How to quantify the Qwen2.5-VL-3B model?

[Gusha-nye]

Hi，big guys!
Recently I wanted to quantize a Qwen2.5-VL-3B model and deploy it locally, I tried to use sglang (https://docs.sglang.ai/backend/quantization.html) to quantize the model but it failed with the following error:

sglang seems to be able to quantize chat models (Qwen2.5-3B) only? I was able to successfully quantize the Qwen2.5-3B model
I would like to ask you guys if there is any way to quantize the Qwen2.5-VL-3B model?

[Gusha-nye]

Now that I have changed the code, the model loads successfully but there are other errors as follows

(sglangEnv) (base) ubuntu@ubun:~/Song/sglang$ python Quantization.py
Loading checkpoint shards: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:01<00:00, 1.43it/s]
Model Load finished
Using a slow image processor as use_fast is unset and a slow processor was saved with this model. use_fast=True will be the default behavior in v4.48, even if the model was saved with a slow processor. This will result in minor differences in outputs. You'll still be able to use a slow processor with use_fast=False.
Tokenizer Load finished
2025-03-20T16:59:08.550688+0800 | main | WARNING - Process rank: 0, device: cuda:0, n_gpu: 1, distributed training: True, 16-bits training: False
2025-03-20T16:59:08.551472+0800 | main | INFO - Training/evaluation parameters TrainingArguments(
_n_gpu=1,
accelerator_config={'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None, 'use_configured_state': False},
adafactor=False,
adam_beta1=0.9,
adam_beta2=0.999,
adam_epsilon=1e-08,
auto_find_batch_size=False,
average_tokens_across_devices=False,
batch_eval_metrics=False,
bf16=False,
bf16_full_eval=False,
data_seed=None,
dataloader_drop_last=False,
dataloader_num_workers=0,
dataloader_persistent_workers=False,
dataloader_pin_memory=True,
dataloader_prefetch_factor=None,
ddp_backend=None,
ddp_broadcast_buffers=None,
ddp_bucket_cap_mb=None,
ddp_find_unused_parameters=None,
ddp_timeout=1800,
debug=[],
deepspeed=None,
disable_tqdm=False,
dispatch_batches=None,
do_eval=False,
do_oneshot=True,
do_predict=False,
do_train=False,
eval_accumulation_steps=None,
eval_delay=0,
eval_do_concat_batches=True,
eval_on_start=False,
eval_steps=None,
eval_strategy=IntervalStrategy.NO,
eval_use_gather_object=False,
evaluation_strategy=None,
fp16=False,
fp16_backend=auto,
fp16_full_eval=False,
fp16_opt_level=O1,
fsdp=[],
fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False},
fsdp_min_num_params=0,
fsdp_transformer_layer_cls_to_wrap=None,
full_determinism=False,
gradient_accumulation_steps=1,
gradient_checkpointing=False,
gradient_checkpointing_kwargs=None,
greater_is_better=None,
group_by_length=False,
half_precision_backend=auto,
hub_always_push=False,
hub_model_id=None,
hub_private_repo=None,
hub_strategy=HubStrategy.EVERY_SAVE,
hub_token=<HUB_TOKEN>,
ignore_data_skip=False,
include_for_metrics=[],
include_inputs_for_metrics=False,
include_num_input_tokens_seen=False,
include_tokens_per_second=False,
jit_mode_eval=False,
label_names=None,
label_smoothing_factor=0.0,
learning_rate=5e-05,
length_column_name=length,
load_best_model_at_end=False,
local_rank=0,
log_level=passive,
log_level_replica=warning,
log_on_each_node=True,
logging_dir=./output/runs/Mar20_16-59-08_ubun,
logging_first_step=False,
logging_nan_inf_filter=True,
logging_steps=500,
logging_strategy=IntervalStrategy.STEPS,
lr_scheduler_kwargs={},
lr_scheduler_type=SchedulerType.LINEAR,
max_grad_norm=1.0,
max_steps=-1,
metric_for_best_model=None,
mp_parameters=,
neftune_noise_alpha=None,
no_cuda=False,
num_train_epochs=3.0,
optim=OptimizerNames.ADAMW_TORCH,
optim_args=None,
optim_target_modules=None,
output_dir=./output,
overwrite_output_dir=False,
past_index=-1,
per_device_eval_batch_size=8,
per_device_train_batch_size=8,
prediction_loss_only=False,
push_to_hub=False,
push_to_hub_model_id=None,
push_to_hub_organization=None,
push_to_hub_token=<PUSH_TO_HUB_TOKEN>,
ray_scope=last,
remove_unused_columns=True,
report_to=[],
restore_callback_states_from_checkpoint=False,
resume_from_checkpoint=None,
run_name=./output,
run_stages=False,
save_on_each_node=False,
save_only_model=False,
save_safetensors=True,
save_steps=500,
save_strategy=SaveStrategy.STEPS,
save_total_limit=None,
seed=42,
skip_memory_metrics=True,
split_batches=None,
tf32=None,
torch_compile=False,
torch_compile_backend=None,
torch_compile_mode=None,
