City
a57094a3d3
Fix for scales being in BF16
...
The main conversion code shouldn't allow this, but apparently some models still end up with quantized 1D tensors, see #383
2025-12-17 15:52:16 +01:00
City
01f8845bf3
Add VL version of Qwen3 4B as valid arch
...
According to #361 the thinking/abliterated versions work even though the model wasn't trained with them, maybe the vision one does too.
Does not have mmproj support, since nothing uses that atm.
2025-12-01 01:48:19 +01:00
City
79379af338
Add tokenizer convert logic for Mistral
...
#367
2025-11-29 22:21:30 +01:00
City
0c9edef33b
Improve invalid tenc file message
2025-11-29 22:20:13 +01:00
City
fa48477de8
Qwen3-4B support for Z-Image
...
for #368
2025-11-26 23:22:52 +01:00
City
d4fbdb0139
Add type string for older mmproj files
2025-11-04 07:04:11 +01:00
City
d247022e3f
Better mmproj detection logic
...
#317
2025-08-20 00:51:28 +02:00
City
0ca2aeb797
Qwen2-VL vision support
...
#317
2025-08-19 17:50:55 +02:00
City
cf0573351a
Qwen2-VL text model support
2025-08-05 15:34:19 +02:00
City
79ba1d4b17
Qwen image arch string
2025-08-05 12:44:57 +02:00
spawner
b1350f6f5a
Merge pull request #304 from spawner1145/main
...
support lumina-image-2.0
2025-08-05 12:42:40 +02:00
City
b3ec875a68
Cosmos convert logic
2025-06-15 00:09:45 +02:00
City
6570efec69
Better warning for incompatible models
...
#256
2025-05-07 05:13:08 +02:00
City
54a4854e0c
Allow embedding shape mismatch
...
#269
2025-04-22 15:50:05 +02:00
City
9d282bcdfd
LLaMA OOM hotfix
2025-04-21 04:25:30 +02:00
City
2ef5a66a9f
HiDream support
...
#248
2025-04-16 01:43:15 +02:00
City
794fc28536
Clean up logging
...
Also fixes numpy warning
2025-04-13 23:01:15 +02:00
City
bc5223b0e3
Fix qtype logging on newer gguf-py
2025-03-16 20:14:33 +01:00
City
4aebf1bf2c
Add compat mode pig
...
closes #230
2025-03-16 19:38:40 +01:00
City
31dffa061b
Add error message for missing packages
2025-03-14 17:37:14 +01:00
City
296b0c1419
Clean up field logic
2025-03-14 17:11:47 +01:00
City
d99b2d931f
force dequant on umT5 to prevent OOM
...
not the best solution but that tensor alone needs ~3GB of runtime memory to dequant
2025-03-14 16:56:48 +01:00
City
01f6e3ff63
spiece tokenizer from gguf metadata
2025-03-05 04:44:53 +01:00
City
298192ed60
Wan model arch
2025-02-26 07:40:44 +01:00
City
c6ad4f2363
Account for not having quantized weights
...
#188
2024-12-22 19:00:03 +01:00
City
51fa2cbb5c
Account for dequantization VRAM to avoid OOM
...
#187
2024-12-22 18:09:51 +01:00
City
8d1b56c5e2
Loader logic for LLaMA models
...
Tokenizer shape may be mismatched depending which model was used as source (128320 VS 128256) - warning instead of error for now.
k and q attn keys need to be reshaped back to HF format (inverse of convert_hf_to_gguf.py logic)
2024-12-20 13:39:17 +01:00
City
0a1723baae
Split loader into separate file
2024-12-20 13:03:27 +01:00
City