llama.cpp
convert-hf : save memory with lazy evaluation
#7075
Merged

convert-hf : save memory with lazy evaluation #7075

compilade merged 26 commits into master from compilade/lazy-convert-hf
compilade
compilade convert-hf : begin refactoring write_tensor
47e02eb7
compilade Merge branch 'master' into compilade/convert-hf-refactor
0d720acb
compilade convert : upgrade to sentencepiece v0.2.0
c33775bc
compilade convert-hf : remove unused n_dims in extra_*_tensors
698f0b34
compilade convert-hf : simplify MoE weights stacking
cde9ea65
compilade convert-hf : flake8 linter doesn't like semicolons
56f60f5d
compilade convert-hf : allow unusual model part names
3870164f
compilade convert : use a string for the SentencePiece tokenizer path
dcd8dfa1
compilade convert-hf : display tensor shape
21068b6b
compilade convert-hf : convert norms to f32 by default
639b374b
compilade convert-hf : sort model part names
644c2696
compilade convert-hf : use an ABC for Model again
ce067af1
compilade convert-hf : use a plain class for Model, and forbid direct instantia…
13f4cf70
compilade Merge branch 'master' into compilade/convert-hf-refactor
6a54973d
compilade Merge branch 'master' into compilade/convert-hf-refactor
3e5e0dce
compilade convert-hf : more consistent formatting of cmdline args
98f2d0e0
compilade convert-hf : align the message logged for converted tensors
f2099c50
compilade compilade added enhancement
compilade compilade added need feedback
bartowski1182
compilade convert-hf : fix Refact conversion
215a0d38
compilade convert-hf : save memory with lazy evaluation
f09674fb
compilade convert-hf : flake8 doesn't like lowercase L as a variable name
0c383328
compilade convert-hf : remove einops requirement for InternLM2
98db4347
compilade compilade force pushed from b361d693 to 98db4347 1 year ago
compilade
compilade convert-hf : faster model parts loading
bc78bf4c
bartowski1182
compilade
slaren
compilade convert-hf : minor changes for consistency
62303e7f
compilade
ggerganov
ggerganov ggerganov added high priority
compilade compilade changed the base branch from compilade/convert-hf-refactor to master 1 year ago
compilade Merge branch 'master' into compilade/lazy-convert-hf
68c5ac62
CISC
compilade gguf-py : add tqdm as a dependency
94e667a9
compilade
ggerganov
ggerganov approved these changes on 2024-05-07
Galunid
ggerganov
compilade
Galunid
slaren
ggerganov
Galunid
Galunid approved these changes on 2024-05-08
compilade Merge branch 'master' into compilade/lazy-convert-hf
bffdaf40
compilade
teleprint-me
teleprint-me
compilade
compilade
compilade
ggerganov
compilade compilade force pushed from 1eccde6f to bffdaf40 1 year ago
compilade
compilade compilade force pushed from cad22e17 to bffdaf40 1 year ago
compilade compilade merged f98eb31c into master 1 year ago
bartowski1182
compilade
bartowski1182

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone