Initial Upload
Browse files- LICENSE.md +70 -0
- README.md +73 -0
- meta-llama_Llama-3.2-1B-Instruct-TEQ-int4-gs128-asym.py +350 -0
- qconfig.json +2262 -0
- quantized_weight.pt +3 -0
- requirements_conda_neural-compressor-3.3.1.txt +164 -0
LICENSE.md
ADDED
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
LLAMA 3.2 COMMUNITY LICENSE AGREEMENT
|
2 |
+
|
3 |
+
Llama 3.2 Version Release Date: September 25, 2024
|
4 |
+
|
5 |
+
“Agreement” means the terms and conditions for use, reproduction, distribution and modification of the Llama Materials set forth herein.
|
6 |
+
|
7 |
+
|
8 |
+
“Documentation” means the specifications, manuals and documentation accompanying Llama 3.2 distributed by Meta at https://www.llama.com/docs/overview.
|
9 |
+
|
10 |
+
|
11 |
+
“Licensee” or “you” means you, or your employer or any other person or entity (if you are entering into this Agreement on such person or entity’s behalf), of the age required under applicable laws, rules or regulations to provide legal consent and that has legal authority to bind your employer or such other person or entity if you are entering in this Agreement on their behalf.
|
12 |
+
|
13 |
+
|
14 |
+
“Llama 3.2” means the foundational large language models and software and algorithms, including machine-learning model code, trained model weights, inference-enabling code, training-enabling code, fine-tuning enabling code and other elements of the foregoing distributed by Meta at https://www.llama.com/llama-downloads.
|
15 |
+
|
16 |
+
|
17 |
+
“Llama Materials” means, collectively, Meta’s proprietary Llama 3.2 and Documentation (and any portion thereof) made available under this Agreement.
|
18 |
+
|
19 |
+
|
20 |
+
“Meta” or “we” means Meta Platforms Ireland Limited (if you are located in or, if you are an entity, your principal place of business is in the EEA or Switzerland) and Meta Platforms, Inc. (if you are located outside of the EEA or Switzerland).
|
21 |
+
|
22 |
+
|
23 |
+
By clicking “I Accept” below or by using or distributing any portion or element of the Llama Materials, you agree to be bound by this Agreement.
|
24 |
+
|
25 |
+
|
26 |
+
1. License Rights and Redistribution.
|
27 |
+
|
28 |
+
|
29 |
+
a. Grant of Rights. You are granted a non-exclusive, worldwide, non-transferable and royalty-free limited license under Meta’s intellectual property or other rights owned by Meta embodied in the Llama Materials to use, reproduce, distribute, copy, create derivative works of, and make modifications to the Llama Materials.
|
30 |
+
|
31 |
+
|
32 |
+
b. Redistribution and Use.
|
33 |
+
|
34 |
+
|
35 |
+
i. If you distribute or make available the Llama Materials (or any derivative works thereof), or a product or service (including another AI model) that contains any of them, you shall (A) provide a copy of this Agreement with any such Llama Materials; and (B) prominently display “Built with Llama” on a related website, user interface, blogpost, about page, or product documentation. If you use the Llama Materials or any outputs or results of the Llama Materials to create, train, fine tune, or otherwise improve an AI model, which is distributed or made available, you shall also include “Llama” at the beginning of any such AI model name.
|
36 |
+
|
37 |
+
|
38 |
+
ii. If you receive Llama Materials, or any derivative works thereof, from a Licensee as part of an integrated end user product, then Section 2 of this Agreement will not apply to you.
|
39 |
+
|
40 |
+
|
41 |
+
iii. You must retain in all copies of the Llama Materials that you distribute the following attribution notice within a “Notice” text file distributed as a part of such copies: “Llama 3.2 is licensed under the Llama 3.2 Community License, Copyright © Meta Platforms, Inc. All Rights Reserved.”
|
42 |
+
|
43 |
+
|
44 |
+
iv. Your use of the Llama Materials must comply with applicable laws and regulations (including trade compliance laws and regulations) and adhere to the Acceptable Use Policy for the Llama Materials (available at https://www.llama.com/llama3_2/use-policy), which is hereby incorporated by reference into this Agreement.
|
45 |
+
|
46 |
+
2. Additional Commercial Terms. If, on the Llama 3.2 version release date, the monthly active users of the products or services made available by or for Licensee, or Licensee’s affiliates, is greater than 700 million monthly active users in the preceding calendar month, you must request a license from Meta, which Meta may grant to you in its sole discretion, and you are not authorized to exercise any of the rights under this Agreement unless or until Meta otherwise expressly grants you such rights.
|
47 |
+
|
48 |
+
|
49 |
+
3. Disclaimer of Warranty. UNLESS REQUIRED BY APPLICABLE LAW, THE LLAMA MATERIALS AND ANY OUTPUT AND RESULTS THEREFROM ARE PROVIDED ON AN “AS IS” BASIS, WITHOUT WARRANTIES OF ANY KIND, AND META DISCLAIMS ALL WARRANTIES OF ANY KIND, BOTH EXPRESS AND IMPLIED, INCLUDING, WITHOUT LIMITATION, ANY WARRANTIES OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE. YOU ARE SOLELY RESPONSIBLE FOR DETERMINING THE APPROPRIATENESS OF USING OR REDISTRIBUTING THE LLAMA MATERIALS AND ASSUME ANY RISKS ASSOCIATED WITH YOUR USE OF THE LLAMA MATERIALS AND ANY OUTPUT AND RESULTS.
|
50 |
+
|
51 |
+
|
52 |
+
4. Limitation of Liability. IN NO EVENT WILL META OR ITS AFFILIATES BE LIABLE UNDER ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, TORT, NEGLIGENCE, PRODUCTS LIABILITY, OR OTHERWISE, ARISING OUT OF THIS AGREEMENT, FOR ANY LOST PROFITS OR ANY INDIRECT, SPECIAL, CONSEQUENTIAL, INCIDENTAL, EXEMPLARY OR PUNITIVE DAMAGES, EVEN IF META OR ITS AFFILIATES HAVE BEEN ADVISED OF THE POSSIBILITY OF ANY OF THE FOREGOING.
|
53 |
+
|
54 |
+
|
55 |
+
5. Intellectual Property.
|
56 |
+
|
57 |
+
|
58 |
+
a. No trademark licenses are granted under this Agreement, and in connection with the Llama Materials, neither Meta nor Licensee may use any name or mark owned by or associated with the other or any of its affiliates, except as required for reasonable and customary use in describing and redistributing the Llama Materials or as set forth in this Section 5(a). Meta hereby grants you a license to use “Llama” (the “Mark”) solely as required to comply with the last sentence of Section 1.b.i. You will comply with Meta’s brand guidelines (currently accessible at https://about.meta.com/brand/resources/meta/company-brand/). All goodwill arising out of your use of the Mark will inure to the benefit of Meta.
|
59 |
+
|
60 |
+
|
61 |
+
b. Subject to Meta’s ownership of Llama Materials and derivatives made by or for Meta, with respect to any derivative works and modifications of the Llama Materials that are made by you, as between you and Meta, you are and will be the owner of such derivative works and modifications.
|
62 |
+
|
63 |
+
|
64 |
+
c. If you institute litigation or other proceedings against Meta or any entity (including a cross-claim or counterclaim in a lawsuit) alleging that the Llama Materials or Llama 3.2 outputs or results, or any portion of any of the foregoing, constitutes infringement of intellectual property or other rights owned or licensable by you, then any licenses granted to you under this Agreement shall terminate as of the date such litigation or claim is filed or instituted. You will indemnify and hold harmless Meta from and against any claim by any third party arising out of or related to your use or distribution of the Llama Materials.
|
65 |
+
|
66 |
+
|
67 |
+
6. Term and Termination. The term of this Agreement will commence upon your acceptance of this Agreement or access to the Llama Materials and will continue in full force and effect until terminated in accordance with the terms and conditions herein. Meta may terminate this Agreement if you are in breach of any term or condition of this Agreement. Upon termination of this Agreement, you shall delete and cease use of the Llama Materials. Sections 3, 4 and 7 shall survive the termination of this Agreement.
|
68 |
+
|
69 |
+
|
70 |
+
7. Governing Law and Jurisdiction. This Agreement will be governed and construed under the laws of the State of California without regard to choice of law principles, and the UN Convention on Contracts for the International Sale of Goods does not apply to this Agreement. The courts of California shall have exclusive jurisdiction of any dispute arising out of this Agreement.
|
README.md
ADDED
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- en
|
4 |
+
- de
|
5 |
+
- fr
|
6 |
+
- it
|
7 |
+
- pt
|
8 |
+
- hi
|
9 |
+
- es
|
10 |
+
- th
|
11 |
+
license: llama3.2
|
12 |
+
library_name: transformers
|
13 |
+
tags:
|
14 |
+
- woq
|
15 |
+
- intel-neural-compressor
|
16 |
+
- inc
|
17 |
+
- neural-compressor
|
18 |
+
- intel
|
19 |
+
- teq
|
20 |
+
- meta
|
21 |
+
- pytorch
|
22 |
+
- llama
|
23 |
+
- llama-3
|
24 |
+
model_name: Llama 3.2 1B Instruct
|
25 |
+
base_model: meta-llama/Llama-3.2-1B-Instruct
|
26 |
+
inference: false
|
27 |
+
model_creator: meta-llama
|
28 |
+
pipeline_tag: text-generation
|
29 |
+
prompt_template: '{prompt}
|
30 |
+
'
|
31 |
+
quantized_by: fbaldassarri
|
32 |
+
---
|
33 |
+
|
34 |
+
## Model Information
|
35 |
+
|
36 |
+
Quantized version of [meta-llama/Llama-3.2-1B-Instruct](https://huggingface.co/meta-llama/Llama-3.2-1B-Instruct) using torch.float32 for quantization tuning.
|
37 |
+
- 4 bits (INT4)
|
38 |
+
- group size = 128
|
39 |
+
- Asymmetrical Quantization
|
40 |
+
- Algorith method: TEQ (Trainable Equivalent Transformation for Quantization of LLMs)
|
41 |
+
|
42 |
+
Quantization framework: [Intel Neural Compressor](https://github.com/intel/neural-compressor/) version 3.3.1
|
43 |
+
|
44 |
+
Note: this INT4 version of Llama-3.2-1B-Instruct has been quantized to run inference through CPU.
|
45 |
+
|
46 |
+
## Disclaimer
|
47 |
+
|
48 |
+
This quantized model comes with no warrenty. It has been developed experimetally only for research purposes.
|
49 |
+
|
50 |
+
This repository only contains contains two files: quantized_model.pt (weights structure) and qconfig.json, and the generated model is a quantized model.
|
51 |
+
It needs to be used in combination with the base model meta-llama/Llama-3.2-1B-Instruct.
|
52 |
+
|
53 |
+
## Replication Recipe
|
54 |
+
|
55 |
+
```
|
56 |
+
$ conda create --name neural-compressor-3.3.1 --file requirements_conda_neural-compressor-3.3.1
|
57 |
+
|
58 |
+
$ python meta-llama_Llama-3.2-1B-Instruct-TEQ-int4-gs128-asym.py
|
59 |
+
```
|
60 |
+
|
61 |
+
## Run Inference
|
62 |
+
|
63 |
+
To run inference you can use [fbaldassarri/woq-inference](https://github.com/fbaldassarri/woq-inference).
|
64 |
+
|
65 |
+
```
|
66 |
+
python teq_inference.py --base meta-llama/Llama-3.2-1B-Instruct --model_dir ./meta-llama_Llama-3.2-1B-Instruct-TEQ-int4-gs128-asym --weights_file quantized_weight.pt --config_file qconfig.json --prompt "What If you have got superpowers?" --device cpu
|
67 |
+
```
|
68 |
+
|
69 |
+
Note: You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
|
70 |
+
|
71 |
+
## License
|
72 |
+
|
73 |
+
[Llama 3.2 Community License](https://github.com/meta-llama/llama-models/blob/main/models/llama3_2/LICENSE)
|
meta-llama_Llama-3.2-1B-Instruct-TEQ-int4-gs128-asym.py
ADDED
@@ -0,0 +1,350 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import sys
|
3 |
+
import time
|
4 |
+
import random
|
5 |
+
import torch
|
6 |
+
|
7 |
+
from collections import UserDict
|
8 |
+
from packaging.version import Version
|
9 |
+
from datasets import load_dataset
|
10 |
+
from torch.utils.data import DataLoader
|
11 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
12 |
+
|
13 |
+
from neural_compressor.common import logger
|
14 |
+
from neural_compressor.torch.utils import is_hpex_available, get_torch_version
|
15 |
+
|
16 |
+
# ====== utils.py content inlined and fixed ======
|
17 |
+
|
18 |
+
class DataloaderPreprocessor:
|
19 |
+
def __init__(self, dataloader_original, use_max_length=False, max_seq_length=2048, nsamples=128) -> None:
|
20 |
+
self.dataloader_original = dataloader_original
|
21 |
+
self.use_max_length = use_max_length
|
22 |
+
self.max_seq_length = max_seq_length
|
23 |
+
self.nsamples = nsamples
|
24 |
+
self.dataloader = []
|
25 |
+
self.is_ready = False
|
26 |
+
|
27 |
+
def get_prepared_dataloader(self):
|
28 |
+
if not self.is_ready:
|
29 |
+
self.prepare_dataloader()
|
30 |
+
return self.dataloader
|
31 |
+
|
32 |
+
def prepare_dataloader(self):
|
33 |
+
if self.use_max_length:
|
34 |
+
self.obtain_first_n_samples_fulllength()
|
35 |
+
else:
|
36 |
+
self.obtain_first_n_samples()
|
37 |
+
self.is_ready = True
|
38 |
+
|
39 |
+
def obtain_first_n_samples(self, seed=0):
|
40 |
+
"""Get first nsample data as the real calibration dataset."""
|
41 |
+
self.dataloader.clear()
|
42 |
+
random.seed(seed)
|
43 |
+
for batch in self.dataloader_original:
|
44 |
+
if len(self.dataloader) == self.nsamples:
|
45 |
+
logger.info(f"Successfully collect {self.nsamples} calibration samples.")
|
46 |
+
break
|
47 |
+
# list, tuple
|
48 |
+
if isinstance(batch, list) or isinstance(batch, tuple):
|
49 |
+
if batch[0].shape[-1] > self.max_seq_length:
|
50 |
+
i = random.randint(0, batch[0].shape[-1] - self.max_seq_length - 1)
|
51 |
+
j = i + self.max_seq_length
|
52 |
+
batch_final = []
|
53 |
+
for item in batch:
|
54 |
+
if isinstance(item, torch.Tensor) and item.ndim == 2:
|
55 |
+
batch_final.append(item[:, i:j])
|
56 |
+
else:
|
57 |
+
batch_final.append(item)
|
58 |
+
else:
|
59 |
+
batch_final = batch[:]
|
60 |
+
# dict
|
61 |
+
elif isinstance(batch, dict):
|
62 |
+
try:
|
63 |
+
length = batch["input_ids"].shape[-1]
|
64 |
+
except Exception:
|
65 |
+
logger.warning("Please make sure your dict'like data contains key of 'input_ids'.")
|
66 |
+
continue
|
67 |
+
batch_final = {}
|
68 |
+
if length > self.max_seq_length:
|
69 |
+
i = random.randint(0, length - self.max_seq_length - 1)
|
70 |
+
j = i + self.max_seq_length
|
71 |
+
for key in batch.keys():
|
72 |
+
if isinstance(batch[key], torch.Tensor):
|
73 |
+
batch_final[key] = batch[key][:, i:j]
|
74 |
+
else:
|
75 |
+
batch_final[key] = batch[key]
|
76 |
+
else:
|
77 |
+
batch_final = batch
|
78 |
+
# tensor
|
79 |
+
else:
|
80 |
+
if batch.shape[-1] > self.max_seq_length:
|
81 |
+
i = random.randint(0, batch.shape[-1] - self.max_seq_length - 1)
|
82 |
+
j = i + self.max_seq_length
|
83 |
+
batch_final = batch[:, i:j]
|
84 |
+
else:
|
85 |
+
batch_final = batch
|
86 |
+
self.dataloader.append(batch_final)
|
87 |
+
if len(self.dataloader) < self.nsamples:
|
88 |
+
logger.warning(f"Try to use {self.nsamples} data, but entire dataset size is {len(self.dataloader)}.")
|
89 |
+
|
90 |
+
def obtain_first_n_samples_fulllength(self, seed=0):
|
91 |
+
self.dataloader.clear()
|
92 |
+
random.seed(seed)
|
93 |
+
unified_length = self.max_seq_length
|
94 |
+
for batch in self.dataloader_original:
|
95 |
+
if len(self.dataloader) == self.nsamples:
|
96 |
+
logger.info(f"Successfully collect {self.nsamples} calibration samples.")
|
97 |
+
break
|
98 |
+
# list & tuple
|
99 |
+
if isinstance(batch, list) or isinstance(batch, tuple):
|
100 |
+
if batch[0].shape[-1] == unified_length:
|
101 |
+
batch_final = batch[:]
|
102 |
+
elif batch[0].shape[-1] > unified_length:
|
103 |
+
i = random.randint(0, batch[0].shape[-1] - unified_length - 1)
|
104 |
+
j = i + unified_length
|
105 |
+
batch_final = []
|
106 |
+
for item in batch:
|
107 |
+
if isinstance(item, torch.Tensor) and item.ndim == 2:
|
108 |
+
batch_final.append(item[:, i:j])
|
109 |
+
else:
|
110 |
+
batch_final.append(item)
|
111 |
+
else:
|
112 |
+
continue
|
113 |
+
# dict
|
114 |
+
elif isinstance(batch, dict):
|
115 |
+
try:
|
116 |
+
length = batch["input_ids"].shape[-1]
|
117 |
+
except Exception:
|
118 |
+
logger.warning("Please make sure your dict'like data contains key of 'input_ids'.")
