llama: add initial support for Falcon-H1 model family #14534
CISC
merged
112 commits into
ggml-org:master
from
tiiuae:add-fh1-rebased
v1
991de6cb
push more fixes
f897efda
another fix
71a6848e
fix
03568c93
more fixes
0c93ef6a
minor fix
fdd5cff4
more cleaning on python code
14c37ec0
python fixes
8bea9226
changed precision for multipliers float 32->64
071f4b7f
fixes
50eadc7b
merge
a39a8423
another fix
1415cd87
fix
243e4d1a
pre-norm -> norm
cce35498
fix
22de62cf
Revert "fix"
2fe057cc
Merge branch 'add-fh1-rebased' of https://github.com/tiiuae/llama.cpp…
d22b4ea4
fix
6c7d9e26
small fix ffn_norm
15138df4
Merge branch 'add-fh1-rebased' of https://github.com/tiiuae/llama.cpp…
a6d0067d
try
1fd0574a
mix instead of max
250b4f10
fix vocab size
3ee79839
Merge branch 'add-fh1-rebased' of https://github.com/tiiuae/llama.cpp…
2aa48dd8
conflict solve
9760c8bc
fixed multipliers
7a25441e
falcon-h1 specefic vocab resolved
280dd2dc
read arch from gguf.MODEL_ARCH
c56ec07a
mamba_d_ssm added to d_inner find_hparam
c4af0f3c
remove unused functions from gguf_writer.py
53304c84
override modify_tensors instead of get_tensors
441d8d66
fix conversion and d_inner
6c39e775
added some cb functions for debugging puposes
8c508938
inp_out_ids moved outside of layers loop
49d74209
mup_vec create as float64
97011d7a
fix rope_theta
286e1fa5
Merge branch 'add-fh1-rebased' of https://github.com/tiiuae/llama.cpp…
b3bc1fb2
injected mup
a9f3a63d
clean ups
e96cc733
Merge pull request #1 from tiiuae/injected-mup
3afb2a89
rm extra space
0ad35028
rm unused MAMBA_CHUNK_SIZE
53446f7e
rm unused key
ae937f44
add bos False
b6df0a49
changed ROPE_TYPE
935d46fa
cleaning debugging stuff
624699c5
cleaning debug quant
042e5ff9
fix comment
f74e266f
some cleanups
632861e6
some cleanups
084873c2
younesbelkada
commented
on 2025-07-07
Update src/llama-model-loader.cpp
fd203302
more cleanups
68cb7845
moe cleanuips
d2f46f18
d_ssm -> d_inner;
7d7da0b3
cleaning unused hparams
67b26642
Merge branch 'add-fh1-rebased' of https://github.com/tiiuae/llama.cpp…
da8a3385
cleanup
e63ee464
more cleanups
d473d428
more cleanups on python conversion;
8555ee8b
minor cleanups
7846c67e
Apply suggestions from code review
2dee7cf9
remove todo
a846d023
Merge branch 'add-fh1-rebased' of https://github.com/tiiuae/llama.cpp…
f028a43a
Merge branch 'add-fh1-rebased' of https://github.com/tiiuae/llama.cpp…
d41f1114
added falcon-h1
f266d145
tensor not required
4bc9e0ca
clean
2834a4ac
remove unneeded attributes
823696ba
more cleanups and fixed conversion
adff470c
remove final_norm
097df0ed
ggerganov
approved these changes
on 2025-07-08
flake8 fixes
9a048d8d
Merge branch 'add-fh1-rebased' of https://github.com/tiiuae/llama.cpp…
52d1ef35
CISC
commented
on 2025-07-08
Update src/llama-model.cpp
58e3866d
Merge branch 'master' into add-fh1-rebased
d28c31a9
flake8 fixes
9b926483
Update src/llama-hparams.cpp
7fe1794c
Update src/llama-model.cpp
40058c04
Update src/llama-model.cpp
debf4e5d
Update src/llama-arch.cpp
212edffd
Update convert_hf_to_gguf.py
90ddf241
CISC
commented
on 2025-07-08
Merge branch 'add-fh1-rebased' of https://github.com/tiiuae/llama.cpp…
7edf3800
added hashes
c3c5d51c
Update src/llama-arch.cpp
f8d7c970
Update src/llama-vocab.cpp
4610ee20
update the update file
082ab4ad
Revert "update the update file"
c5515e36
fix: address suggestions
1ef53b31
fix: update convert_hf_to_gguf.py
d5efbd01
CISC
commented
on 2025-07-08
Update gguf-py/gguf/constants.py
a5afc8bc
CISC
commented
on 2025-07-08
Update src/llama-model-loader.cpp
99f9a3dd
d_inner fixed
c3c64c35
CISC
commented
on 2025-07-08
Update src/llama-model.cpp
63e3afc9
CISC
approved these changes
on 2025-07-08
reshaping ssm_norm for 34B
d7585783
Merge branch 'add-fh1-rebased' of https://github.com/tiiuae/llama.cpp…
8972c15f
removing generate_mup
7897c217
remove duplicates metadata keys
6403caae
rm comment
710630a5
compilade
commented
on 2025-07-08
Merge branch 'master' into add-fh1-rebased
7b9aa7ba
final comment
ecc5253d
fix unused args
bbca33eb
fix constants
9f514e39
CISC
commented
on 2025-07-08
fix bad merge
34c5d830
Update src/llama-model.cpp
521e8235
falcon-h1: remove unused ssm_in_b and bad merge
6943f4ea
compilade
approved these changes
on 2025-07-08
CISC
commented
on 2025-07-08
Update src/llama-model.cpp
4d2c94b7
falcon-h1: fix last comment
b7c9a995
CISC
commented
on 2025-07-08
compilade
requested changes
on 2025-07-08
Update convert_hf_to_gguf.py
9fd308d8
falcon-h1: revert add_add_bos(False)
51f50bfb
compilade
approved these changes
on 2025-07-09
CISC
commented
on 2025-07-09
falcon-h1: fix tied weights
367d8c58
falcon-h1: remove whitespace
1fa361bb
falcon-h1: fix wrong size param
6dde986a
falcon-h1: fix whitespace issues
94ab3a88
gabe-l-hart
commented
on 2025-07-09
Reviewers
Assignees
No one assigned
Labels
python
hot
Milestone
No milestone
Login to write a write a comment.
Login via GitHub