pooja-ganesh commited on
Commit
18e759a
·
verified ·
1 Parent(s): 7c4f159

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +62 -0
README.md CHANGED
@@ -2,5 +2,67 @@
2
  license: mit
3
  base_model:
4
  - deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
 
 
5
  ---
6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2
  license: mit
3
  base_model:
4
  - deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
5
+ language:
6
+ - en
7
  ---
8
 
9
+ # DeepSeek-R1-Distill-Qwen2.5-Math-1.5B-awq-asym-uint4-g128-lmhead-onnx-hybrid
10
+ - ## Introduction
11
+ This model was created by applying [Quark](https://quark.docs.amd.com/latest/index.html) with calibration samples from Pile dataset.
12
+ - ## Quantization Strategy
13
+ - ***Quantized Layers***: All linear layers
14
+ - ***Weight***: uint4 asymmetric per-group, group_size=128
15
+ - ## Quick Start
16
+ 1. [Download and install Quark](https://quark.docs.amd.com/latest/install.html)
17
+ 2. Run the quantization script in the example folder using the following command line:
18
+ ```sh
19
+ export MODEL_DIR = [local model checkpoint folder] or Qwen2.5-Math-1.5B
20
+ # single GPU
21
+ python quantize_quark.py --model_dir $MODEL_DIR \
22
+ --output_dir output_dir $MODEL_NAME-awq-asym-uint4-g128-lmhead \
23
+ --quant_scheme w_uint4_per_group_asym \
24
+ --num_calib_data 128 \
25
+ --quant_algo awq \
26
+ --dataset pileval_for_awq_benchmark \
27
+ --seq_len 512 \
28
+ --model_export hf_format \
29
+ --data_type bfloat16 \
30
+ --exclude_layers
31
+ # cpu
32
+ python quantize_quark.py --model_dir $MODEL_DIR \
33
+ --output_dir output_dir $MODEL_NAME-awq-asym-uint4-g128-lmhead \
34
+ --quant_scheme w_uint4_per_group_asym \
35
+ --num_calib_data 128 \
36
+ --quant_algo awq \
37
+ --dataset pileval_for_awq_benchmark \
38
+ --seq_len 512 \
39
+ --model_export hf_format \
40
+ --data_type bfloat16 \
41
+ --exclude_layers \
42
+ --device cpu
43
+ ```
44
+
45
+ #### License
46
+ Modifications copyright(c) 2024 Advanced Micro Devices,Inc. All rights reserved.
47
+
48
+ MIT License
49
+
50
+ Copyright (c) 2023 DeepSeek
51
+
52
+ Permission is hereby granted, free of charge, to any person obtaining a copy
53
+ of this software and associated documentation files (the "Software"), to deal
54
+ in the Software without restriction, including without limitation the rights
55
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
56
+ copies of the Software, and to permit persons to whom the Software is
57
+ furnished to do so, subject to the following conditions:
58
+
59
+ The above copyright notice and this permission notice shall be included in all
60
+ copies or substantial portions of the Software.
61
+
62
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
63
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
64
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
65
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
66
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
67
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
68
+ SOFTWARE.