|
119 |
+
continue
|
120 |
+
batch_final = {}
|
121 |
+
if length == self.max_seq_length:
|
122 |
+
batch_final = batch
|
123 |
+
elif length > self.max_seq_length:
|
124 |
+
i = random.randint(0, length - self.max_seq_length - 1)
|
125 |
+
j = i + self.max_seq_length
|
126 |
+
for key in batch.keys():
|
127 |
+
if isinstance(batch[key], torch.Tensor):
|
128 |
+
batch_final[key] = batch[key][:, i:j]
|
129 |
+
else:
|
130 |
+
batch_final[key] = batch[key]
|
131 |
+
else:
|
132 |
+
continue
|
133 |
+
# tensor
|
134 |
+
else:
|
135 |
+
if batch.shape[-1] == unified_length:
|
136 |
+
batch_final = batch
|
137 |
+
elif batch.shape[-1] > unified_length:
|
138 |
+
i = random.randint(0, batch.shape[-1] - unified_length - 1)
|
139 |
+
j = i + unified_length
|
140 |
+
batch_final = batch[:, i:j]
|
141 |
+
else:
|
142 |
+
continue
|
143 |
+
self.dataloader.append(batch_final)
|
144 |
+
if len(self.dataloader) < self.nsamples:
|
145 |
+
logger.warning(
|
146 |
+
f"Trying to allocate {self.nsamples} data with fixed length {unified_length}, "
|
147 |
+
f"but only {len(self.dataloader)} samples are found. Please use smaller 'self.max_seq_length' value."
|
148 |
+
)
|
149 |
+
|
150 |
+
def get_example_inputs(model, dataloader):
|
151 |
+
version = get_torch_version()
|
152 |
+
from neural_compressor.torch.algorithms.smooth_quant import move_input_to_device
|
153 |
+
if dataloader is None:
|
154 |
+
return None
|
155 |
+
device = next(model.parameters()).device
|
156 |
+
try:
|
157 |
+
for idx, (input, label) in enumerate(dataloader):
|
158 |
+
input = move_input_to_device(input, device)
|
159 |
+
if isinstance(input, (dict, UserDict)):
|
160 |
+
assert version.release >= Version("1.12.0").release, "INC support IPEX version >= 1.12.0"
|
161 |
+
if "label" in input.keys():
|
162 |
+
input.pop("label")
|
163 |
+
if version.release <= Version("2.0.1").release:
|
164 |
+
return tuple(input.values())
|
165 |
+
else:
|
166 |
+
return dict(input)
|
167 |
+
if isinstance(input, (list, tuple)):
|
168 |
+
return tuple(input)
|
169 |
+
if isinstance(input, torch.Tensor):
|
170 |
+
return input
|
171 |
+
break
|
172 |
+
except Exception as e:
|
173 |
+
for idx, input in enumerate(dataloader):
|
174 |
+
input = move_input_to_device(input, device)
|
175 |
+
if isinstance(input, (dict, UserDict)):
|
176 |
+
assert version.release >= Version("1.12.0").release, "INC support IPEX version >= 1.12.0"
|
177 |
+
if "label" in input.keys():
|
178 |
+
input.pop("label")
|
179 |
+
if version.release <= Version("2.0.1").release:
|
180 |
+
return tuple(input.values())
|
181 |
+
else:
|
182 |
+
return dict(input)
|
183 |
+
if isinstance(input, list) or isinstance(input, tuple):
|
184 |
+
return tuple(input)
|
185 |
+
if isinstance(input, torch.Tensor):
|
186 |
+
return input
|
187 |
+
break
|
188 |
+
if idx == 0:
|
189 |
+
assert False, "Please checkout the example_inputs format."
|
190 |
+
|
191 |
+
# ====== End of utils.py content ======
|
192 |
+
|
193 |
+
# ====== Hardcoded arguments ======
|
194 |
+
class Args:
|
195 |
+
model = "meta-llama/Llama-3.2-1B-Instruct"
|
196 |
+
trust_remote_code = True
|
197 |
+
revision = None
|
198 |
+
dataset = "neuralmagic/LLM_compression_calibration"
|
199 |
+
output_dir = "meta-llama_Llama-3.2-1B-Instruct-TEQ-int4-gs128-asym"
|
200 |
+
quantize = True
|
201 |
+
seed = 42
|
202 |
+
load = False
|
203 |
+
accuracy = False
|
204 |
+
performance = False
|
205 |
+
iters = 100
|
206 |
+
batch_size = 1
|
207 |
+
pad_max_length = 512
|
208 |
+
calib_iters = 512
|
209 |
+
tasks = "lambada_openai,hellaswag,winogrande,piqa"
|
210 |
+
peft_model_id = None
|
211 |
+
|
212 |
+
# Weight-only quantization configs
|
213 |
+
woq_algo = "TEQ"
|
214 |
+
woq_bits = 4
|
215 |
+
woq_dtype = "int"
|
216 |
+
woq_group_size = 128
|
217 |
+
woq_group_dim = 1
|
218 |
+
woq_scheme = "asym"
|
219 |
+
woq_use_mse_search = False
|
220 |
+
woq_use_full_range = False
|
221 |
+
quant_lm_head = True
|
222 |
+
use_hf_format = False
|
223 |
+
|
224 |
+
# TEQ/AWQ configs
|
225 |
+
use_auto_scale = False
|
226 |
+
use_auto_clip = False
|
227 |
+
folding = False
|
228 |
+
absorb_layer_dict = {}
|
229 |
+
|
230 |
+
# DoubleQuant configs
|
231 |
+
double_quant_type = None
|
232 |
+
double_quant_dtype = "fp32"
|
233 |
+
double_quant_bits = 8
|
234 |
+
double_quant_use_sym = True
|
235 |
+
double_quant_group_size = 256
|
236 |
+
|
237 |
+
args = Args()
|
238 |
+
calib_size = 1
|
239 |
+
|
240 |
+
if is_hpex_available():
|
241 |
+
import habana_frameworks.torch.core as htcore
|
242 |
+
htcore.hpu_set_inference_env()
|
243 |
+
device = "hpu"
|
244 |
+
else:
|
245 |
+
device = "cpu"
|
246 |
+
|
247 |
+
# ====== Helper functions ======
|
248 |
+
def get_user_model():
|
249 |
+
torchscript = False
|
250 |
+
if args.woq_algo in ["AWQ", "TEQ"]:
|
251 |
+
torchscript = True
|
252 |
+
user_model = AutoModelForCausalLM.from_pretrained(
|
253 |
+
args.model,
|
254 |
+
torchscript=torchscript,
|
255 |
+
trust_remote_code=args.trust_remote_code,
|
256 |
+
revision=args.revision,
|
257 |
+
)
|
258 |
+
tokenizer = AutoTokenizer.from_pretrained(args.model)
|
259 |
+
user_model = user_model.float()
|
260 |
+
user_model = user_model.to(memory_format=torch.channels_last)
|
261 |
+
user_model.eval()
|
262 |
+
return user_model, tokenizer
|
263 |
+
|
264 |
+
def calib_func(prepared_model):
|
265 |
+
for i, calib_input in enumerate(calib_dataloader):
|
266 |
+
if i > args.calib_iters:
|
267 |
+
break
|
268 |
+
prepared_model(calib_input[0])
|
269 |
+
|
270 |
+
# ====== Main quantization logic ======
|
271 |
+
if args.quantize:
|
272 |
+
user_model, tokenizer = get_user_model()
|
273 |
+
calib_dataset = load_dataset(args.dataset, split="train")
|
274 |
+
calib_dataset = calib_dataset.shuffle(seed=args.seed)
|
275 |
+
|
276 |
+
class Evaluator:
|
277 |
+
def __init__(self, dataset, tokenizer, batch_size=8, pad_val=1, pad_max=196, is_calib=False):
|
278 |
+
self.dataset = dataset
|
279 |
+
self.tokenizer = tokenizer
|
280 |
+
self.batch_size = batch_size
|
281 |
+
self.pad_val = pad_val
|
282 |
+
self.pad_max = pad_max
|
283 |
+
self.is_calib = is_calib
|
284 |
+
self.dataset = self.dataset.map(self.tokenize_function, batched=True)
|
285 |
+
self.dataset.set_format(type="torch", columns=["input_ids"])
|
286 |
+
|
287 |
+
@torch.no_grad()
|
288 |
+
def tokenize_function(self, examples):
|
289 |
+
if args.woq_algo in ['TEQ']:
|
290 |
+
if self.tokenizer.pad_token is None:
|
291 |
+
self.tokenizer.pad_token = self.tokenizer.eos_token
|
292 |
+
example = self.tokenizer(examples["text"], padding="max_length", max_length=self.pad_max)
|
293 |
+
else:
|
294 |
+
example = self.tokenizer(examples["text"])
|
295 |
+
return example
|
296 |
+
|
297 |
+
@torch.no_grad()
|
298 |
+
def collate_batch(self, batch):
|
299 |
+
input_ids_padded = []
|
300 |
+
last_ind = []
|
301 |
+
for text in batch:
|
302 |
+
input_ids = text["input_ids"]
|
303 |
+
pad_len = self.pad_max - input_ids.shape[0]
|
304 |
+
last_ind.append(input_ids.shape[0] - 1)
|
305 |
+
input_ids = input_ids[:self.pad_max] if len(input_ids) > self.pad_max else input_ids
|
306 |
+
input_ids = torch.nn.functional.pad(input_ids, (0, pad_len), value=self.pad_val)
|
307 |
+
input_ids_padded.append(input_ids)
|
308 |
+
return (torch.vstack(input_ids_padded), torch.tensor(last_ind))
|
309 |
+
|
310 |
+
calib_evaluator = Evaluator(calib_dataset, tokenizer, args.batch_size, pad_max=args.pad_max_length, is_calib=True)
|
311 |
+
calib_dataloader = DataLoader(
|
312 |
+
calib_evaluator.dataset,
|
313 |
+
batch_size=calib_size,
|
314 |
+
shuffle=False,
|
315 |
+
collate_fn=calib_evaluator.collate_batch,
|
316 |
+
)
|
317 |
+
|
318 |
+
# === TEQ quantization ===
|
319 |
+
from neural_compressor.torch.quantization import TEQConfig, prepare, convert
|
320 |
+
|
321 |
+
weight_sym = True if args.woq_scheme == "sym" else False
|
322 |
+
quant_config = TEQConfig(
|
323 |
+
dtype=args.woq_dtype,
|
324 |
+
bits=args.woq_bits,
|
325 |
+
use_sym=weight_sym,
|
326 |
+
group_size=args.woq_group_size,
|
327 |
+
group_dim=args.woq_group_dim,
|
328 |
+
folding=args.folding,
|
329 |
+
quant_lm_head=args.quant_lm_head,
|
330 |
+
)
|
331 |
+
|
332 |
+
example_inputs = torch.ones([1, args.pad_max_length], dtype=torch.long)
|
333 |
+
run_fn = calib_func
|
334 |
+
|
335 |
+
user_model = prepare(model=user_model, quant_config=quant_config, example_inputs=example_inputs)
|
336 |
+
run_fn(user_model)
|
337 |
+
user_model = convert(user_model)
|
338 |
+
|
339 |
+
# === Save quantized model ===
|
340 |
+
os.makedirs(args.output_dir, exist_ok=True)
|
341 |
+
print("Saving weight-only quantized model to", args.output_dir)
|
342 |
+
if args.use_hf_format:
|
343 |
+
user_model.save(args.output_dir, format="huggingface")
|
344 |
+
tokenizer.save_pretrained(args.output_dir)
|
345 |
+
else:
|
346 |
+
user_model.save(args.output_dir)
|
347 |
+
print("Saved weight-only quantized model.")
|
348 |
+
|
349 |
+
else:
|
350 |
+
print("Quantization not enabled. Exiting.")
|
qconfig.json
ADDED
@@ -0,0 +1,2262 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"('model.layers.0.self_attn.q_proj.orig_layer', 'Linear')": {
|
3 |
+
"teq": {
|
4 |
+
"dtype": "int",
|
5 |
+
"bits": 4,
|
6 |
+
"use_sym": false,
|
7 |
+
"group_size": 128,
|
8 |
+
"group_dim": 1,
|
9 |
+
"use_full_range": false,
|
10 |
+
"use_mse_search": false,
|
11 |
+
"use_layer_wise": false,
|
12 |
+
"use_double_quant": false,
|
13 |
+
"double_quant_bits": 8,
|
14 |
+
"double_quant_dtype": "int",
|
15 |
+
"double_quant_use_sym": true,
|
16 |
+
"double_quant_group_size": 256,
|
17 |
+
"quant_lm_head": true,
|
18 |
+
"absorb_to_layer": {},
|
19 |
+
"folding": false
|
20 |
+
}
|
21 |
+
},
|
22 |
+
"('model.layers.0.self_attn.k_proj.orig_layer', 'Linear')": {
|
23 |
+
"teq": {
|
24 |
+
"dtype": "int",
|
25 |
+
"bits": 4,
|
26 |
+
"use_sym": false,
|
27 |
+
"group_size": 128,
|
28 |
+
"group_dim": 1,
|
29 |
+
"use_full_range": false,
|
30 |
+
"use_mse_search": false,
|
31 |
+
"use_layer_wise": false,
|
32 |
+
"use_double_quant": false,
|
33 |
+
"double_quant_bits": 8,
|
34 |
+
"double_quant_dtype": "int",
|
35 |
+
"double_quant_use_sym": true,
|
36 |
+
"double_quant_group_size": 256,
|
37 |
+
"quant_lm_head": true,
|
38 |
+
"absorb_to_layer": {},
|
39 |
+
"folding": false
|
40 |
+
}
|
41 |
+
},
|
42 |
+
"('model.layers.0.self_attn.v_proj.orig_layer', 'Linear')": {
|
43 |
+
"teq": {
|
44 |
+
"dtype": "int",
|
45 |
+
"bits": 4,
|
46 |
+
"use_sym": false,
|
47 |
+
"group_size": 128,
|
48 |
+
"group_dim": 1,
|
49 |
+
"use_full_range": false,
|
50 |
+
"use_mse_search": false,
|
51 |
+
"use_layer_wise": false,
|
52 |
+
"use_double_quant": false,
|
53 |
+
"double_quant_bits": 8,
|
54 |
+
"double_quant_dtype": "int",
|
55 |
+
"double_quant_use_sym": true,
|
56 |
+
"double_quant_group_size": 256,
|
57 |
+
"quant_lm_head": true,
|
58 |
+
"absorb_to_layer": {},
|
59 |
+
"folding": false
|
60 |
+
}
|
61 |
+
},
|
62 |
+
"('model.layers.0.self_attn.o_proj.orig_layer', 'Linear')": {
|
63 |
+
"teq": {
|
64 |
+
"dtype": "int",
|
65 |
+
"bits": 4,
|
66 |
+
"use_sym": false,
|
67 |
+
"group_size": 128,
|
68 |
+
"group_dim": 1,
|
69 |
+
"use_full_range": false,
|
70 |
+
"use_mse_search": false,
|
71 |
+
"use_layer_wise": false,
|
72 |
+
"use_double_quant": false,
|
73 |
+
"double_quant_bits": 8,
|
74 |
+
"double_quant_dtype": "int",
|
75 |
+
"double_quant_use_sym": true,
|
76 |
+
"double_quant_group_size": 256,
|
77 |
+
"quant_lm_head": true,
|
78 |
+
"absorb_to_layer": {},
|
79 |
+
"folding": false
|
80 |
+
}
|
81 |
+
},
|
82 |
+
"('model.layers.0.mlp.gate_proj.orig_layer', 'Linear')": {
|
83 |
+
"teq": {
|
84 |
+
"dtype": "int",
|
85 |
+
"bits": 4,
|
86 |
+
"use_sym": false,
|
87 |
+
"group_size": 128,
|
88 |
+
"group_dim": 1,
|
89 |
+
"use_full_range": false,
|
90 |
+
"use_mse_search": false,
|
91 |
+
"use_layer_wise": false,
|
92 |
+
"use_double_quant": false,
|
93 |
+
"double_quant_bits": 8,
|
94 |
+
"double_quant_dtype": "int",
|
95 |
+
"double_quant_use_sym": true,
|
96 |
+
"double_quant_group_size": 256,
|
97 |
+
"quant_lm_head": true,
|
98 |
+
"absorb_to_layer": {},
|
99 |
+
"folding": false
|
100 |
+
}
|
101 |
+
},
|
102 |
+
"('model.layers.0.mlp.up_proj.orig_layer', 'Linear')": {
|
103 |
+
"teq": {
|
104 |
+
"dtype": "int",
|
105 |
+
"bits": 4,
|
106 |
+
"use_sym": false,
|
107 |
+
"group_size": 128,
|
108 |
+
"group_dim": 1,
|
109 |
+
"use_full_range": false,
|
110 |
+
"use_mse_search": false,
|
111 |
+
"use_layer_wise": false,
|
112 |
+
"use_double_quant": false,
|
113 |
+
"double_quant_bits": 8,
|
114 |
+