torch_empty_cache_steps=None,
torchdynamo=None,
tpu_metrics_debug=False,
tpu_num_cores=None,
use_cpu=False,
use_ipex=False,
use_legacy_prediction_loop=False,
use_liger_kernel=False,
use_mps_device=False,
warmup_ratio=0.0,
warmup_steps=0,
weight_decay=0.0,
)
2025-03-20T16:59:08.858207+0800 | _check_create_state | INFO - State created for compression lifecycle
2025-03-20T16:59:08.858472+0800 | pre_initialize_structure | INFO - Compression lifecycle structure pre-initialized for 0 modifiers
2025-03-20T16:59:08.858533+0800 | pre_initialize_structure | INFO - Compression lifecycle structure pre-initialized for 0 modifiers
2025-03-20T16:59:08.858563+0800 | populate_datasets | INFO - Running oneshot without calibration data. This is expected for weight-only and dynamic quantization
2025-03-20T16:59:08.861435+0800 | one_shot | INFO - *** One Shot ***
2025-03-20T16:59:08.861538+0800 | from_modifiers | INFO - Creating recipe from modifiers
2025-03-20T16:59:08.875834+0800 | _check_compile_recipe | INFO - Recipe compiled and 1 modifiers created
/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/compressed_tensors/utils/offload.py:205: UserWarning: Shape of parameter being updated torch.Size([1280, 26]) does not match shape of update data torch.Size([1280, 27])
warnings.warn(
Traceback (most recent call last):
File "/home/ubuntu/Song/sglang/Quantization.py", line 20, in
oneshot(model=model, recipe=recipe)
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/llmcompressor/transformers/finetune/text_generation.py", line 94, in oneshot
main(model_args, data_args, recipe_args, training_args)
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/llmcompressor/transformers/finetune/text_generation.py", line 447, in main
stage_runner.one_shot()
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/llmcompressor/transformers/finetune/runner.py", line 168, in one_shot
self.trainer.one_shot(calibration_data=calib_data, stage=stage)
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/llmcompressor/transformers/finetune/session_mixin.py", line 454, in one_shot
apply(
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/llmcompressor/core/session_functions.py", line 184, in apply
return active_session().apply(
^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/llmcompressor/core/session.py", line 212, in apply
self.initialize(**kwargs)
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/llmcompressor/core/session.py", line 158, in initialize
mod_data = self._lifecycle.initialize(
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/llmcompressor/core/lifecycle.py", line 126, in initialize
data = mod.initialize(state=self.state, **extras)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/llmcompressor/modifiers/stage.py", line 124, in initialize
modifier.initialize(state, **kwargs)
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/llmcompressor/modifiers/modifier.py", line 118, in initialize
initialized = self.on_initialize(state=state, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/llmcompressor/modifiers/quantization/quantization/base.py", line 110, in on_initialize
module.apply(update_weight_zp_scale)
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1029, in apply
module.apply(fn)
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1029, in apply
module.apply(fn)
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1029, in apply
module.apply(fn)
[Previous line repeated 2 more times]
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1030, in apply
fn(self)
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/llmcompressor/modifiers/quantization/calibration.py", line 123, in update_weight_zp_scale
call_observer(module=module, base_name="weight")
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/llmcompressor/modifiers/quantization/calibration.py", line 87, in call_observer
update_parameter_data(module, updated_scale, f"{base_name}scale")
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/compressed_tensors/utils/offload.py", line 155, in update_parameter_data
update_offload_parameter(module, param_name, new_param_data)
File "/home/ubuntu/Song/Env/sglangEnv/lib/python3.12/site-packages/compressed_tensors/utils/offload.py", line 212, in update_offload_parameter
param.data.copy(data)
RuntimeError: The size of tensor a (26) must match the size of tensor b (27) at non-singleton dimension 1