Skip to content

Add HQQ quant loader for ooba #4888

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
oobabooga merged 9 commits into from
Dec 19, 2023
Merged

Add HQQ quant loader for ooba #4888

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
81ef984
done
waters222 Dec 11, 2023
f74669b
address review comment
waters222 Dec 12, 2023
f2c0f6a
address review comment
waters222 Dec 14, 2023
829b316
Small adjustments
oobabooga Dec 15, 2023
05819d3
Make the backend customizable
oobabooga Dec 15, 2023
1a9ead5
Merge branch 'dev' into waters222-main
oobabooga Dec 15, 2023
f017b4b
Merge branch 'dev' into waters222-main
oobabooga Dec 19, 2023
165b466
Add hqq to requirements
oobabooga Dec 19, 2023
346f92c
Update README
oobabooga Dec 19, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
38 changes: 38 additions & 0 deletions modules/loaders.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -147,6 +147,11 @@
'trust_remote_code',
'no_use_fast',
'no_flash_attn',
],
'HQQ': [
'trust_remote_code',
'no_use_fast',
'no_flash_attn',
]
})

Expand Down Expand Up @@ -495,6 +500,39 @@
'skip_special_tokens',
'auto_max_new_tokens',
},
'HQQ': {
'temperature',
'temperature_last',
'top_p',
'min_p',
'top_k',
'typical_p',
'epsilon_cutoff',
'eta_cutoff',
'tfs',
'top_a',
'repetition_penalty',
'presence_penalty',
'frequency_penalty',
'repetition_penalty_range',
'encoder_repetition_penalty',
'no_repeat_ngram_size',
'min_length',
'seed',
'do_sample',
'mirostat_mode',
'mirostat_tau',
'mirostat_eta',
'grammar_file_row',
'grammar_string',
'guidance_scale',
'negative_prompt',
'ban_eos_token',
'custom_token_bans',
'add_bos_token',
'skip_special_tokens',
'auto_max_new_tokens',
},
}

loaders_model_types = {
Expand Down
10 changes: 10 additions & 0 deletions modules/models.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -72,6 +72,7 @@ def load_model(model_name, loader=None):
'ctransformers': ctransformers_loader,
'AutoAWQ': AutoAWQ_loader,
'QuIP#': QuipSharp_loader,
'HQQ': HQQ_loader,
}

metadata = get_model_metadata(model_name)
Expand Down Expand Up @@ -404,6 +405,15 @@ def ExLlamav2_HF_loader(model_name):

return Exllamav2HF.from_pretrained(model_name)

def HQQ_loader(model_name):
from hqq.engine.hf import HQQModelForCausalLM
from hqq.core.quantize import HQQLinear, HQQBackend

model_dir = f'{shared.args.model_dir}/{model_name}'
logger.warning(f"loading HQQ model from {model_dir} with HQQLinear backend")
model = HQQModelForCausalLM.from_quantized(model_dir)
HQQLinear.set_backend(HQQBackend.PYTORCH_COMPILE)
return model

def RWKV_loader(model_name):
'''
Expand Down
2 changes: 2 additions & 0 deletions modules/shared.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -243,6 +243,8 @@ def fix_loader_name(name):
return 'AutoAWQ'
elif name in ['quip#', 'quip-sharp', 'quipsharp', 'quip_sharp']:
return 'QuIP#'
elif name in ['HQQ']:
return 'HQQ'


def add_extension(name, last=False):
Expand Down
1 change: 1 addition & 0 deletions requirements.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@ tqdm
wandb

git+https://github.com/oobabooga/torch-grammar.git
git+https://github.com/mobiusml/hqq.git

# bitsandbytes
bitsandbytes==0.41.1; platform_system != "Windows"
Expand Down