"double_quant_dtype": "int",
|
115 |
+
"double_quant_use_sym": true,
|
116 |
+
"double_quant_group_size": 256,
|
117 |
+
"quant_lm_head": true,
|
118 |
+
"absorb_to_layer": {},
|
119 |
+
"folding": false
|
120 |
+
}
|
121 |
+
},
|
122 |
+
"('model.layers.0.mlp.down_proj.orig_layer', 'Linear')": {
|
123 |
+
"teq": {
|
124 |
+
"dtype": "int",
|
125 |
+
"bits": 4,
|
126 |
+
"use_sym": false,
|
127 |
+
"group_size": 128,
|
128 |
+
"group_dim": 1,
|
129 |
+
"use_full_range": false,
|
130 |
+
"use_mse_search": false,
|
131 |
+
"use_layer_wise": false,
|
132 |
+
"use_double_quant": false,
|
133 |
+
"double_quant_bits": 8,
|
134 |
+
"double_quant_dtype": "int",
|
135 |
+
"double_quant_use_sym": true,
|
136 |
+
"double_quant_group_size": 256,
|
137 |
+
"quant_lm_head": true,
|
138 |
+
"absorb_to_layer": {},
|
139 |
+
"folding": false
|
140 |
+
}
|
141 |
+
},
|
142 |
+
"('model.layers.1.self_attn.q_proj.orig_layer', 'Linear')": {
|
143 |
+
"teq": {
|
144 |
+
"dtype": "int",
|
145 |
+
"bits": 4,
|
146 |
+
"use_sym": false,
|
147 |
+
"group_size": 128,
|
148 |
+
"group_dim": 1,
|
149 |
+
"use_full_range": false,
|
150 |
+
"use_mse_search": false,
|
151 |
+
"use_layer_wise": false,
|
152 |
+
"use_double_quant": false,
|
153 |
+
"double_quant_bits": 8,
|
154 |
+
"double_quant_dtype": "int",
|
155 |
+
"double_quant_use_sym": true,
|
156 |
+
"double_quant_group_size": 256,
|
157 |
+
"quant_lm_head": true,
|
158 |
+
"absorb_to_layer": {},
|
159 |
+
"folding": false
|
160 |
+
}
|
161 |
+
},
|
162 |
+
"('model.layers.1.self_attn.k_proj.orig_layer', 'Linear')": {
|
163 |
+
"teq": {
|
164 |
+
"dtype": "int",
|
165 |
+
"bits": 4,
|
166 |
+
"use_sym": false,
|
167 |
+
"group_size": 128,
|
168 |
+
"group_dim": 1,
|
169 |
+
"use_full_range": false,
|
170 |
+
"use_mse_search": false,
|
171 |
+
"use_layer_wise": false,
|
172 |
+
"use_double_quant": false,
|
173 |
+
"double_quant_bits": 8,
|
174 |
+
"double_quant_dtype": "int",
|
175 |
+
"double_quant_use_sym": true,
|
176 |
+
"double_quant_group_size": 256,
|
177 |
+
"quant_lm_head": true,
|
178 |
+
"absorb_to_layer": {},
|
179 |
+
"folding": false
|
180 |
+
}
|
181 |
+
},
|
182 |
+
"('model.layers.1.self_attn.v_proj.orig_layer', 'Linear')": {
|
183 |
+
"teq": {
|
184 |
+
"dtype": "int",
|
185 |
+
"bits": 4,
|
186 |
+
"use_sym": false,
|
187 |
+
"group_size": 128,
|
188 |
+
"group_dim": 1,
|
189 |
+
"use_full_range": false,
|
190 |
+
"use_mse_search": false,
|
191 |
+
"use_layer_wise": false,
|
192 |
+
"use_double_quant": false,
|
193 |
+
"double_quant_bits": 8,
|
194 |
+
"double_quant_dtype": "int",
|
195 |
+
"double_quant_use_sym": true,
|
196 |
+
"double_quant_group_size": 256,
|
197 |
+
"quant_lm_head": true,
|
198 |
+
"absorb_to_layer": {},
|
199 |
+
"folding": false
|
200 |
+
}
|
201 |
+
},
|
202 |
+
"('model.layers.1.self_attn.o_proj.orig_layer', 'Linear')": {
|
203 |
+
"teq": {
|
204 |
+
"dtype": "int",
|
205 |
+
"bits": 4,
|
206 |
+
"use_sym": false,
|
207 |
+
"group_size": 128,
|
208 |
+
"group_dim": 1,
|
209 |
+
"use_full_range": false,
|
210 |
+
"use_mse_search": false,
|
211 |
+
"use_layer_wise": false,
|
212 |
+
"use_double_quant": false,
|
213 |
+
"double_quant_bits": 8,
|
214 |
+
"double_quant_dtype": "int",
|
215 |
+
"double_quant_use_sym": true,
|
216 |
+
"double_quant_group_size": 256,
|
217 |
+
"quant_lm_head": true,
|
218 |
+
"absorb_to_layer": {},
|
219 |
+
"folding": false
|
220 |
+
}
|
221 |
+
},
|
222 |
+
"('model.layers.1.mlp.gate_proj.orig_layer', 'Linear')": {
|
223 |
+
"teq": {
|
224 |
+
"dtype": "int",
|
225 |
+
"bits": 4,
|
226 |
+
"use_sym": false,
|
227 |
+
"group_size": 128,
|
228 |
+
"group_dim": 1,
|
229 |
+
"use_full_range": false,
|
230 |
+
"use_mse_search": false,
|
231 |
+
"use_layer_wise": false,
|
232 |
+
"use_double_quant": false,
|
233 |
+
"double_quant_bits": 8,
|
234 |
+
"double_quant_dtype": "int",
|
235 |
+
"double_quant_use_sym": true,
|
236 |
+
"double_quant_group_size": 256,
|
237 |
+
"quant_lm_head": true,
|
238 |
+
"absorb_to_layer": {},
|
239 |
+
"folding": false
|
240 |
+
}
|
241 |
+
},
|
242 |
+
"('model.layers.1.mlp.up_proj.orig_layer', 'Linear')": {
|
243 |
+
"teq": {
|
244 |
+
"dtype": "int",
|
245 |
+
"bits": 4,
|
246 |
+
"use_sym": false,
|
247 |
+
"group_size": 128,
|
248 |
+
"group_dim": 1,
|
249 |
+
"use_full_range": false,
|
250 |
+
"use_mse_search": false,
|
251 |
+
"use_layer_wise": false,
|
252 |
+
"use_double_quant": false,
|
253 |
+
"double_quant_bits": 8,
|
254 |
+
"double_quant_dtype": "int",
|
255 |
+
"double_quant_use_sym": true,
|
256 |
+
"double_quant_group_size": 256,
|
257 |
+
"quant_lm_head": true,
|
258 |
+
"absorb_to_layer": {},
|
259 |
+
"folding": false
|
260 |
+
}
|
261 |
+
},
|
262 |
+
"('model.layers.1.mlp.down_proj.orig_layer', 'Linear')": {
|
263 |
+
"teq": {
|
264 |
+
"dtype": "int",
|
265 |
+
"bits": 4,
|
266 |
+
"use_sym": false,
|
267 |
+
"group_size": 128,
|
268 |
+
"group_dim": 1,
|
269 |
+
"use_full_range": false,
|
270 |
+
"use_mse_search": false,
|
271 |
+
"use_layer_wise": false,
|
272 |
+
"use_double_quant": false,
|
273 |
+
"double_quant_bits": 8,
|
274 |
+
"double_quant_dtype": "int",
|
275 |
+
"double_quant_use_sym": true,
|
276 |
+
"double_quant_group_size": 256,
|
277 |
+
"quant_lm_head": true,
|
278 |
+
"absorb_to_layer": {},
|
279 |
+
"folding": false
|
280 |
+
}
|
281 |
+
},
|
282 |
+
"('model.layers.2.self_attn.q_proj.orig_layer', 'Linear')": {
|
283 |
+
"teq": {
|
284 |
+
"dtype": "int",
|
285 |
+
"bits": 4,
|
286 |
+
"use_sym": false,
|
287 |
+
"group_size": 128,
|
288 |
+
"group_dim": 1,
|
289 |
+
"use_full_range": false,
|
290 |
+
"use_mse_search": false,
|
291 |
+
"use_layer_wise": false,
|
292 |
+
"use_double_quant": false,
|
293 |
+
"double_quant_bits": 8,
|
294 |
+
"double_quant_dtype": "int",
|
295 |
+
"double_quant_use_sym": true,
|
296 |
+
"double_quant_group_size": 256,
|
297 |
+
"quant_lm_head": true,
|
298 |
+
"absorb_to_layer": {},
|
299 |
+
"folding": false
|
300 |
+
}
|
301 |
+
},
|
302 |
+
"('model.layers.2.self_attn.k_proj.orig_layer', 'Linear')": {
|
303 |
+
"teq": {
|
304 |
+
"dtype": "int",
|
305 |
+
"bits": 4,
|
306 |
+
"use_sym": false,
|
307 |
+
"group_size": 128,
|
308 |
+
"group_dim": 1,
|
309 |
+
"use_full_range": false,
|
310 |
+
"use_mse_search": false,
|
311 |
+
"use_layer_wise": false,
|
312 |
+
"use_double_quant": false,
|
313 |
+
"double_quant_bits": 8,
|
314 |
+
"double_quant_dtype": "int",
|
315 |
+
"double_quant_use_sym": true,
|
316 |
+
"double_quant_group_size": 256,
|
317 |
+
"quant_lm_head": true,
|
318 |
+
"absorb_to_layer": {},
|
319 |
+
"folding": false
|
320 |
+
}
|
321 |
+
},
|
322 |
+
"('model.layers.2.self_attn.v_proj.orig_layer', 'Linear')": {
|
323 |
+
"teq": {
|
324 |
+
"dtype": "int",
|
325 |
+
"bits": 4,
|
326 |
+
"use_sym": false,
|
327 |
+
"group_size": 128,
|
328 |
+
"group_dim": 1,
|
329 |
+
"use_full_range": false,
|
330 |
+
"use_mse_search": false,
|
331 |
+
"use_layer_wise": false,
|
332 |
+
"use_double_quant": false,
|
333 |
+
"double_quant_bits": 8,
|
334 |
+
"double_quant_dtype": "int",
|
335 |
+
"double_quant_use_sym": true,
|
336 |
+
"double_quant_group_size": 256,
|
337 |
+
"quant_lm_head": true,
|
338 |
+
"absorb_to_layer": {},
|
339 |
+
"folding": false
|
340 |
+
}
|
341 |
+
},
|
342 |
+
"('model.layers.2.self_attn.o_proj.orig_layer', 'Linear')": {
|
343 |
+
"teq": {
|
344 |
+
"dtype": "int",
|
345 |
+
"bits": 4,
|
346 |
+
"use_sym": false,
|
347 |
+
"group_size": 128,
|
348 |
+
"group_dim": 1,
|
349 |
+
"use_full_range": false,
|
350 |
+
"use_mse_search": false,
|
351 |
+
"use_layer_wise": false,
|
352 |
+
"use_double_quant": false,
|
353 |
+
"double_quant_bits": 8,
|
354 |
+
"double_quant_dtype": "int",
|
355 |
+
"double_quant_use_sym": true,
|
356 |
+
"double_quant_group_size": 256,
|
357 |
+
"quant_lm_head": true,
|
358 |
+
"absorb_to_layer": {},
|
359 |
+
"folding": false
|
360 |
+
}
|
361 |
+
},
|
362 |
+
"('model.layers.2.mlp.gate_proj.orig_layer', 'Linear')": {
|
363 |
+
"teq": {
|
364 |
+
"dtype": "int",
|
365 |
+
"bits": 4,
|
366 |
+
"use_sym": false,
|
367 |
+
"group_size": 128,
|
368 |
+
"group_dim": 1,
|
369 |
+
"use_full_range": false,
|
370 |
+
"use_mse_search": false,
|
371 |
+
"use_layer_wise": false,
|
372 |
+
"use_double_quant": false,
|
373 |
+
"double_quant_bits": 8,
|
374 |
+
"double_quant_dtype": "int",
|
375 |
+
"double_quant_use_sym": true,
|
376 |
+
"double_quant_group_size": 256,
|
377 |
+
"quant_lm_head": true,
|
378 |
+
"absorb_to_layer": {},
|
379 |
+
"folding": false
|
380 |
+
}
|
381 |
+
},
|
382 |
+
"('model.layers.2.mlp.up_proj.orig_layer', 'Linear')": {
|
383 |
+
"teq": {
|
384 |
+
"dtype": "int",
|
385 |
+
"bits": 4,
|
386 |
+
"use_sym": false,
|
387 |
+
"group_size": 128,
|
388 |
+
"group_dim": 1,
|
389 |
+
"use_full_range": false,
|
390 |
+
"use_mse_search": false,
|
391 |
+
"use_layer_wise": false,
|
392 |
+
"use_double_quant": false,
|
393 |
+
"double_quant_bits": 8,
|
394 |
+
"double_quant_dtype": "int",
|
395 |
+
"double_quant_use_sym": true,
|
396 |
+
"double_quant_group_size": 256,
|
397 |
+
"quant_lm_head": true,
|
398 |
+
"absorb_to_layer": {},
|
399 |
+
"folding": false
|
400 |
+
}
|
401 |
+
},
|
402 |
+
"('model.layers.2.mlp.down_proj.orig_layer', 'Linear')": {
|
403 |
+
"teq": {
|
404 |
+
"dtype": "int",
|
405 |
+
"bits": 4,
|
406 |
+
"use_sym": false,
|
407 |
+
"group_size": 128,
|
408 |
+
"group_dim": 1,
|
409 |
+
"use_full_range": false,
|
410 |
+
"use_mse_search": false,
|
411 |
+
"use_layer_wise": false,
|
412 |
+
"use_double_quant": false,
|
413 |
+
"double_quant_bits": 8,
|
414 |
+
"double_quant_dtype": "int",
|
415 |
+
"double_quant_use_sym": true,
|
416 |
+
"double_quant_group_size": 256,
|
417 |
+
"quant_lm_head": true,
|
418 |
+
"absorb_to_layer": {},
|
419 |
+
"folding": false
|
420 |
+
}
|
421 |
+
},
|
422 |
+
"('model.layers.3.self_attn.q_proj.orig_layer', 'Linear')": {
|
423 |
+
"teq": {
|
424 |
+
"dtype": "int",
|
425 |
+
"bits": 4,
|
426 |
+
"use_sym": false,
|
427 |
+
"group_size": 128,
|
428 |
+
"group_dim": 1,
|
429 |
+
"use_full_range": false,
|
430 |
+
"use_mse_search": false,
|
431 |
+
"use_layer_wise": false,
|
432 |
+
"use_double_quant": false,
|
433 |
+
"double_quant_bits": 8,
|
434 |
+
"double_quant_dtype": "int",
|
435 |
+
"double_quant_use_sym": true,
|
436 |
+
"double_quant_group_size": 256,
|
437 |
+
"quant_lm_head": true,
|
438 |
+
"absorb_to_layer": {},
|
439 |
+
"folding": false
|
440 |
+
}
|
441 |
+
},
|
442 |
+
"('model.layers.3.self_attn.k_proj.orig_layer', 'Linear')": {
|
443 |
+
"teq": {
|
444 |
+
"dtype": "int",
|
445 |
+
"bits": 4,
|
446 |
+
"use_sym": false,
|
447 |
+
"group_size": 128,
|
448 |
+
"group_dim": 1,
|
449 |
+
"use_full_range": false,
|
450 |
+
"use_mse_search": false,
|
451 |
+
"use_layer_wise": false,
|
452 |
+
"use_double_quant": false,
|
453 |
+
"double_quant_bits": 8,
|
454 |
+
"double_quant_dtype": "int",
|
455 |
+
"double_quant_use_sym": true,
|
456 |
+
"double_quant_group_size": 256,
|
457 |
+
"quant_lm_head": true,
|
458 |
+
"absorb_to_layer": {},
|
459 |
+
"folding": false
|
460 |
+
}
|
461 |
+
},
|
462 |
+
"('model.layers.3.self_attn.v_proj.orig_layer', 'Linear')": {
|
463 |
+
"teq": {
|
464 |
+
"dtype": "int",
|
465 |
+
"bits": 4,
|
466 |
+
"use_sym": false,
|
467 |
+
"group_size": 128,
|
468 |
+
"group_dim": 1,
|
469 |
+
"use_full_range": false,
|
470 |
+
"use_mse_search": false,
|
471 |
+
"use_layer_wise": false,
|
472 |
+
"use_double_quant": false,
|
473 |
+
"double_quant_bits": 8,
|
474 |
+
"double_quant_dtype": "int",
|
475 |
+
"double_quant_use_sym": true,
|
476 |
+
"double_quant_group_size": 256,
|
477 |
+
"quant_lm_head": true,
|
478 |
+
"absorb_to_layer": {},
|
479 |
+
"folding": false
|
480 |
+
}
|
481 |
+
},
|
482 |
+
"('model.layers.3.self_attn.o_proj.orig_layer', 'Linear')": {
|
483 |
+
"teq": {
|
484 |
+
"dtype": "int",
|
485 |
+
"bits": 4,
|
486 |
+
"use_sym": false,
|
487 |
+
"group_size": 128,
|
488 |
+
"group_dim": 1,
|
489 |
+
"use_full_range": false,
|
490 |
+
"use_mse_search": false,
|
491 |
+
"use_layer_wise": false,
|
492 |
+
"use_double_quant": false,
|
493 |
+
"double_quant_bits": 8,
|
494 |
+
"double_quant_dtype": "int",
|
495 |
+
"double_quant_use_sym": true,
|
496 |
+
"double_quant_group_size": 256,
|
497 |
+
"quant_lm_head": true,
|
498 |
+
"absorb_to_layer": {},
|
499 |
+
"folding": false
|
500 |
+
}
|
501 |
+
},
|
502 |
+
"('model.layers.3.mlp.gate_proj.orig_layer', 'Linear')": {
|
503 |
+
"teq": {
|
504 |
+
"dtype": "int",
|
505 |
+
"bits": 4,
|
506 |
+
"use_sym": false,
|
507 |
+
"group_size": 128,
|
508 |
+
"group_dim": 1,
|
509 |
+
"use_full_range": false,
|
510 |
+
"use_mse_search": false,
|
511 |
+
"use_layer_wise": false,
|
512 |
+
"use_double_quant": false,
|
513 |
+
"double_quant_bits": 8,
|
514 |
+
"double_quant_dtype": "int",
|
515 |
+
"double_quant_use_sym": true,
|
516 |
+
"double_quant_group_size": 256,
|
517 |
+
"quant_lm_head": true,
|
518 |
+
"absorb_to_layer": {},
|
519 |
+
"folding": false
|
520 |
+
}
|
521 |
+
},
|
522 |
+
"('model.layers.3.mlp.up_proj.orig_layer', 'Linear')": {
|
523 |
+
"teq": {
|
524 |
+
"dtype": "int",
|
525 |
+
"bits": 4,
|
526 |
+
"use_sym": false,
|
527 |
+
"group_size": 128,
|
528 |
+
"group_dim": 1,
|
529 |
+
"use_full_range": false,
|
530 |
+
"use_mse_search": false,
|
531 |
+
"use_layer_wise": false,
|
532 |
+
"use_double_quant": false,
|
533 |
+
"double_quant_bits": 8,
|
534 |
+
"double_quant_dtype": "int",
|
535 |
+
"double_quant_use_sym": true,
|
536 |
+
"double_quant_group_size": 256,
|
537 |
+
"quant_lm_head": true,
|
538 |
+
"absorb_to_layer": {},
|
539 |
+
"folding": false
|
540 |
+
}
|
541 |
+
},
|
542 |
+
"('model.layers.3.mlp.down_proj.orig_layer', 'Linear')": {
|
543 |
+
"teq": {
|
544 |
+
"dtype": "int",
|
545 |
+
"bits": 4,
|
546 |
+
"use_sym": false,
|
547 |
+
"group_size": 128,
|
548 |
+
"group_dim": 1,
|
549 |
+
"use_full_range": false,
|
550 |
+
"use_mse_search": false,
|
551 |
+
"use_layer_wise": false,
|
552 |
+
"use_double_quant": false,
|
553 |
+
"double_quant_bits": 8,
|
554 |
+
"double_quant_dtype": "int",
|
555 |
+
"double_quant_use_sym": true,
|
556 |
+
"double_quant_group_size": 256,
|
557 |
+
"quant_lm_head": true,
|
558 |
+
"absorb_to_layer": {},
|
559 |
+
"folding": false
|
560 |
+
}
|
561 |
+
},
|
562 |
+
"('model.layers.4.self_attn.q_proj.orig_layer', 'Linear')": {
|
563 |
+
"teq": {
|
564 |
+
"dtype": "int",
|
565 |
+
"bits": 4,
|
566 |
+
"use_sym": false,
|
567 |
+
"group_size": 128,
|
568 |
+
"group_dim": 1,
|
569 |
+
"use_full_range": false,
|
570 |
+
"use_mse_search": false,
|
571 |
+
"use_layer_wise": false,
|
572 |
+
"use_double_quant": false,
|
573 |
+
"double_quant_bits": 8,
|
574 |
+
"double_quant_dtype": "int",
|
575 |
+
"double_quant_use_sym": true,
|
576 |
+
"double_quant_group_size": 256,
|
577 |
+
"quant_lm_head": true,
|
578 |
+
"absorb_to_layer": {},
|
579 |
+
"folding": false
|
580 |
+
}
|
581 |
+
},
|
582 |
+
"('model.layers.4.self_attn.k_proj.orig_layer', 'Linear')": {
|
583 |
+
"teq": {
|
584 |
+
"dtype": "int",
|
585 |
+
"bits": 4,
|
586 |
+
"use_sym": false,
|
587 |
+
"group_size": 128,
|
588 |
+
"group_dim": 1,
|
589 |
+
"use_full_range": false,
|
590 |
+
"use_mse_search": false,
|
591 |
+
"use_layer_wise": false,
|
592 |
+
"use_double_quant": false,
|
593 |
+
"double_quant_bits": 8,
|
594 |
+
"double_quant_dtype": "int",
|
595 |
+
"double_quant_use_sym": true,
|
596 |
+
"double_quant_group_size": 256,
|
597 |
+
"quant_lm_head": true,
|
598 |
+
"absorb_to_layer": {},
|
599 |
+
"folding": false
|
600 |
+
}
|
601 |
+
},
|
602 |
+
"('model.layers.4.self_attn.v_proj.orig_layer', 'Linear')": {
|
603 |
+
"teq": {
|
604 |
+
"dtype": "int",
|
605 |
+
"bits": 4,
|
606 |
+
"use_sym": false,
|
607 |
+
"group_size": 128,
|
608 |
+
"group_dim": 1,
|
609 |
+
"use_full_range": false,
|
610 |
+
"use_mse_search": false,
|
611 |
+
"use_layer_wise": false,
|
612 |
+
"use_double_quant": false,
|
613 |
+
"double_quant_bits": 8,
|
614 |
+
"double_quant_dtype": "int",
|
615 |
+
"double_quant_use_sym": true,
|
616 |
+
"double_quant_group_size": 256,
|
617 |
+
"quant_lm_head": true,
|
618 |
+
"absorb_to_layer": {},
|
619 |
+
"folding": false
|
620 |
+
}
|
621 |
+
},
|
622 |
+
"('model.layers.4.self_attn.o_proj.orig_layer', 'Linear')": {
|
623 |
+
"teq": {
|
624 |
+
"dtype": "int",
|
625 |
+
"bits": 4,
|
626 |
+
"use_sym": false,
|
627 |
+
"group_size": 128,
|
628 |
+
"group_dim": 1,
|
629 |
+
"use_full_range": false,
|
630 |
+
"use_mse_search": false,
|
631 |
+
"use_layer_wise": false,
|
632 |
+
"use_double_quant": false,
|
633 |
+
"double_quant_bits": 8,
|
634 |
+
"double_quant_dtype": "int",
|
635 |
+
"double_quant_use_sym": true,
|
636 |
+
"double_quant_group_size": 256,
|
637 |
+
"quant_lm_head": true,
|
638 |
+
"absorb_to_layer": {},
|
639 |
+
"folding": false
|
640 |
+
}
|
641 |
+
},
|
642 |
+
"('model.layers.4.mlp.gate_proj.orig_layer', 'Linear')": {
|
643 |
+
"teq": {
|
644 |
+
"dtype": "int",
|
645 |
+
"bits": 4,
|
646 |
+
"use_sym": false,
|
647 |
+
"group_size": 128,
|
648 |
+
"group_dim": 1,
|
649 |
+
"use_full_range": false,
|
650 |
+
"use_mse_search": false,
|
651 |
+
"use_layer_wise": false,
|
652 |
+
"use_double_quant": false,
|
653 |
+
"double_quant_bits": 8,
|
654 |
+
"double_quant_dtype": "int",
|
655 |
+
"double_quant_use_sym": true,
|
656 |
+
"double_quant_group_size": 256,
|
657 |
+
"quant_lm_head": true,
|
658 |
+
"absorb_to_layer": {},
|
659 |
+
"folding": false
|
660 |
+
}
|
661 |
+
},
|
662 |
+
"('model.layers.4.mlp.up_proj.orig_layer', 'Linear')": {
|
663 |
+
"teq": {
|
664 |
+
"dtype": "int",
|
665 |
+
"bits": 4,
|
666 |
+
"use_sym": false,
|
667 |
+
"group_size": 128,
|
668 |
+
"group_dim": 1,
|
669 |
+
"use_full_range": false,
|
670 |
+
"use_mse_search": false,
|
671 |
+
"use_layer_wise": false,
|
672 |
+
"use_double_quant": false,
|
673 |
+
"double_quant_bits": 8,
|
674 |
+
"double_quant_dtype": "int",
|
675 |
+
"double_quant_use_sym": true,
|
676 |
+
"double_quant_group_size": 256,
|
677 |
+
"quant_lm_head": true,
|
678 |
+
"absorb_to_layer": {},
|
679 |
+
"folding": false
|
680 |
+
}
|
681 |
+
},
|
682 |
+
"('model.layers.4.mlp.down_proj.orig_layer', 'Linear')": {
|
683 |
+
"teq": {
|
684 |
+
"dtype": "int",
|
685 |
+
"bits": 4,
|
686 |
+
"use_sym": false,
|
687 |
+
"group_size": 128,
|
688 |
+
"group_dim": 1,
|
689 |
+
"use_full_range": false,
|
690 |
+
"use_mse_search": false,
|
691 |
+
"use_layer_wise": false,
|
692 |
+
"use_double_quant": false,
|
693 |
+
"double_quant_bits": 8,
|
694 |
+
"double_quant_dtype": "int",
|
695 |
+
"double_quant_use_sym": true,
|
696 |
+
"double_quant_group_size": 256,
|
697 |
+
"quant_lm_head": true,
|
698 |
+
"absorb_to_layer": {},
|
699 |
+
"folding": false
|
700 |
+
}
|
701 |
+
},
|
702 |
+
"('model.layers.5.self_attn.q_proj.orig_layer', 'Linear')": {
|
703 |
+
"teq": {
|
704 |
+
"dtype": "int",
|
705 |
+
"bits": 4,
|
706 |
+
"use_sym": false,
|
707 |
+
"group_size": 128,
|
708 |
+
"group_dim": 1,
|
709 |
+
"use_full_range": false,
|
710 |
+
"use_mse_search": false,
|
711 |
+
"use_layer_wise": false,
|
712 |
+
"use_double_quant": false,
|
713 |
+
"double_quant_bits": 8,
|
714 |
+
"double_quant_dtype": "int",
|
715 |
+
"double_quant_use_sym": true,
|
716 |
+
"double_quant_group_size": 256,
|
717 |
+
"quant_lm_head": true,
|
718 |
+
"absorb_to_layer": {},
|
719 |
+
"folding": false
|
720 |
+
}
|
721 |
+
},
|
722 |
+
"('model.layers.5.self_attn.k_proj.orig_layer', 'Linear')": {
|
723 |
+
"teq": {
|
724 |
+
"dtype": "int",
|
725 |
+
"bits": 4,
|
726 |
+
"use_sym": false,
|
727 |
+
"group_size": 128,
|
728 |
+
"group_dim": 1,
|
729 |
+
"use_full_range": false,
|
730 |
+
"use_mse_search": false,
|
731 |
+
"use_layer_wise": false,
|
732 |
+
"use_double_quant": false,
|
733 |
+
"double_quant_bits": 8,
|
734 |
+
"double_quant_dtype": "int",
|
735 |
+
"double_quant_use_sym": true,
|
736 |
+
"double_quant_group_size": 256,
|
737 |
+
"quant_lm_head": true,
|
738 |
+
"absorb_to_layer": {},
|
739 |
+
"folding": false
|
740 |
+
}
|
741 |
+
},
|
742 |
+
"('model.layers.5.self_attn.v_proj.orig_layer', 'Linear')": {
|
743 |
+
"teq": {
|
744 |
+
"dtype": "int",
|
745 |
+
"bits": 4,
|
746 |
+
"use_sym": false,
|
747 |
+
"group_size": 128,
|
748 |
+
"group_dim": 1,
|
749 |
+
"use_full_range": false,
|
750 |
+
"use_mse_search": false,
|
751 |
+
"use_layer_wise": false,
|
752 |
+
"use_double_quant": false,
|
753 |
+
"double_quant_bits": 8,
|
754 |
+
"double_quant_dtype": "int",
|
755 |
+
"double_quant_use_sym": true,
|
756 |
+
"double_quant_group_size": 256,
|
757 |
+
"quant_lm_head": true,
|
758 |
+
"absorb_to_layer": {},
|
759 |
+
"folding": false
|
760 |
+
}
|
761 |
+
},
|
762 |
+
"('model.layers.5.self_attn.o_proj.orig_layer', 'Linear')": {
|
763 |
+
"teq": {
|
764 |
+
"dtype": "int",
|
765 |
+
"bits": 4,
|
766 |
+
"use_sym": false,
|
767 |
+
"group_size": 128,
|
768 |
+
"group_dim": 1,
|
769 |
+
"use_full_range": false,
|
770 |
+
"use_mse_search": false,
|
771 |
+
"use_layer_wise": false,
|
772 |
+
"use_double_quant": false,
|
773 |
+
"double_quant_bits": 8,
|
774 |
+
"double_quant_dtype": "int",
|
775 |
+
"double_quant_use_sym": true,
|
776 |
+
"double_quant_group_size": 256,
|
777 |
+
"quant_lm_head": true,
|
778 |
+
"absorb_to_layer": {},
|
779 |
+
"folding": false
|
780 |
+
}
|
781 |
+
},
|
782 |
+
"('model.layers.5.mlp.gate_proj.orig_layer', 'Linear')": {
|
783 |
+
"teq": {
|
784 |
+
"dtype": "int",
|
785 |
+
"bits": 4,
|
786 |
+
"use_sym": false,
|
787 |
+
"group_size": 128,
|
788 |
+
"group_dim": 1,
|
789 |
+
"use_full_range": false,
|
790 |
+
"use_mse_search": false,
|
791 |
+
"use_layer_wise": false,
|
792 |
+
"use_double_quant": false,
|
793 |
+
"double_quant_bits": 8,
|
794 |
+
"double_quant_dtype": "int",
|
795 |
+
"double_quant_use_sym": true,
|
796 |
+
"double_quant_group_size": 256,
|
797 |
+
"quant_lm_head": true,
|
798 |
+
"absorb_to_layer": {},
|
799 |
+
"folding": false
|
800 |
+
}
|
801 |
+
},
|
802 |
+
"('model.layers.5.mlp.up_proj.orig_layer', 'Linear')": {
|
803 |
+
"teq": {
|
804 |
+
"dtype": "int",
|
805 |
+
"bits": 4,
|
806 |
+
"use_sym": false,
|
807 |
+
"group_size": 128,
|
808 |
+
"group_dim": 1,
|
809 |
+
"use_full_range": false,
|
810 |
+
"use_mse_search": false,
|
811 |
+
"use_layer_wise": false,
|
812 |
+
"use_double_quant": false,
|
813 |
+
"double_quant_bits": 8,
|
814 |
+
"double_quant_dtype": "int",
|
815 |
+
"double_quant_use_sym": true,
|
816 |
+
"double_quant_group_size": 256,
|
817 |
+
"quant_lm_head": true,
|
818 |
+
"absorb_to_layer": {},
|
819 |
+
"folding": false
|
820 |
+
}
|
821 |
+
},
|
822 |
+
"('model.layers.5.mlp.down_proj.orig_layer', 'Linear')": {
|
823 |
+
"teq": {
|
824 |
+
"dtype": "int",
|
825 |
+
"bits": 4,
|
826 |
+
"use_sym": false,
|
827 |
+
"group_size": 128,
|
828 |
+
"group_dim": 1,
|
829 |
+
"use_full_range": false,
|
830 |
+
"use_mse_search": false,
|
831 |
+
"use_layer_wise": false,
|
832 |
+
"use_double_quant": false,
|
833 |
+
"double_quant_bits": 8,
|
834 |
+
"double_quant_dtype": "int",
|
835 |
+
"double_quant_use_sym": true,
|
836 |
+
"double_quant_group_size": 256,
|
837 |
+
"quant_lm_head": true,
|
838 |
+
"absorb_to_layer": {},
|
839 |
+
"folding": false
|
840 |
+
}
|
841 |
+
},
|
842 |
+
"('model.layers.6.self_attn.q_proj.orig_layer', 'Linear')": {
|
843 |
+
"teq": {
|
844 |
+
"dtype": "int",
|
845 |
+
"bits": 4,
|
846 |
+
"use_sym": false,
|
847 |
+
"group_size": 128,
|
848 |
+
"group_dim": 1,
|
849 |
+
"use_full_range": false,
|
850 |
+
"use_mse_search": false,
|
851 |
+
"use_layer_wise": false,
|
852 |
+
"use_double_quant": false,
|
853 |
+
"double_quant_bits": 8,
|
854 |
+
"double_quant_dtype": "int",
|
855 |
+
"double_quant_use_sym": true,
|
856 |
+
"double_quant_group_size": 256,
|
857 |
+
"quant_lm_head": true,
|
858 |
+
"absorb_to_layer": {},
|
859 |
+
"folding": false
|
860 |
+
}
|
861 |
+
},
|
862 |
+
"('model.layers.6.self_attn.k_proj.orig_layer', 'Linear')": {
|
863 |
+
"teq": {
|
864 |
+
"dtype": "int",
|
865 |
+
"bits": 4,
|
866 |
+
"use_sym": false,
|
867 |
+
"group_size": 128,
|
868 |
+
"group_dim": 1,
|
869 |
+
"use_full_range": false,
|
870 |
+
"use_mse_search": false,
|
871 |
+
"use_layer_wise": false,
|
872 |
+
"use_double_quant": false,
|
873 |
+
"double_quant_bits": 8,
|
874 |
+
"double_quant_dtype": "int",
|
875 |
+
"double_quant_use_sym": true,
|
876 |
+
"double_quant_group_size": 256,
|
877 |
+
"quant_lm_head": true,
|
878 |
+
"absorb_to_layer": {},
|
879 |
+
"folding": false
|
880 |
+
}
|
881 |
+
},
|
882 |
+
"('model.layers.6.self_attn.v_proj.orig_layer', 'Linear')": {
|
883 |
+
"teq": {
|
884 |
+
"dtype": "int",
|
885 |
+
"bits": 4,
|
886 |
+
"use_sym": false,
|
887 |
+
"group_size": 128,
|
888 |
+
"group_dim": 1,
|
889 |
+
"use_full_range": false,
|
890 |
+
"use_mse_search": false,
|
891 |
+
"use_layer_wise": false,
|
892 |
+
"use_double_quant": false,
|
893 |
+
"double_quant_bits": 8,
|
894 |
+
"double_quant_dtype": "int",
|
895 |
+
"double_quant_use_sym": true,
|
896 |
+
"double_quant_group_size": 256,
|
897 |
+
"quant_lm_head": true,
|
898 |
+
"absorb_to_layer": {},
|
899 |
+
"folding": false
|
900 |
+
}
|
901 |
+
},
|
902 |
+
"('model.layers.6.self_attn.o_proj.orig_layer', 'Linear')": {
|
903 |
+
"teq": {
|
904 |
+
"dtype": "int",
|
905 |
+
"bits": 4,
|
906 |
+
"use_sym": false,
|
907 |
+
"group_size": 128,
|
908 |
+
"group_dim": 1,
|
909 |
+
"use_full_range": false,
|
910 |
+
"use_mse_search": false,
|
911 |
+
"use_layer_wise": false,
|
912 |
+
"use_double_quant": false,
|
913 |
+
"double_quant_bits": 8,
|
914 |
+
"double_quant_dtype": "int",
|
915 |
+
"double_quant_use_sym": true,
|
916 |
+
"double_quant_group_size": 256,
|
917 |
+
"quant_lm_head": true,
|
918 |
+
"absorb_to_layer": {},
|
919 |
+
"folding": false
|
920 |
+
}
|
921 |
+
},
|
922 |
+
"('model.layers.6.mlp.gate_proj.orig_layer', 'Linear')": {
|
923 |
+
"teq": {
|
924 |
+
"dtype": "int",
|
925 |
+
"bits": 4,
|
926 |
+
"use_sym": false,
|
927 |
+
"group_size": 128,
|
928 |
+
"group_dim": 1,
|
929 |
+
"use_full_range": false,
|
930 |
+
"use_mse_search": false,
|
931 |
+
"use_layer_wise": false,
|
932 |
+
"use_double_quant": false,
|
933 |
+
"double_quant_bits": 8,
|
934 |
+
"double_quant_dtype": "int",
|
935 |
+
"double_quant_use_sym": true,
|
936 |
+
"double_quant_group_size": 256,
|
937 |
+
"quant_lm_head": true,
|
938 |
+
"absorb_to_layer": {},
|
939 |
+
"folding": false
|
940 |
+
}
|
941 |
+
},
|
942 |
+
"('model.layers.6.mlp.up_proj.orig_layer', 'Linear')": {
|
943 |
+
"teq": {
|
944 |
+
"dtype": "int",
|
945 |
+
"bits": 4,
|
946 |
+
"use_sym": false,
|
947 |
+
"group_size": 128,
|
948 |
+
"group_dim": 1,
|
949 |
+
"use_full_range": false,
|
950 |
+
"use_mse_search": false,
|
951 |
+
"use_layer_wise": false,
|
952 |
+
"use_double_quant": false,
|
953 |
+
"double_quant_bits": 8,
|
954 |
+
"double_quant_dtype": "int",
|
955 |
+
"double_quant_use_sym": true,
|
956 |
+
"double_quant_group_size": 256,
|
957 |
+
"quant_lm_head": true,
|
958 |
+
"absorb_to_layer": {},
|
959 |
+
"folding": false
|
960 |
+
}
|
961 |
+
},
|
962 |
+
"('model.layers.6.mlp.down_proj.orig_layer', 'Linear')": {
|
963 |
+
"teq": {
|
964 |
+
"dtype": "int",
|
965 |
+
"bits": 4,
|
966 |
+
"use_sym": false,
|
967 |
+
"group_size": 128,
|
968 |
+
"group_dim": 1,
|
969 |
+
"use_full_range": false,
|
970 |
+
"use_mse_search": false,
|
971 |
+
"use_layer_wise": false,
|
972 |
+
"use_double_quant": false,
|
973 |
+
"double_quant_bits": 8,
|
974 |
+
"double_quant_dtype": "int",
|
975 |
+
"double_quant_use_sym": true,
|
976 |
+
"double_quant_group_size": 256,
|
977 |
+
"quant_lm_head": true,
|
978 |
+
"absorb_to_layer": {},
|
979 |
+
"folding": false
|
980 |
+
}
|
981 |
+
},
|
982 |
+
"('model.layers.7.self_attn.q_proj.orig_layer', 'Linear')": {
|
983 |
+
"teq": {
|
984 |
+
"dtype": "int",
|
985 |
+
"bits": 4,
|
986 |
+
"use_sym": false,
|
987 |
+
"group_size": 128,
|
988 |
+
"group_dim": 1,
|
989 |
+
"use_full_range": false,
|
990 |
+
"use_mse_search": false,
|
991 |
+
"use_layer_wise": false,
|
992 |
+
"use_double_quant": false,
|
993 |
+
"double_quant_bits": 8,
|
994 |
+
"double_quant_dtype": "int",
|
995 |
+
"double_quant_use_sym": true,
|
996 |
+
"double_quant_group_size": 256,
|
997 |
+
"quant_lm_head": true,
|
998 |
+
"absorb_to_layer": {},
|
999 |
+
"folding": false
|
1000 |
+
}
|
1001 |
+
},
|
1002 |
+
"('model.layers.7.self_attn.k_proj.orig_layer', 'Linear')": {
|
1003 |
+
"teq": {
|
1004 |
+
"dtype": "int",
|
1005 |
+
"bits": 4,
|
1006 |
+
"use_sym": false,
|
1007 |
+
"group_size": 128,
|
1008 |
+
"group_dim": 1,
|
1009 |
+
"use_full_range": false,
|
1010 |
+
"use_mse_search": false,
|
1011 |
+
"use_layer_wise": false,
|
1012 |
+
"use_double_quant": false,
|
1013 |
+
"double_quant_bits": 8,
|
1014 |
+
"double_quant_dtype": "int",
|
1015 |
+
"double_quant_use_sym": true,
|
1016 |
+
"double_quant_group_size": 256,
|
1017 |
+
"quant_lm_head": true,
|
1018 |
+
"absorb_to_layer": {},
|
1019 |
+
"folding": false
|
1020 |
+
}
|
1021 |
+
},
|
1022 |
+
"('model.layers.7.self_attn.v_proj.orig_layer', 'Linear')": {
|
1023 |
+
"teq": {
|
1024 |
+
"dtype": "int",
|
1025 |
+
"bits": 4,
|
1026 |
+
"use_sym": false,
|
1027 |
+
"group_size": 128,
|
1028 |
+
"group_dim": 1,
|
1029 |
+
"use_full_range": false,
|
1030 |
+
"use_mse_search": false,
|
1031 |
+
"use_layer_wise": false,
|
1032 |
+
"use_double_quant": false,
|
1033 |
+
"double_quant_bits": 8,
|
1034 |
+
"double_quant_dtype": "int",
|
1035 |
+
"double_quant_use_sym": true,
|
1036 |
+
"double_quant_group_size": 256,
|
1037 |
+
"quant_lm_head": true,
|
1038 |
+
"absorb_to_layer": {},
|
1039 |
+
"folding": false
|
1040 |
+
}
|
1041 |
+
},
|
1042 |
+
"('model.layers.7.self_attn.o_proj.orig_layer', 'Linear')": {
|
1043 |
+
"teq": {
|
1044 |
+
"dtype": "int",
|
1045 |
+
"bits": 4,
|
1046 |
+
"use_sym": false,
|
1047 |
+
"group_size": 128,
|
1048 |
+
"group_dim": 1,
|
1049 |
+
"use_full_range": false,
|
1050 |
+
"use_mse_search": false,
|
1051 |
+
"use_layer_wise": false,
|
1052 |
+
"use_double_quant": false,
|
1053 |
+
"double_quant_bits": 8,
|
1054 |
+
"double_quant_dtype": "int",
|
1055 |
+
"double_quant_use_sym": true,
|
1056 |
+
"double_quant_group_size": 256,
|
1057 |
+
"quant_lm_head": true,
|
1058 |
+
"absorb_to_layer": {},
|
1059 |
+
"folding": false
|
1060 |
+
}
|
1061 |
+
},
|
1062 |
+
"('model.layers.7.mlp.gate_proj.orig_layer', 'Linear')": {
|
1063 |
+
"teq": {
|
1064 |
+
"dtype": "int",
|
1065 |
+
"bits": 4,
|
1066 |
+
"use_sym": false,
|
1067 |
+
"group_size": 128,
|
1068 |
+
"group_dim": 1,
|
1069 |
+
"use_full_range": false,
|
1070 |
+
"use_mse_search": false,
|
1071 |
+
"use_layer_wise": false,
|
1072 |
+
"use_double_quant": false,
|
1073 |
+
"double_quant_bits": 8,
|
1074 |
+
"double_quant_dtype": "int",
|
1075 |
+
"double_quant_use_sym": true,
|
1076 |
+
"double_quant_group_size": 256,
|
1077 |
+
"quant_lm_head": true,
|
1078 |
+
"absorb_to_layer": {},
|
1079 |
+
"folding": false
|
1080 |
+
}
|
1081 |
+
},
|
1082 |
+
"('model.layers.7.mlp.up_proj.orig_layer', 'Linear')": {
|
1083 |
+
"teq": {
|
1084 |
+
"dtype": "int",
|
1085 |
+
"bits": 4,
|
1086 |
+
"use_sym": false,
|
1087 |
+
"group_size": 128,
|
1088 |
+
"group_dim": 1,
|
1089 |
+
"use_full_range": false,
|
1090 |
+
"use_mse_search": false,
|
1091 |
+
"use_layer_wise": false,
|
1092 |
+
"use_double_quant": false,
|
1093 |
+
"double_quant_bits": 8,
|
1094 |
+
"double_quant_dtype": "int",
|
1095 |
+
"double_quant_use_sym": true,
|
1096 |
+
"double_quant_group_size": 256,
|
1097 |
+
"quant_lm_head": true,
|
1098 |
+
"absorb_to_layer": {},
|
1099 |
+
"folding": false
|
1100 |
+
}
|
1101 |
+
},
|
1102 |
+
"('model.layers.7.mlp.down_proj.orig_layer', 'Linear')": {
|
1103 |
+
"teq": {
|
1104 |
+
"dtype": "int",
|
1105 |
+
"bits": 4,
|
1106 |
+
"use_sym": false,
|
1107 |
+
"group_size": 128,
|
1108 |
+
"group_dim": 1,
|
1109 |
+
"use_full_range": false,
|
1110 |
+
"use_mse_search": false,
|
1111 |
+
"use_layer_wise": false,
|
1112 |
+
"use_double_quant": false,
|
1113 |
+
"double_quant_bits": 8,
|
1114 |
+
"double_quant_dtype": "int",
|
1115 |
+
"double_quant_use_sym": true,
|
1116 |
+
"double_quant_group_size": 256,
|
1117 |
+
"quant_lm_head": true,
|
1118 |
+
"absorb_to_layer": {},
|
1119 |
+
"folding": false
|
1120 |
+
}
|
1121 |
+
},
|
1122 |
+
"('model.layers.8.self_attn.q_proj.orig_layer', 'Linear')": {
|
1123 |
+
"teq": {
|
1124 |
+
"dtype": "int",
|
1125 |
+
"bits": 4,
|
1126 |
+
"use_sym": false,
|
1127 |
+
"group_size": 128,
|
1128 |
+
"group_dim": 1,
|
1129 |
+
"use_full_range": false,
|
1130 |
+
"use_mse_search": false,
|
1131 |
+
"use_layer_wise": false,
|
1132 |
+
"use_double_quant": false,
|
1133 |
+
"double_quant_bits": 8,
|
1134 |
+
"double_quant_dtype": "int",
|
1135 |
+
"double_quant_use_sym": true,
|
1136 |
+
"double_quant_group_size": 256,
|
1137 |
+
"quant_lm_head": true,
|
1138 |
+
"absorb_to_layer": {},
|
1139 |
+
"folding": false
|
1140 |
+
}
|
1141 |
+
},
|
1142 |
+
"('model.layers.8.self_attn.k_proj.orig_layer', 'Linear')": {
|
1143 |
+
"teq": {
|
1144 |
+
"dtype": "int",
|
1145 |
+
"bits": 4,
|
1146 |
+
"use_sym": false,
|
1147 |
+
"group_size": 128,
|
1148 |
+
"group_dim": 1,
|
1149 |
+
"use_full_range": false,
|
1150 |
+
"use_mse_search": false,
|
1151 |
+
"use_layer_wise": false,
|
1152 |
+
"use_double_quant": false,
|
1153 |
+
"double_quant_bits": 8,
|
1154 |
+
"double_quant_dtype": "int",
|
1155 |
+
"double_quant_use_sym": true,
|
1156 |
+
"double_quant_group_size": 256,
|
1157 |
+
"quant_lm_head": true,
|
1158 |
+
"absorb_to_layer": {},
|
1159 |
+
"folding": false
|
1160 |
+
}
|
1161 |
+
},
|
1162 |
+
"('model.layers.8.self_attn.v_proj.orig_layer', 'Linear')": {
|
1163 |
+
"teq": {
|
1164 |
+
"dtype": "int",
|
1165 |
+
"bits": 4,
|
1166 |
+
"use_sym": false,
|
1167 |
+
"group_size": 128,
|
1168 |
+
"group_dim": 1,
|
1169 |
+
"use_full_range": false,
|
1170 |
+
"use_mse_search": false,
|
1171 |
+
"use_layer_wise": false,
|
1172 |
+
"use_double_quant": false,
|
1173 |
+
"double_quant_bits": 8,
|
1174 |
+
"double_quant_dtype": "int",
|
1175 |
+
"double_quant_use_sym": true,
|
1176 |
+
"double_quant_group_size": 256,
|
1177 |
+
"quant_lm_head": true,
|
1178 |
+
"absorb_to_layer": {},
|
1179 |
+
"folding": false
|
1180 |
+
}
|
1181 |
+
},
|
1182 |
+
"('model.layers.8.self_attn.o_proj.orig_layer', 'Linear')": {
|
1183 |
+
"teq": {
|
1184 |
+
"dtype": "int",
|
1185 |
+
"bits": 4,
|
1186 |
+
"use_sym": false,
|
1187 |
+
"group_size": 128,
|
1188 |
+
"group_dim": 1,
|
1189 |
+
"use_full_range": false,
|
1190 |
+
"use_mse_search": false,
|
1191 |
+
"use_layer_wise": false,
|
1192 |
+
"use_double_quant": false,
|
1193 |
+
"double_quant_bits": 8,
|
1194 |
+
"double_quant_dtype": "int",
|
1195 |
+
"double_quant_use_sym": true,
|
1196 |
+
"double_quant_group_size": 256,
|
1197 |
+
"quant_lm_head": true,
|
1198 |
+
"absorb_to_layer": {},
|
1199 |
+
"folding": false
|
1200 |
+
}
|
1201 |
+
},
|
1202 |
+
"('model.layers.8.mlp.gate_proj.orig_layer', 'Linear')": {
|
1203 |
+
"teq": {
|
1204 |
+
"dtype": "int",
|
1205 |
+
"bits": 4,
|
1206 |
+
"use_sym": false,
|
1207 |
+
"group_size": 128,
|
1208 |
+
"group_dim": 1,
|
1209 |
+
"use_full_range": false,
|
1210 |
+
"use_mse_search": false,
|
1211 |
+
"use_layer_wise": false,
|
1212 |
+
"use_double_quant": false,
|
1213 |
+
"double_quant_bits": 8,
|
1214 |
+
"double_quant_dtype": "int",
|
1215 |
+
"double_quant_use_sym": true,
|
1216 |
+
"double_quant_group_size": 256,
|
1217 |
+
"quant_lm_head": true,
|
1218 |
+
"absorb_to_layer": {},
|
1219 |
+
"folding": false
|
1220 |
+
}
|
1221 |
+
},
|
1222 |
+
"('model.layers.8.mlp.up_proj.orig_layer', 'Linear')": {
|
1223 |
+
"teq": {
|
1224 |
+
"dtype": "int",
|
1225 |
+
"bits": 4,
|
1226 |
+
"use_sym": false,
|
1227 |
+
"group_size": 128,
|
1228 |
+
"group_dim": 1,
|
1229 |
+
"use_full_range": false,
|
1230 |
+
"use_mse_search": false,
|
1231 |
+
"use_layer_wise": false,
|
1232 |
+
"use_double_quant": false,
|
1233 |
+
"double_quant_bits": 8,
|
1234 |
+
"double_quant_dtype": "int",
|
1235 |
+
"double_quant_use_sym": true,
|
1236 |
+
"double_quant_group_size": 256,
|
1237 |
+
"quant_lm_head": true,
|
1238 |
+
"absorb_to_layer": {},
|
1239 |
+
"folding": false
|
1240 |
+
}
|
1241 |
+
},
|
1242 |
+
"('model.layers.8.mlp.down_proj.orig_layer', 'Linear')": {
|
1243 |
+
"teq": {
|
1244 |
+
"dtype": "int",
|
1245 |
+
"bits": 4,
|
1246 |
+
"use_sym": false,
|
1247 |
+
"group_size": 128,
|
1248 |
+
"group_dim": 1,
|
1249 |
+
"use_full_range": false,
|
1250 |
+
"use_mse_search": false,
|
1251 |
+
"use_layer_wise": false,
|
1252 |
+
"use_double_quant": false,
|
1253 |
+
"double_quant_bits": 8,
|
1254 |
+
"double_quant_dtype": "int",
|
1255 |
+
"double_quant_use_sym": true,
|
1256 |
+
"double_quant_group_size": 256,
|
1257 |
+
"quant_lm_head": true,
|
1258 |
+
"absorb_to_layer": {},
|
1259 |
+
"folding": false
|
1260 |
+
}
|
1261 |
+
},
|
1262 |
+
"('model.layers.9.self_attn.q_proj.orig_layer', 'Linear')": {
|
1263 |
+
"teq": {
|
1264 |
+
"dtype": "int",
|
1265 |
+
"bits": 4,
|
1266 |
+
"use_sym": false,
|
1267 |
+
"group_size": 128,
|
1268 |
+
"group_dim": 1,
|
1269 |
+
"use_full_range": false,
|
1270 |
+
"use_mse_search": false,
|
1271 |
+
"use_layer_wise": false,
|
1272 |
+
"use_double_quant": false,
|
1273 |
+
"double_quant_bits": 8,
|
1274 |
+
"double_quant_dtype": "int",
|
1275 |
+
"double_quant_use_sym": true,
|
1276 |
+
"double_quant_group_size": 256,
|
1277 |
+
"quant_lm_head": true,
|
1278 |
+
"absorb_to_layer": {},
|
1279 |
+
"folding": false
|
1280 |
+
}
|
1281 |
+
},
|
1282 |
+
"('model.layers.9.self_attn.k_proj.orig_layer', 'Linear')": {
|
1283 |
+
"teq": {
|
1284 |
+
"dtype": "int",
|
1285 |
+
"bits": 4,
|
1286 |
+
"use_sym": false,
|
1287 |
+
"group_size": 128,
|
1288 |
+
"group_dim": 1,
|
1289 |
+
"use_full_range": false,
|
1290 |
+
"use_mse_search": false,
