensor uuid if missing 2024-07-16 06:38:40 +10:00
brian khuu
4dc8ddd35a convert_hf_to_gguf.py: Remove code that is already in fill_templated_filename() and GGUFWriter() 2024-07-16 06:38:40 +10:00
brian khuu
2f23927d37 convert_hf_to_gguf.py: rebase error correction 2024-07-16 06:38:40 +10:00
brian khuu
5011eefeaf convert_hf_to_gguf.py: optional, dataclass removed from type as it was unused 2024-07-16 06:38:40 +10:00
brian khuu
e9734434bd convert-*.py: Remove self.model_name that was left in since last rebase 2024-07-16 06:38:40 +10:00
brian khuu
eaa47f5546 convert-*.py: separated unit test, hf_repo to repo_url 2024-07-16 06:38:40 +10:00
brian khuu
d060fcdbe2 convert-*.py: adjusted authorship KV store 2024-07-16 06:38:40 +10:00
brian khuu
91e65d9485 convert-*.py: add unittest to metadata class 2024-07-16 06:38:38 +10:00
brian khuu
3625a42061 convert-*.py: add heuristic to directory name fallback 
Also add source_url for huggingface url
 2024-07-16 06:37:42 +10:00
brian khuu
39472a09da convert-*.py: need to include self in per_model_weight_count_estimation() 2024-07-16 06:37:42 +10:00
brian khuu
54918ad14e convert-*.py: refactor parameter weight class 2024-07-16 06:37:42 +10:00
brian khuu
32e80e094c convert-*.py: base_model is actually in spec for model cards 2024-07-16 06:37:42 +10:00
brian khuu
4d5cd0670a convert-*.py: use heuristics to parse _name_or_path 2024-07-16 06:37:42 +10:00
brian khuu
b0553f42da convert-*.py: adjust help message 2024-07-16 06:37:42 +10:00
brian khuu
dd1571211e convert-*.py: add quantized_by and enhance heuristics 2024-07-16 06:37:38 +10:00
brian khuu
5a86dfaa1c convert-*.py: add general.organization to kv store 2024-07-16 06:36:03 +10:00
brian khuu
f7c20793b9 convert-*.py: enable --model-name direct metadata override 2024-07-16 06:36:03 +10:00
brian khuu
b1927eed82 convert-*.py: move per model weight estimation away from util back to main script 
plus some refactoring
 2024-07-16 06:36:03 +10:00
brian khuu
684c604eca convert-*.py: add datasets and language to KV store 2024-07-16 06:36:03 +10:00
brian khuu
0f1d50fab7 convert-*.py: add parameter size class 2024-07-16 06:36:03 +10:00
brian khuu
8f734083dd convert-*.py: add base_version and add tags 2024-07-16 06:36:03 +10:00
brian khuu
b36e391b87 convert-*.py: parse model card in metadata util. Add license_link and license_name to kv store 2024-07-16 06:36:03 +10:00
brian khuu
5c263cb257 convert-*.py: encoding_scheme --> output_type 2024-07-16 06:36:03 +10:00
brian khuu
4d5f18a0e6 convert-*.py: metadata class moved to utility 2024-07-16 06:36:03 +10:00
brian khuu
916872f72f convert-*.py: model card metadata 2024-07-16 06:36:03 +10:00
brian khuu
a42c2b7efc convert-*.py: add basename and finetune metadata 2024-07-16 06:36:03 +10:00
brian khuu
dbb1b471e4 convert-*.py: add --get-outfile command and refactor 2024-07-16 06:36:03 +10:00
brian khuu
d3a936fd0e convert-*.py: licence -> license 2024-07-16 06:36:03 +10:00
Xuan Son Nguyen
97bdd26eee
Refactor lora adapter support (#8332) 
* lora: load to devide buft

* add patch tensor function

* correct tensor patch

* llama_lora_adapter_apply

* correct ggml_backend_tensor_copy

* add llm_build_mm

* fix auto merge

* update based on review comments

* add convert script

* no more transpose A

* add f16 convert

* add metadata check

* add sanity check

* fix ftype

* add requirements

* fix requirements

* fix outfile

* conversion: only allow selected models

* fix types

* cuda : do not use dmmv if the tensor does not have enough cols

* llama : lora fixes

* do not disable mmap with lora

Co-authored-by: slaren <slarengh@gmail.com>

* llm_build_lora_mm_id

* convert_lora : MoE LoRA conversion support

* convert_lora : prefer safetensors, similarly to convert_hf

* convert_hf : simplify modify_tensors for InternLM2

* convert_lora : lazy conversion

* llama : load and use alpha from LoRA adapters

* llama : use llm_build_lora_mm in most model graphs

* auto scale

* Revert "auto scale"

This reverts commit 42415a4874.

* remove redundant params

* Apply suggestions from code review

Co-authored-by: slaren <slarengh@gmail.com>

* change kv metadata

* move add_type to __init__

* convert_hf : move add_type to main()

* convert_lora : use the GGUFWriter from Model instead of overwriting it

---------

Co-authored-by: slaren <slarengh@gmail.com>
Co-authored-by: Francis Couture-Harpin <git@compilade.net>
 2024-07-15 20:50:47 +02:00
Xuan Son Nguyen
4db8f60fe7
fix ci (#8494) 2024-07-15 19:23:10 +02:00
Daniel Bevenius
8fac431b06
ggml : suppress unknown pragma 'GCC' on windows (#8460) 
This commit adds a macro guard to pragma GCC to avoid the following
warning on windows:

```console
C:\llama.cpp\ggml\src\ggml-aarch64.c(17,9): warning C4068:
unknown pragma 'GCC' [C:\lama.cpp\build\ggml\src\ggml.vcxproj]
```
 2024-07-15 15:48:17 +03:00