|
1291 |
+
"use_layer_wise": false,
|
1292 |
+
"use_double_quant": false,
|
1293 |
+
"double_quant_bits": 8,
|
1294 |
+
"double_quant_dtype": "int",
|
1295 |
+
"double_quant_use_sym": true,
|
1296 |
+
"double_quant_group_size": 256,
|
1297 |
+
"quant_lm_head": true,
|
1298 |
+
"absorb_to_layer": {},
|
1299 |
+
"folding": false
|
1300 |
+
}
|
1301 |
+
},
|
1302 |
+
"('model.layers.9.self_attn.v_proj.orig_layer', 'Linear')": {
|
1303 |
+
"teq": {
|
1304 |
+
"dtype": "int",
|
1305 |
+
"bits": 4,
|
1306 |
+
"use_sym": false,
|
1307 |
+
"group_size": 128,
|
1308 |
+
"group_dim": 1,
|
1309 |
+
"use_full_range": false,
|
1310 |
+
"use_mse_search": false,
|
1311 |
+
"use_layer_wise": false,
|
1312 |
+
"use_double_quant": false,
|
1313 |
+
"double_quant_bits": 8,
|
1314 |
+
"double_quant_dtype": "int",
|
1315 |
+
"double_quant_use_sym": true,
|
1316 |
+
"double_quant_group_size": 256,
|
1317 |
+
"quant_lm_head": true,
|
1318 |
+
"absorb_to_layer": {},
|
1319 |
+
"folding": false
|
1320 |
+
}
|
1321 |
+
},
|
1322 |
+
"('model.layers.9.self_attn.o_proj.orig_layer', 'Linear')": {
|
1323 |
+
"teq": {
|
1324 |
+
"dtype": "int",
|
1325 |
+
"bits": 4,
|
1326 |
+
"use_sym": false,
|
1327 |
+
"group_size": 128,
|
1328 |
+
"group_dim": 1,
|
1329 |
+
"use_full_range": false,
|
1330 |
+
"use_mse_search": false,
|
1331 |
+
"use_layer_wise": false,
|
1332 |
+
"use_double_quant": false,
|
1333 |
+
"double_quant_bits": 8,
|
1334 |
+
"double_quant_dtype": "int",
|
1335 |
+
"double_quant_use_sym": true,
|
1336 |
+
"double_quant_group_size": 256,
|
1337 |
+
"quant_lm_head": true,
|
1338 |
+
"absorb_to_layer": {},
|
1339 |
+
"folding": false
|
1340 |
+
}
|
1341 |
+
},
|
1342 |
+
"('model.layers.9.mlp.gate_proj.orig_layer', 'Linear')": {
|
1343 |
+
"teq": {
|
1344 |
+
"dtype": "int",
|
1345 |
+
"bits": 4,
|
1346 |
+
"use_sym": false,
|
1347 |
+
"group_size": 128,
|
1348 |
+
"group_dim": 1,
|
1349 |
+
"use_full_range": false,
|
1350 |
+
"use_mse_search": false,
|
1351 |
+
"use_layer_wise": false,
|
1352 |
+
"use_double_quant": false,
|
1353 |
+
"double_quant_bits": 8,
|
1354 |
+
"double_quant_dtype": "int",
|
1355 |
+
"double_quant_use_sym": true,
|
1356 |
+
"double_quant_group_size": 256,
|
1357 |
+
"quant_lm_head": true,
|
1358 |
+
"absorb_to_layer": {},
|
1359 |
+
"folding": false
|
1360 |
+
}
|
1361 |
+
},
|
1362 |
+
"('model.layers.9.mlp.up_proj.orig_layer', 'Linear')": {
|
1363 |
+
"teq": {
|
1364 |
+
"dtype": "int",
|
1365 |
+
"bits": 4,
|
1366 |
+
"use_sym": false,
|
1367 |
+
"group_size": 128,
|
1368 |
+
"group_dim": 1,
|
1369 |
+
"use_full_range": false,
|
1370 |
+
"use_mse_search": false,
|
1371 |
+
"use_layer_wise": false,
|
1372 |
+
"use_double_quant": false,
|
1373 |
+
"double_quant_bits": 8,
|
1374 |
+
"double_quant_dtype": "int",
|
1375 |
+
"double_quant_use_sym": true,
|
1376 |
+
"double_quant_group_size": 256,
|
1377 |
+
"quant_lm_head": true,
|
1378 |
+
"absorb_to_layer": {},
|
1379 |
+
"folding": false
|
1380 |
+
}
|
1381 |
+
},
|
1382 |
+
"('model.layers.9.mlp.down_proj.orig_layer', 'Linear')": {
|
1383 |
+
"teq": {
|
1384 |
+
"dtype": "int",
|
1385 |
+
"bits": 4,
|
1386 |
+
"use_sym": false,
|
1387 |
+
"group_size": 128,
|
1388 |
+
"group_dim": 1,
|
1389 |
+
"use_full_range": false,
|
1390 |
+
"use_mse_search": false,
|
1391 |
+
"use_layer_wise": false,
|
1392 |
+
"use_double_quant": false,
|
1393 |
+
"double_quant_bits": 8,
|
1394 |
+
"double_quant_dtype": "int",
|
1395 |
+
"double_quant_use_sym": true,
|
1396 |
+
"double_quant_group_size": 256,
|
1397 |
+
"quant_lm_head": true,
|
1398 |
+
"absorb_to_layer": {},
|
1399 |
+
"folding": false
|
1400 |
+
}
|
1401 |
+
},
|
1402 |
+
"('model.layers.10.self_attn.q_proj.orig_layer', 'Linear')": {
|
1403 |
+
"teq": {
|
1404 |
+
"dtype": "int",
|
1405 |
+
"bits": 4,
|
1406 |
+
"use_sym": false,
|
1407 |
+
"group_size": 128,
|
1408 |
+
"group_dim": 1,
|
1409 |
+
"use_full_range": false,
|
1410 |
+
"use_mse_search": false,
|
1411 |
+
"use_layer_wise": false,
|
1412 |
+
"use_double_quant": false,
|
1413 |
+
"double_quant_bits": 8,
|
1414 |
+
"double_quant_dtype": "int",
|
1415 |
+
"double_quant_use_sym": true,
|
1416 |
+
"double_quant_group_size": 256,
|
1417 |
+
"quant_lm_head": true,
|
1418 |
+
"absorb_to_layer": {},
|
1419 |
+
"folding": false
|
1420 |
+
}
|
1421 |
+
},
|
1422 |
+
"('model.layers.10.self_attn.k_proj.orig_layer', 'Linear')": {
|
1423 |
+
"teq": {
|
1424 |
+
"dtype": "int",
|
1425 |
+
"bits": 4,
|
1426 |
+
"use_sym": false,
|
1427 |
+
"group_size": 128,
|
1428 |
+
"group_dim": 1,
|
1429 |
+
"use_full_range": false,
|
1430 |
+
"use_mse_search": false,
|
1431 |
+
"use_layer_wise": false,
|
1432 |
+
"use_double_quant": false,
|
1433 |
+
"double_quant_bits": 8,
|
1434 |
+
"double_quant_dtype": "int",
|
1435 |
+
"double_quant_use_sym": true,
|
1436 |
+
"double_quant_group_size": 256,
|
1437 |
+
"quant_lm_head": true,
|
1438 |
+
"absorb_to_layer": {},
|
1439 |
+
"folding": false
|
1440 |
+
}
|
1441 |
+
},
|
1442 |
+
"('model.layers.10.self_attn.v_proj.orig_layer', 'Linear')": {
|
1443 |
+
"teq": {
|
1444 |
+
"dtype": "int",
|
1445 |
+
"bits": 4,
|
1446 |
+
"use_sym": false,
|
1447 |
+
"group_size": 128,
|
1448 |
+
"group_dim": 1,
|
1449 |
+
"use_full_range": false,
|
1450 |
+
"use_mse_search": false,
|
1451 |
+
"use_layer_wise": false,
|
1452 |
+
"use_double_quant": false,
|
1453 |
+
"double_quant_bits": 8,
|
1454 |
+
"double_quant_dtype": "int",
|
1455 |
+
"double_quant_use_sym": true,
|
1456 |
+
"double_quant_group_size": 256,
|
1457 |
+
"quant_lm_head": true,
|
1458 |
+
"absorb_to_layer": {},
|
1459 |
+
"folding": false
|
1460 |
+
}
|
1461 |
+
},
|
1462 |
+
"('model.layers.10.self_attn.o_proj.orig_layer', 'Linear')": {
|
1463 |
+
"teq": {
|
1464 |
+
"dtype": "int",
|
1465 |
+
"bits": 4,
|
1466 |
+
"use_sym": false,
|
1467 |
+
"group_size": 128,
|
1468 |
+
"group_dim": 1,
|
1469 |
+
"use_full_range": false,
|
1470 |
+
"use_mse_search": false,
|
1471 |
+
"use_layer_wise": false,
|
1472 |
+
"use_double_quant": false,
|
1473 |
+
"double_quant_bits": 8,
|
1474 |
+
"double_quant_dtype": "int",
|
1475 |
+
"double_quant_use_sym": true,
|
1476 |
+
"double_quant_group_size": 256,
|
1477 |
+
"quant_lm_head": true,
|
1478 |
+
"absorb_to_layer": {},
|
1479 |
+
"folding": false
|
1480 |
+
}
|
1481 |
+
},
|
1482 |
+
"('model.layers.10.mlp.gate_proj.orig_layer', 'Linear')": {
|
1483 |
+
"teq": {
|
1484 |
+
"dtype": "int",
|
1485 |
+
"bits": 4,
|
1486 |
+
"use_sym": false,
|
1487 |
+
"group_size": 128,
|
1488 |
+
"group_dim": 1,
|
1489 |
+
"use_full_range": false,
|
1490 |
+
"use_mse_search": false,
|
1491 |
+
"use_layer_wise": false,
|
1492 |
+
"use_double_quant": false,
|
1493 |
+
"double_quant_bits": 8,
|
1494 |
+
"double_quant_dtype": "int",
|
1495 |
+
"double_quant_use_sym": true,
|
1496 |
+
"double_quant_group_size": 256,
|
1497 |
+
"quant_lm_head": true,
|
1498 |
+
"absorb_to_layer": {},
|
1499 |
+
"folding": false
|
1500 |
+
}
|
1501 |
+
},
|
1502 |
+
"('model.layers.10.mlp.up_proj.orig_layer', 'Linear')": {
|
1503 |
+
"teq": {
|
1504 |
+
"dtype": "int",
|
1505 |
+
"bits": 4,
|
1506 |
+
"use_sym": false,
|
1507 |
+
"group_size": 128,
|
1508 |
+
"group_dim": 1,
|
1509 |
+
"use_full_range": false,
|
1510 |
+
"use_mse_search": false,
|
1511 |
+
"use_layer_wise": false,
|
1512 |
+
"use_double_quant": false,
|
1513 |
+
"double_quant_bits": 8,
|
1514 |
+
"double_quant_dtype": "int",
|
1515 |
+
"double_quant_use_sym": true,
|
1516 |
+
"double_quant_group_size": 256,
|
1517 |
+
"quant_lm_head": true,
|
1518 |
+
"absorb_to_layer": {},
|
1519 |
+
"folding": false
|
1520 |
+
}
|
1521 |
+
},
|
1522 |
+
"('model.layers.10.mlp.down_proj.orig_layer', 'Linear')": {
|
1523 |
+
"teq": {
|
1524 |
+
"dtype": "int",
|
1525 |
+
"bits": 4,
|
1526 |
+
"use_sym": false,
|
1527 |
+
"group_size": 128,
|
1528 |
+
"group_dim": 1,
|
1529 |
+
"use_full_range": false,
|
1530 |
+
"use_mse_search": false,
|
1531 |
+
"use_layer_wise": false,
|
1532 |
+
"use_double_quant": false,
|
1533 |
+
"double_quant_bits": 8,
|
1534 |
+
"double_quant_dtype": "int",
|
1535 |
+
"double_quant_use_sym": true,
|
1536 |
+
"double_quant_group_size": 256,
|
1537 |
+
"quant_lm_head": true,
|
1538 |
+
"absorb_to_layer": {},
|
1539 |
+
"folding": false
|
1540 |
+
}
|
1541 |
+
},
|
1542 |
+
"('model.layers.11.self_attn.q_proj.orig_layer', 'Linear')": {
|
1543 |
+
"teq": {
|
1544 |
+
"dtype": "int",
|
1545 |
+
"bits": 4,
|
1546 |
+
"use_sym": false,
|
1547 |
+
"group_size": 128,
|
1548 |
+
"group_dim": 1,
|
1549 |
+
"use_full_range": false,
|
1550 |
+
"use_mse_search": false,
|
1551 |
+
"use_layer_wise": false,
|
1552 |
+
"use_double_quant": false,
|
1553 |
+
"double_quant_bits": 8,
|
1554 |
+
"double_quant_dtype": "int",
|
1555 |
+
"double_quant_use_sym": true,
|
1556 |
+
"double_quant_group_size": 256,
|
1557 |
+
"quant_lm_head": true,
|
1558 |
+
"absorb_to_layer": {},
|
1559 |
+
"folding": false
|
1560 |
+
}
|
1561 |
+
},
|
1562 |
+
"('model.layers.11.self_attn.k_proj.orig_layer', 'Linear')": {
|
1563 |
+
"teq": {
|
1564 |
+
"dtype": "int",
|
1565 |
+
"bits": 4,
|
1566 |
+
"use_sym": false,
|
1567 |
+
"group_size": 128,
|
1568 |
+
"group_dim": 1,
|
1569 |
+
"use_full_range": false,
|
1570 |
+
"use_mse_search": false,
|
1571 |
+
"use_layer_wise": false,
|
1572 |
+
"use_double_quant": false,
|
1573 |
+
"double_quant_bits": 8,
|
1574 |
+
"double_quant_dtype": "int",
|
1575 |
+
"double_quant_use_sym": true,
|
1576 |
+
"double_quant_group_size": 256,
|
1577 |
+
"quant_lm_head": true,
|
1578 |
+
"absorb_to_layer": {},
|
1579 |
+
"folding": false
|
1580 |
+
}
|
1581 |
+
},
|
1582 |
+
"('model.layers.11.self_attn.v_proj.orig_layer', 'Linear')": {
|
1583 |
+
"teq": {
|
1584 |
+
"dtype": "int",
|
1585 |
+
"bits": 4,
|
1586 |
+
"use_sym": false,
|
1587 |
+
"group_size": 128,
|
1588 |
+
"group_dim": 1,
|
1589 |
+
"use_full_range": false,
|
1590 |
+
"use_mse_search": false,
|
1591 |
+
"use_layer_wise": false,
|
1592 |
+
"use_double_quant": false,
|
1593 |
+
"double_quant_bits": 8,
|
1594 |
+
"double_quant_dtype": "int",
|
1595 |
+
"double_quant_use_sym": true,
|
1596 |
+
"double_quant_group_size": 256,
|
1597 |
+
"quant_lm_head": true,
|
1598 |
+
"absorb_to_layer": {},
|
1599 |
+
"folding": false
|
1600 |
+
}
|
1601 |
+
},
|
1602 |
+
"('model.layers.11.self_attn.o_proj.orig_layer', 'Linear')": {
|
1603 |
+
"teq": {
|
1604 |
+
"dtype": "int",
|
1605 |
+
"bits": 4,
|
1606 |
+
"use_sym": false,
|
1607 |
+
"group_size": 128,
|
1608 |
+
"group_dim": 1,
|
1609 |
+
"use_full_range": false,
|
1610 |
+
"use_mse_search": false,
|
1611 |
+
"use_layer_wise": false,
|
1612 |
+
"use_double_quant": false,
|
1613 |
+
"double_quant_bits": 8,
|
1614 |
+
"double_quant_dtype": "int",
|
1615 |
+
"double_quant_use_sym": true,
|
1616 |
+
"double_quant_group_size": 256,
|
1617 |
+
"quant_lm_head": true,
|
1618 |
+
"absorb_to_layer": {},
|
1619 |
+
"folding": false
|
1620 |
+
}
|
1621 |
+
},
|
1622 |
+
"('model.layers.11.mlp.gate_proj.orig_layer', 'Linear')": {
|
1623 |
+
"teq": {
|
1624 |
+
"dtype": "int",
|
1625 |
+
"bits": 4,
|
1626 |
+
"use_sym": false,
|
1627 |
+
"group_size": 128,
|
1628 |
+
"group_dim": 1,
|
1629 |
+
"use_full_range": false,
|
1630 |
+
"use_mse_search": false,
|
1631 |
+
"use_layer_wise": false,
|
1632 |
+
"use_double_quant": false,
|
1633 |
+
"double_quant_bits": 8,
|
1634 |
+
"double_quant_dtype": "int",
|
1635 |
+
"double_quant_use_sym": true,
|
1636 |
+
"double_quant_group_size": 256,
|
1637 |
+
"quant_lm_head": true,
|
1638 |
+
"absorb_to_layer": {},
|
1639 |
+
"folding": false
|
1640 |
+
}
|
1641 |
+
},
|
1642 |
+
"('model.layers.11.mlp.up_proj.orig_layer', 'Linear')": {
|
1643 |
+
"teq": {
|
1644 |
+
"dtype": "int",
|
1645 |
+
"bits": 4,
|
1646 |
+
"use_sym": false,
|
1647 |
+
"group_size": 128,
|
1648 |
+
"group_dim": 1,
|
1649 |
+
"use_full_range": false,
|
1650 |
+
"use_mse_search": false,
|
1651 |
+
"use_layer_wise": false,
|
1652 |
+
"use_double_quant": false,
|
1653 |
+
"double_quant_bits": 8,
|
1654 |
+
"double_quant_dtype": "int",
|
1655 |
+
"double_quant_use_sym": true,
|
1656 |
+
"double_quant_group_size": 256,
|
1657 |
+
"quant_lm_head": true,
|
1658 |
+
"absorb_to_layer": {},
|
1659 |
+
"folding": false
|
1660 |
+
}
|
1661 |
+
},
|
1662 |
+
"('model.layers.11.mlp.down_proj.orig_layer', 'Linear')": {
|
1663 |
+
"teq": {
|
1664 |
+
"dtype": "int",
|
1665 |
+
"bits": 4,
|
1666 |
+
"use_sym": false,
|
1667 |
+
"group_size": 128,
|
1668 |
+
"group_dim": 1,
|
1669 |
+
"use_full_range": false,
|
1670 |
+
"use_mse_search": false,
|
1671 |
+
"use_layer_wise": false,
|
1672 |
+
"use_double_quant": false,
|
1673 |
+
"double_quant_bits": 8,
|
1674 |
+
"double_quant_dtype": "int",
|
1675 |
+
"double_quant_use_sym": true,
|
1676 |
+
"double_quant_group_size": 256,
|
1677 |
+
"quant_lm_head": true,
|
1678 |
+
"absorb_to_layer": {},
|
1679 |
+
"folding": false
|
1680 |
+
}
|
1681 |
+
},
|
1682 |
+
"('model.layers.12.self_attn.q_proj.orig_layer', 'Linear')": {
|
1683 |
+
"teq": {
|
1684 |
+
"dtype": "int",
|
1685 |
+
"bits": 4,
|
1686 |
+
"use_sym": false,
|
1687 |
+
"group_size": 128,
|
1688 |
+
"group_dim": 1,
|
1689 |
+
"use_full_range": false,
|
1690 |
+
"use_mse_search": false,
|
1691 |
+
"use_layer_wise": false,
|
1692 |
+
"use_double_quant": false,
|
1693 |
+
"double_quant_bits": 8,
|
1694 |
+
"double_quant_dtype": "int",
|
1695 |
+
"double_quant_use_sym": true,
|
1696 |
+
"double_quant_group_size": 256,
|
1697 |
+
"quant_lm_head": true,
|
1698 |
+
"absorb_to_layer": {},
|
1699 |
+
"folding": false
|
1700 |
+
}
|
1701 |
+
},
|
1702 |
+
"('model.layers.12.self_attn.k_proj.orig_layer', 'Linear')": {
|
1703 |
+
"teq": {
|
1704 |
+
"dtype": "int",
|
1705 |
+
"bits": 4,
|
1706 |
+
"use_sym": false,
|
1707 |
+
"group_size": 128,
|
1708 |
+
"group_dim": 1,
|
1709 |
+
"use_full_range": false,
|
1710 |
+
"use_mse_search": false,
|
1711 |
+
"use_layer_wise": false,
|
1712 |
+
"use_double_quant": false,
|
1713 |
+
"double_quant_bits": 8,
|
1714 |
+
"double_quant_dtype": "int",
|
1715 |
+
"double_quant_use_sym": true,
|
1716 |
+
"double_quant_group_size": 256,
|
1717 |
+
"quant_lm_head": true,
|
1718 |
+
"absorb_to_layer": {},
|
1719 |
+
"folding": false
|
1720 |
+
}
|
1721 |
+
},
|
1722 |
+
"('model.layers.12.self_attn.v_proj.orig_layer', 'Linear')": {
|
1723 |
+
"teq": {
|
1724 |
+
"dtype": "int",
|
1725 |
+
"bits": 4,
|
1726 |
+
"use_sym": false,
|
1727 |
+
"group_size": 128,
|
1728 |
+
"group_dim": 1,
|
1729 |
+
"use_full_range": false,
|
1730 |
+
"use_mse_search": false,
|
1731 |
+
"use_layer_wise": false,
|
1732 |
+
"use_double_quant": false,
|
1733 |
+
"double_quant_bits": 8,
|
1734 |
+
"double_quant_dtype": "int",
|
1735 |
+
"double_quant_use_sym": true,
|
1736 |
+
"double_quant_group_size": 256,
|
1737 |
+
"quant_lm_head": true,
|
1738 |
+
"absorb_to_layer": {},
|
1739 |
+
"folding": false
|
1740 |
+
}
|
1741 |
+
},
|
1742 |
+
"('model.layers.12.self_attn.o_proj.orig_layer', 'Linear')": {
|
1743 |
+
"teq": {
|
1744 |
+
"dtype": "int",
|
1745 |
+
"bits": 4,
|
1746 |
+
"use_sym": false,
|
1747 |
+
"group_size": 128,
|
1748 |
+
"group_dim": 1,
|
1749 |
+
"use_full_range": false,
|
1750 |
+
"use_mse_search": false,
|
1751 |
+
"use_layer_wise": false,
|
1752 |
+
"use_double_quant": false,
|
1753 |
+
"double_quant_bits": 8,
|
1754 |
+
"double_quant_dtype": "int",
|
1755 |
+
"double_quant_use_sym": true,
|
1756 |
+
"double_quant_group_size": 256,
|
1757 |
+
"quant_lm_head": true,
|
1758 |
+
"absorb_to_layer": {},
|
1759 |
+
"folding": false
|
1760 |
+
}
|
1761 |
+
},
|
1762 |
+
"('model.layers.12.mlp.gate_proj.orig_layer', 'Linear')": {
|
1763 |
+
"teq": {
|
1764 |
+
"dtype": "int",
|
1765 |
+
"bits": 4,
|
1766 |
+
"use_sym": false,
|
1767 |
+
"group_size": 128,
|
1768 |
+
"group_dim": 1,
|
1769 |
+
"use_full_range": false,
|
1770 |
+
"use_mse_search": false,
|
1771 |
+
"use_layer_wise": false,
|
1772 |
+
"use_double_quant": false,
|
1773 |
+
"double_quant_bits": 8,
|
1774 |
+
"double_quant_dtype": "int",
|
1775 |
+
"double_quant_use_sym": true,
|
1776 |
+
"double_quant_group_size": 256,
|
1777 |
+
"quant_lm_head": true,
|
1778 |
+
"absorb_to_layer": {},
|
1779 |
+
"folding": false
|
1780 |
+
}
|
1781 |
+
},
|
1782 |
+
"('model.layers.12.mlp.up_proj.orig_layer', 'Linear')": {
|
1783 |
+
"teq": {
|
1784 |
+
"dtype": "int",
|
1785 |
+
"bits": 4,
|
1786 |
+
"use_sym": false,
|
1787 |
+
"group_size": 128,
|
1788 |
+
"group_dim": 1,
|
1789 |
+
"use_full_range": false,
|
1790 |
+
"use_mse_search": false,
|
1791 |
+
"use_layer_wise": false,
|
1792 |
+
"use_double_quant": false,
|
1793 |
+
"double_quant_bits": 8,
|
1794 |
+
"double_quant_dtype": "int",
|
1795 |
+
"double_quant_use_sym": true,
|
1796 |
+
"double_quant_group_size": 256,
|
1797 |
+
"quant_lm_head": true,
|
1798 |
+
"absorb_to_layer": {},
|
1799 |
+
"folding": false
|
1800 |
+
}
|
1801 |
+
},
|
1802 |
+
"('model.layers.12.mlp.down_proj.orig_layer', 'Linear')": {
|
1803 |
+
"teq": {
|
1804 |
+
"dtype": "int",
|
1805 |
+
"bits": 4,
|
1806 |
+
"use_sym": false,
|
1807 |
+
"group_size": 128,
|
1808 |
+
"group_dim": 1,
|
1809 |
+
"use_full_range": false,
|
1810 |
+
"use_mse_search": false,
|
1811 |
+
"use_layer_wise": false,
|
1812 |
+
"use_double_quant": false,
|
1813 |
+
"double_quant_bits": 8,
|
1814 |
+
"double_quant_dtype": "int",
|
1815 |
+
"double_quant_use_sym": true,
|
1816 |
+
"double_quant_group_size": 256,
|
1817 |
+
"quant_lm_head": true,
|
1818 |
+
"absorb_to_layer": {},
|
1819 |
+
"folding": false
|
1820 |
+
}
|
1821 |
+
},
|
1822 |
+
"('model.layers.13.self_attn.q_proj.orig_layer', 'Linear')": {
|
1823 |
+
"teq": {
|
1824 |
+
"dtype": "int",
|
1825 |
+
"bits": 4,
|
1826 |
+
"use_sym": false,
|
1827 |
+
"group_size": 128,
|
1828 |
+
"group_dim": 1,
|
1829 |
+
"use_full_range": false,
|
1830 |
+
"use_mse_search": false,
|
1831 |
+
"use_layer_wise": false,
|
1832 |
+
"use_double_quant": false,
|
1833 |
+
"double_quant_bits": 8,
|
1834 |
+
"double_quant_dtype": "int",
|
1835 |
+
"double_quant_use_sym": true,
|
1836 |
+
"double_quant_group_size": 256,
|
1837 |
+
"quant_lm_head": true,
|
1838 |
+
"absorb_to_layer": {},
|
1839 |
+
"folding": false
|
1840 |
+
}
|
1841 |
+
},
|
1842 |
+
"('model.layers.13.self_attn.k_proj.orig_layer', 'Linear')": {
|
1843 |
+
"teq": {
|
1844 |
+
"dtype": "int",
|
1845 |
+
"bits": 4,
|
1846 |
+
"use_sym": false,
|
1847 |
+
"group_size": 128,
|
1848 |
+
"group_dim": 1,
|
1849 |
+
"use_full_range": false,
|
1850 |
+
"use_mse_search": false,
|
1851 |
+
"use_layer_wise": false,
|
1852 |
+
"use_double_quant": false,
|
1853 |
+
"double_quant_bits": 8,
|
1854 |
+
"double_quant_dtype": "int",
|
1855 |
+
"double_quant_use_sym": true,
|
1856 |
+
"double_quant_group_size": 256,
|
1857 |
+
"quant_lm_head": true,
|
1858 |
+
"absorb_to_layer": {},
|
1859 |
+
"folding": false
|
1860 |
+
}
|
1861 |
+
},
|
1862 |
+
"('model.layers.13.self_attn.v_proj.orig_layer', 'Linear')": {
|
1863 |
+
"teq": {
|
1864 |
+
"dtype": "int",
|
1865 |
+
"bits": 4,
|
1866 |
+
"use_sym": false,
|
1867 |
+
"group_size": 128,
|
1868 |
+
"group_dim": 1,
|
1869 |
+
"use_full_range": false,
|
1870 |
+
"use_mse_search": false,
|
1871 |
+
"use_layer_wise": false,
|
1872 |
+
"use_double_quant": false,
|
1873 |
+
"double_quant_bits": 8,
|
1874 |
+
"double_quant_dtype": "int",
|
1875 |
+
"double_quant_use_sym": true,
|
1876 |
+
"double_quant_group_size": 256,
|
1877 |
+
"quant_lm_head": true,
|
1878 |
+
"absorb_to_layer": {},
|
1879 |
+
"folding": false
|
1880 |
+
}
|
1881 |
+
},
|
1882 |
+
"('model.layers.13.self_attn.o_proj.orig_layer', 'Linear')": {
|
1883 |
+
"teq": {
|
1884 |
+
"dtype": "int",
|
1885 |
+
"bits": 4,
|
1886 |
+
"use_sym": false,
|
1887 |
+
"group_size": 128,
|
1888 |
+
"group_dim": 1,
|
1889 |
+
"use_full_range": false,
|
1890 |
+
"use_mse_search": false,
|
1891 |
+
"use_layer_wise": false,
|
1892 |
+
"use_double_quant": false,
|
1893 |
+
"double_quant_bits": 8,
|
1894 |
+
"double_quant_dtype": "int",
|
1895 |
+
"double_quant_use_sym": true,
|
1896 |
+
"double_quant_group_size": 256,
|
1897 |
+
"quant_lm_head": true,
|
1898 |
+
"absorb_to_layer": {},
|
1899 |
+
"folding": false
|
1900 |
+
}
|
1901 |
+
},
|
1902 |
+
"('model.layers.13.mlp.gate_proj.orig_layer', 'Linear')": {
|
1903 |
+
"teq": {
|
1904 |
+
"dtype": "int",
|
1905 |
+
"bits": 4,
|
1906 |
+
"use_sym": false,
|
1907 |
+
"group_size": 128,
|
1908 |
+
"group_dim": 1,
|
1909 |
+
"use_full_range": false,
|
1910 |
+
"use_mse_search": false,
|
1911 |
+
"use_layer_wise": false,
|
1912 |
+
"use_double_quant": false,
|
1913 |
+
"double_quant_bits": 8,
|
1914 |
+
"double_quant_dtype": "int",
|
1915 |
+
"double_quant_use_sym": true,
|
1916 |
+
"double_quant_group_size": 256,
|
1917 |
+
"quant_lm_head": true,
|
1918 |
+
"absorb_to_layer": {},
|
1919 |
+
"folding": false
|
1920 |
+
}
|
1921 |
+
},
|
1922 |
+
"('model.layers.13.mlp.up_proj.orig_layer', 'Linear')": {
|
1923 |
+
"teq": {
|
1924 |
+
"dtype": "int",
|
1925 |
+
"bits": 4,
|
1926 |
+
"use_sym": false,
|
1927 |
+
"group_size": 128,
|
1928 |
+
"group_dim": 1,
|
1929 |
+
"use_full_range": false,
|
1930 |
+
"use_mse_search": false,
|
1931 |
+
"use_layer_wise": false,
|
1932 |
+
"use_double_quant": false,
|
1933 |
+
"double_quant_bits": 8,
|
1934 |
+
"double_quant_dtype": "int",
|
1935 |
+
"double_quant_use_sym": true,
|
1936 |
+
"double_quant_group_size": 256,
|
1937 |
+
"quant_lm_head": true,
|
1938 |
+
"absorb_to_layer": {},
|
1939 |
+
"folding": false
|
1940 |
+
}
|
1941 |
+
},
|
1942 |
+
"('model.layers.13.mlp.down_proj.orig_layer', 'Linear')": {
|
1943 |
+
"teq": {
|
1944 |
+
"dtype": "int",
|
1945 |
+
"bits": 4,
|
1946 |
+
"use_sym": false,
|
1947 |
+
"group_size": 128,
|
1948 |
+
"group_dim": 1,
|
1949 |
+
"use_full_range": false,
|
1950 |
+
"use_mse_search": false,
|
1951 |
+
"use_layer_wise": false,
|
1952 |
+
"use_double_quant": false,
|
1953 |
+
"double_quant_bits": 8,
|
1954 |
+
"double_quant_dtype": "int",
|
1955 |
+
"double_quant_use_sym": true,
|
1956 |
+
"double_quant_group_size": 256,
|
1957 |
+
"quant_lm_head": true,
|
1958 |
+
"absorb_to_layer": {},
|
1959 |
+
"folding": false
|
1960 |
+
}
|
1961 |
+
},
|
1962 |
+
"('model.layers.14.self_attn.q_proj.orig_layer', 'Linear')": {
|
1963 |
+
"teq": {
|
1964 |
+
"dtype": "int",
|
1965 |
+
"bits": 4,
|
1966 |
+
"use_sym": false,
|
1967 |
+
"group_size": 128,
|
1968 |
+
"group_dim": 1,
|
1969 |
+
"use_full_range": false,
|
1970 |
+
"use_mse_search": false,
|
1971 |
+
"use_layer_wise": false,
|
1972 |
+
"use_double_quant": false,
|
1973 |
+
"double_quant_bits": 8,
|
1974 |
+
"double_quant_dtype": "int",
|
1975 |
+
"double_quant_use_sym": true,
|
1976 |
+
"double_quant_group_size": 256,
|
1977 |
+
"quant_lm_head": true,
|
1978 |
+
"absorb_to_layer": {},
|
1979 |
+
"folding": false
|
1980 |
+
}
|
1981 |
+
},
|
1982 |
+
"('model.layers.14.self_attn.k_proj.orig_layer', 'Linear')": {
|
1983 |
+
"teq": {
|
1984 |
+
"dtype": "int",
|
1985 |
+
"bits": 4,
|
1986 |
+
"use_sym": false,
|
1987 |
+
"group_size": 128,
|
1988 |
+
"group_dim": 1,
|
1989 |
+
"use_full_range": false,
|
1990 |
+
"use_mse_search": false,
|
1991 |
+
"use_layer_wise": false,
|
1992 |
+
"use_double_quant": false,
|
1993 |
+
"double_quant_bits": 8,
|
1994 |
+
"double_quant_dtype": "int",
|
1995 |
+
"double_quant_use_sym": true,
|
1996 |
+
"double_quant_group_size": 256,
|
1997 |
+
"quant_lm_head": true,
|
1998 |
+
"absorb_to_layer": {},
|
1999 |
+
"folding": false
|
2000 |
+
}
|
2001 |
+
},
|
2002 |
+
"('model.layers.14.self_attn.v_proj.orig_layer', 'Linear')": {
|
2003 |
+
"teq": {
|
2004 |
+
"dtype": "int",
|
2005 |
+
"bits": 4,
|
2006 |
+
"use_sym": false,
|
2007 |
+
"group_size": 128,
|
2008 |
+
"group_dim": 1,
|
2009 |
+
"use_full_range": false,
|
2010 |
+
"use_mse_search": false,
|
2011 |
+
"use_layer_wise": false,
|
2012 |
+
"use_double_quant": false,
|
2013 |
+
"double_quant_bits": 8,
|
2014 |
+
"double_quant_dtype": "int",
|
2015 |
+
"double_quant_use_sym": true,
|
2016 |
+
"double_quant_group_size": 256,
|
2017 |
+
"quant_lm_head": true,
|
2018 |
+
"absorb_to_layer": {},
|
2019 |
+
"folding": false
|
2020 |
+
}
|
2021 |
+
},
|
2022 |
+
"('model.layers.14.self_attn.o_proj.orig_layer', 'Linear')": {
|
2023 |
+
"teq": {
|
2024 |
+
"dtype": "int",
|
2025 |
+
"bits": 4,
|
2026 |
+
"use_sym": false,
|
2027 |
+
"group_size": 128,
|
2028 |
+
"group_dim": 1,
|
2029 |
+
"use_full_range": false,
|
2030 |
+
"use_mse_search": false,
|
2031 |
+
"use_layer_wise": false,
|
2032 |
+
"use_double_quant": false,
|
2033 |
+
"double_quant_bits": 8,
|
2034 |
+
"double_quant_dtype": "int",
|
2035 |
+
"double_quant_use_sym": true,
|
2036 |
+
"double_quant_group_size": 256,
|
2037 |
+
"quant_lm_head": true,
|
2038 |
+
"absorb_to_layer": {},
|
2039 |
+
"folding": false
|
2040 |
+
}
|
2041 |
+
},
|
2042 |
+
"('model.layers.14.mlp.gate_proj.orig_layer', 'Linear')": {
|
2043 |
+
"teq": {
|
2044 |
+
"dtype": "int",
|
2045 |
+
"bits": 4,
|
2046 |
+
"use_sym": false,
|
2047 |
+
"group_size": 128,
|
2048 |
+
"group_dim": 1,
|
2049 |
+
"use_full_range": false,
|
2050 |
+
"use_mse_search": false,
|
2051 |
+
"use_layer_wise": false,
|
2052 |
+
"use_double_quant": false,
|
2053 |
+
"double_quant_bits": 8,
|
2054 |
+
"double_quant_dtype": "int",
|
2055 |
+
"double_quant_use_sym": true,
|
2056 |
+
"double_quant_group_size": 256,
|
2057 |
+
"quant_lm_head": true,
|
2058 |
+
"absorb_to_layer": {},
|
2059 |
+
"folding": false
|
2060 |
+
}
|
2061 |
+
},
|
2062 |
+
"('model.layers.14.mlp.up_proj.orig_layer', 'Linear')": {
|
2063 |
+
"teq": {
|
2064 |
+
"dtype": "int",
|
2065 |
+
"bits": 4,
|
2066 |
+
"use_sym": false,
|
2067 |
+
"group_size": 128,
|
2068 |
+
"group_dim": 1,
|
2069 |
+
"use_full_range": false,
|
2070 |
+
"use_mse_search": false,
|
2071 |
+
"use_layer_wise": false,
|
2072 |
+
"use_double_quant": false,
|
2073 |
+
"double_quant_bits": 8,
|
2074 |
+
"double_quant_dtype": "int",
|
2075 |
+
"double_quant_use_sym": true,
|
2076 |
+
"double_quant_group_size": 256,
|
2077 |
+
"quant_lm_head": true,
|
2078 |
+
"absorb_to_layer": {},
|
2079 |
+
"folding": false
|
2080 |
+
}
|
2081 |
+
},
|
2082 |
+
"('model.layers.14.mlp.down_proj.orig_layer', 'Linear')": {
|
2083 |
+
"teq": {
|
2084 |
+
"dtype": "int",
|
2085 |
+
"bits": 4,
|
2086 |
+
"use_sym": false,
|
2087 |
+
"group_size": 128,
|
2088 |
+
"group_dim": 1,
|
2089 |
+
"use_full_range": false,
|
2090 |
+
"use_mse_search": false,
|
2091 |
+
"use_layer_wise": false,
|
2092 |
+
"use_double_quant": false,
|
2093 |
+
"double_quant_bits": 8,
|
2094 |
+
"double_quant_dtype": "int",
|
2095 |
+
"double_quant_use_sym": true,
|
2096 |
+
"double_quant_group_size": 256,
|
2097 |
+
"quant_lm_head": true,
|
2098 |
+
"absorb_to_layer": {},
|
2099 |
+
"folding": false
|
2100 |
+
}
|
2101 |
+
},
|
2102 |
+
"('model.layers.15.self_attn.q_proj.orig_layer', 'Linear')": {
|
2103 |
+
"teq": {
|
2104 |
+
"dtype": "int",
|
2105 |
+
"bits": 4,
|
2106 |
+
"use_sym": false,
|
2107 |
+
"group_size": 128,
|
2108 |
+
"group_dim": 1,
|
2109 |
+
"use_full_range": false,
|
2110 |
+
"use_mse_search": false,
|
2111 |
+
"use_layer_wise": false,
|
2112 |
+
"use_double_quant": false,
|
2113 |
+
"double_quant_bits": 8,
|
2114 |
+
"double_quant_dtype": "int",
|
2115 |
+
"double_quant_use_sym": true,
|
2116 |
+
"double_quant_group_size": 256,
|
2117 |
+
"quant_lm_head": true,
|
2118 |
+
"absorb_to_layer": {},
|
2119 |
+
"folding": false
|
2120 |
+
}
|
2121 |
+
},
|
2122 |
+
"('model.layers.15.self_attn.k_proj.orig_layer', 'Linear')": {
|
2123 |
+
"teq": {
|
2124 |
+
"dtype": "int",
|
2125 |
+
"bits": 4,
|
2126 |
+
"use_sym": false,
|
2127 |
+
"group_size": 128,
|
2128 |
+
"group_dim": 1,
|
2129 |
+
"use_full_range": false,
|
2130 |
+
"use_mse_search": false,
|
2131 |
+
"use_layer_wise": false,
|
2132 |
+
"use_double_quant": false,
|
2133 |
+
"double_quant_bits": 8,
|
2134 |
+
"double_quant_dtype": "int",
|
2135 |
+
"double_quant_use_sym": true,
|
2136 |
+
"double_quant_group_size": 256,
|
2137 |
+
"quant_lm_head": true,
|
2138 |
+
"absorb_to_layer": {},
|
2139 |
+
"folding": false
|
2140 |
+
}
|
2141 |
+
},
|
2142 |
+
"('model.layers.15.self_attn.v_proj.orig_layer', 'Linear')": {
|
2143 |
+
"teq": {
|
2144 |
+
"dtype": "int",
|
2145 |
+
"bits": 4,
|
2146 |
+
"use_sym": false,
|
2147 |
+
"group_size": 128,
|
2148 |
+
"group_dim": 1,
|
2149 |
+
"use_full_range": false,
|
2150 |
+
"use_mse_search": false,
|
2151 |
+
"use_layer_wise": false,
|
2152 |
+
"use_double_quant": false,
|
2153 |
+
"double_quant_bits": 8,
|
2154 |
+
"double_quant_dtype": "int",
|
2155 |
+
"double_quant_use_sym": true,
|
2156 |
+
"double_quant_group_size": 256,
|
2157 |
+
"quant_lm_head": true,
|
2158 |
+
"absorb_to_layer": {},
|
2159 |
+
"folding": false
|
2160 |
+
}
|
2161 |
+
},
|
2162 |
+
"('model.layers.15.self_attn.o_proj.orig_layer', 'Linear')": {
|
2163 |
+
"teq": {
|
2164 |
+
"dtype": "int",
|
2165 |
+
"bits": 4,
|
2166 |
+
"use_sym": false,
|
2167 |
+
"group_size": 128,
|
2168 |
+
"group_dim": 1,
|
2169 |
+
"use_full_range": false,
|
2170 |
+
"use_mse_search": false,
|
2171 |
+
"use_layer_wise": false,
|
2172 |
+
"use_double_quant": false,
|
2173 |
+
"double_quant_bits": 8,
|
2174 |
+
"double_quant_dtype": "int",
|
2175 |
+
"double_quant_use_sym": true,
|
2176 |
+
"double_quant_group_size": 256,
|
2177 |
+
"quant_lm_head": true,
|
2178 |
+
"absorb_to_layer": {},
|
2179 |
+
"folding": false
|
2180 |
+
}
|
2181 |
+
},
|
2182 |
+
"('model.layers.15.mlp.gate_proj.orig_layer', 'Linear')": {
|
2183 |
+
"teq": {
|
2184 |
+
"dtype": "int",
|
2185 |
+
"bits": 4,
|
2186 |
+
"use_sym": false,
|
2187 |
+
"group_size": 128,
|
2188 |
+
"group_dim": 1,
|
2189 |
+
"use_full_range": false,
|
2190 |
+
"use_mse_search": false,
|
2191 |
+
"use_layer_wise": false,
|
2192 |
+
"use_double_quant": false,
|
2193 |
+
"double_quant_bits": 8,
|
2194 |
+
"double_quant_dtype": "int",
|
2195 |
+
"double_quant_use_sym": true,
|
2196 |
+
"double_quant_group_size": 256,
|
2197 |
+
"quant_lm_head": true,
|
2198 |
+
"absorb_to_layer": {},
|
2199 |
+
"folding": false
|
2200 |
+
}
|
2201 |
+
},
|
2202 |
+
"('model.layers.15.mlp.up_proj.orig_layer', 'Linear')": {
|
2203 |
+
"teq": {
|
2204 |
+
"dtype": "int",
|
2205 |
+
"bits": 4,
|
2206 |
+
"use_sym": false,
|
2207 |
+
"group_size": 128,
|
2208 |
+
"group_dim": 1,
|
2209 |
+
"use_full_range": false,
|
2210 |
+
"use_mse_search": false,
|
2211 |
+
"use_layer_wise": false,
|
2212 |
+
"use_double_quant": false,
|
2213 |
+
"double_quant_bits": 8,
|
2214 |
+
"double_quant_dtype": "int",
|
2215 |
+
"double_quant_use_sym": true,
|
2216 |
+
"double_quant_group_size": 256,
|
2217 |
+
"quant_lm_head": true,
|
2218 |
+
"absorb_to_layer": {},
|
2219 |
+
"folding": false
|
2220 |
+
}
|
2221 |
+
},
|
2222 |
+
"('model.layers.15.mlp.down_proj.orig_layer', 'Linear')": {
|
2223 |
+
"teq": {
|
2224 |
+
"dtype": "int",
|
2225 |
+
"bits": 4,
|
2226 |
+
"use_sym": false,
|
2227 |
+
"group_size": 128,
|
2228 |
+
"group_dim": 1,
|
2229 |
+
"use_full_range": false,
|
2230 |
+
"use_mse_search": false,
|
2231 |
+
"use_layer_wise": false,
|
2232 |
+
"use_double_quant": false,
|
2233 |
+
"double_quant_bits": 8,
|
2234 |
+
"double_quant_dtype": "int",
|
2235 |
+
"double_quant_use_sym": true,
|
2236 |
+
"double_quant_group_size": 256,
|
2237 |
+
"quant_lm_head": true,
|
2238 |
+
"absorb_to_layer": {},
|
2239 |
+
"folding": false
|
2240 |
+
}
|
2241 |
+
},
|
2242 |
+
"('lm_head.orig_layer', 'Linear')": {
|
2243 |
+
"teq": {
|
2244 |
+
"dtype": "int",
|
2245 |
+
"bits": 4,
|
2246 |
+
"use_sym": false,
|
2247 |
+
"group_size": 128,
|
2248 |
+
"group_dim": 1,
|
2249 |
+
"use_full_range": false,
|
2250 |
+
"use_mse_search": false,
|
2251 |
+
"use_layer_wise": false,
|
2252 |
+
"use_double_quant": false,
|
2253 |
+
"double_quant_bits": 8,
|
2254 |
+
"double_quant_dtype": "int",
|
2255 |
+
"double_quant_use_sym": true,
|
2256 |
+
"double_quant_group_size": 256,
|
2257 |
+
"quant_lm_head": true,
|
2258 |
+
"absorb_to_layer": {},
|
2259 |
+
"folding": false
|
2260 |
+
}
|
2261 |
+
}
|
2262 |
+
}
|
quantized_weight.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:881d451ec7358c02de3f131c885e50277efd9d0d68d77b86aff6f0418b290345
|
3 |
+
size 1695497302
|
requirements_conda_neural-compressor-3.3.1.txt
ADDED
@@ -0,0 +1,164 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# This file may be used to create an environment using:
|
2 |
+
# $ conda create --name <env> --file <this file>
|
3 |
+
# platform: linux-64
|
4 |
+
# created-by: conda 24.11.3
|
5 |
+
_libgcc_mutex=0.1=main
|
6 |
+
_openmp_mutex=5.1=1_gnu
|
7 |
+
absl-py=2.2.2=pypi_0
|
8 |
+
accelerate=1.6.0=pypi_0
|
9 |
+
aiohappyeyeballs=2.6.1=pypi_0
|
10 |
+
aiohttp=3.11.18=pypi_0
|
11 |
+
aiosignal=1.3.2=pypi_0
|
12 |
+
annotated-types=0.7.0=pypi_0
|
13 |
+
astunparse=1.6.3=pypi_0
|
14 |
+
attrs=25.3.0=pypi_0
|
15 |
+
auto-round=0.5.1=pypi_0
|
16 |
+
bzip2=1.0.8=h5eee18b_6
|
17 |
+
ca-certificates=2025.2.25=h06a4308_0
|
18 |
+
certifi=2025.1.31=pypi_0
|
19 |
+
chardet=5.2.0=pypi_0
|
20 |
+
charset-normalizer=3.4.1=pypi_0
|
21 |
+
click=8.1.8=pypi_0
|
22 |
+
colorama=0.4.6=pypi_0
|
23 |
+
contourpy=1.3.2=pypi_0
|
24 |
+
cycler=0.12.1=pypi_0
|
25 |
+
dataproperty=1.1.0=pypi_0
|
26 |
+
datasets=3.5.0=pypi_0
|
27 |
+
deprecated=1.2.18=pypi_0
|
28 |
+
dill=0.3.8=pypi_0
|
29 |
+
docker-pycreds=0.4.0=pypi_0
|
30 |
+
einops=0.8.1=pypi_0
|
31 |
+
evaluate=0.4.3=pypi_0
|
32 |
+
filelock=3.13.1=pypi_0
|
33 |
+
flatbuffers=25.2.10=pypi_0
|
34 |
+
fonttools=4.57.0=pypi_0
|
35 |
+
frozenlist=1.6.0=pypi_0
|
36 |
+
fsspec=2024.6.1=pypi_0
|
37 |
+
gast=0.6.0=pypi_0
|
38 |
+
gitdb=4.0.12=pypi_0
|
39 |
+
gitpython=3.1.44=pypi_0
|
40 |
+
google-pasta=0.2.0=pypi_0
|
41 |
+
grpcio=1.71.0=pypi_0
|
42 |
+
h5py=3.13.0=pypi_0
|
43 |
+
huggingface-hub=0.30.2=pypi_0
|
44 |
+
idna=3.10=pypi_0
|
45 |
+
iniconfig=2.1.0=pypi_0
|
46 |
+
jinja2=3.1.4=pypi_0
|
47 |
+
joblib=1.4.2=pypi_0
|
48 |
+
jsonlines=4.0.0=pypi_0
|
49 |
+
keras=3.9.2=pypi_0
|
50 |
+
kiwisolver=1.4.8=pypi_0
|
51 |
+
ld_impl_linux-64=2.40=h12ee557_0
|
52 |
+
libclang=18.1.1=pypi_0
|
53 |
+
libffi=3.4.4=h6a678d5_1
|
54 |
+
libgcc-ng=11.2.0=h1234567_1
|
55 |
+
libgomp=11.2.0=h1234567_1
|
56 |
+
libstdcxx-ng=11.2.0=h1234567_1
|
57 |
+
libuuid=1.41.5=h5eee18b_0
|
58 |
+
llvmlite=0.44.0=pypi_0
|
59 |
+
lm-eval=0.4.3=pypi_0
|
60 |
+
lxml=5.4.0=pypi_0
|
61 |
+
markdown=3.8=pypi_0
|
62 |
+
markdown-it-py=3.0.0=pypi_0
|
63 |
+
markupsafe=2.1.5=pypi_0
|
64 |
+
matplotlib=3.10.1=pypi_0
|
65 |
+
mbstrdecoder=1.1.4=pypi_0
|
66 |
+
mdurl=0.1.2=pypi_0
|
67 |
+
ml-dtypes=0.5.1=pypi_0
|
68 |
+
more-itertools=10.7.0=pypi_0
|
69 |
+
mpmath=1.3.0=pypi_0
|
70 |
+
multidict=6.4.3=pypi_0
|
71 |
+
multiprocess=0.70.16=pypi_0
|
72 |
+
namex=0.0.9=pypi_0
|
73 |
+
ncurses=6.4=h6a678d5_0
|
74 |
+
networkx=3.3=pypi_0
|
75 |
+
neural-compressor=3.3.1=pypi_0
|
76 |
+
nltk=3.9.1=pypi_0
|
77 |
+
numba=0.61.2=pypi_0
|
78 |
+
numexpr=2.10.2=pypi_0
|
79 |
+
numpy=1.26.4=pypi_0
|
80 |
+
opencv-python-headless=4.11.0.86=pypi_0
|
81 |
+
openssl=3.0.16=h5eee18b_0
|
82 |
+
opt-einsum=3.4.0=pypi_0
|
83 |
+
optree=0.15.0=pypi_0
|
84 |
+
packaging=25.0=pypi_0
|
85 |
+
pandas=2.2.3=pypi_0
|
86 |
+
pathvalidate=3.2.3=pypi_0
|
87 |
+
peft=0.15.2=pypi_0
|
88 |
+
pillow=11.0.0=pypi_0
|
89 |
+
pip=25.0=py311h06a4308_0
|
90 |
+
platformdirs=4.3.7=pypi_0
|
91 |
+
pluggy=1.5.0=pypi_0
|
92 |
+
portalocker=3.1.1=pypi_0
|
93 |
+
prettytable=3.16.0=pypi_0
|
94 |
+
propcache=0.3.1=pypi_0
|
95 |
+
protobuf=5.29.4=pypi_0
|
96 |
+
psutil=7.0.0=pypi_0
|
97 |
+
py-cpuinfo=9.0.0=pypi_0
|
98 |
+
pyarrow=19.0.1=pypi_0
|
99 |
+
pybind11=2.13.6=pypi_0
|
100 |
+
pycocotools=2.0.8=pypi_0
|
101 |
+
pydantic=2.11.3=pypi_0
|
102 |
+
pydantic-core=2.33.1=pypi_0
|
103 |
+
pygments=2.19.1=pypi_0
|
104 |
+
pyparsing=3.2.3=pypi_0
|
105 |
+
pytablewriter=1.2.1=pypi_0
|
106 |
+
pytest=8.3.5=pypi_0
|
107 |
+
python=3.11.11=he870216_0
|
108 |
+
python-dateutil=2.9.0.post0=pypi_0
|
109 |
+
pytz=2025.2=pypi_0
|
110 |
+
pyyaml=6.0.2=pypi_0
|
111 |
+
readline=8.2=h5eee18b_0
|
112 |
+
regex=2024.11.6=pypi_0
|
113 |
+
requests=2.32.3=pypi_0
|
114 |
+
rich=14.0.0=pypi_0
|
115 |
+
rouge-score=0.1.2=pypi_0
|
116 |
+
sacrebleu=2.5.1=pypi_0
|
117 |
+
safetensors=0.5.3=pypi_0
|
118 |
+
schema=0.7.7=pypi_0
|
119 |
+
scikit-learn=1.6.1=pypi_0
|
120 |
+
scipy=1.15.2=pypi_0
|
121 |
+
sentencepiece=0.2.0=pypi_0
|
122 |
+
sentry-sdk=2.27.0=pypi_0
|
123 |
+
setproctitle=1.3.5=pypi_0
|
124 |
+
setuptools=75.8.0=py311h06a4308_0
|
125 |
+
six=1.17.0=pypi_0
|
126 |
+
smmap=5.0.2=pypi_0
|
127 |
+
sqlite=3.45.3=h5eee18b_0
|
128 |
+
sqlitedict=2.1.0=pypi_0
|
129 |
+
sympy=1.13.3=pypi_0
|
130 |
+
tabledata=1.3.4=pypi_0
|
131 |
+
tabulate=0.9.0=pypi_0
|
132 |
+
tbb=2022.1.0=pypi_0
|
133 |
+
tcmlib=1.3.0=pypi_0
|
134 |
+
tcolorpy=0.1.7=pypi_0
|
135 |
+
tensorboard=2.19.0=pypi_0
|
136 |
+
tensorboard-data-server=0.7.2=pypi_0
|
137 |
+
tensorflow-cpu=2.19.0=pypi_0
|
138 |
+
tensorflow-io-gcs-filesystem=0.37.1=pypi_0
|
139 |
+
termcolor=3.0.1=pypi_0
|
140 |
+
threadpoolctl=3.6.0=pypi_0
|
141 |
+
tk=8.6.14=h39e8969_0
|
142 |
+
tokenizers=0.21.1=pypi_0
|
143 |
+
torch=2.7.0+cpu=pypi_0
|
144 |
+
torchaudio=2.7.0+cpu=pypi_0
|
145 |
+
torchvision=0.22.0+cpu=pypi_0
|
146 |
+
tqdm=4.67.1=pypi_0
|
147 |
+
tqdm-multiprocess=0.0.11=pypi_0
|
148 |
+
transformers=4.51.3=pypi_0
|
149 |
+
typepy=1.3.4=pypi_0
|
150 |
+
typing-extensions=4.12.2=pypi_0
|
151 |
+
typing-inspection=0.4.0=pypi_0
|
152 |
+
tzdata=2025.2=pypi_0
|
153 |
+
urllib3=2.4.0=pypi_0
|
154 |
+
wandb=0.19.10=pypi_0
|
155 |
+
wcwidth=0.2.13=pypi_0
|
156 |
+
werkzeug=3.1.3=pypi_0
|
157 |
+
wheel=0.45.1=py311h06a4308_0
|
158 |
+
word2number=1.1=pypi_0
|
159 |
+
wrapt=1.17.2=pypi_0
|
160 |
+
xxhash=3.5.0=pypi_0
|
161 |
+
xz=5.6.4=h5eee18b_1
|
162 |
+
yarl=1.20.0=pypi_0
|
163 |
+
zlib=1.2.13=h5eee18b_1
|
164 |
+
zstandard=0.23.0=pypi_0
|