Skip to content

update quant docs #425

New issue

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

Sign up for GitHub

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

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Apr 24, 2024
Merged

update quant docs #425

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 6 additions & 6 deletions docs/quantization.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -214,7 +214,7 @@ python3 generate.py [--compile] --checkpoint-path ${MODEL_PATH} --prompt "Hello,
```

```
python3 export.py --checkpoint-path ${MODEL_PATH} -d fp32 --quantize "{'linear:int4': {'groupsize' : 32} }" [ --output-pte-path ${MODEL_OUT}/${MODEL_NAME}_int4-gw32.pte | --output-dso-path ${MODEL_OUT}/${MODEL_NAME}_int4-gw32.dso]
python3 export.py --checkpoint-path ${MODEL_PATH} -d fp32 --quantize '{"linear:int4": {"groupsize" : 32} }' [ --output-pte-path ${MODEL_OUT}/${MODEL_NAME}_int4-gw32.pte | --output-dso-path ${MODEL_OUT}/${MODEL_NAME}_int4-gw32.dso]
```
Now you can run your model with the same command as before:

Expand All @@ -227,7 +227,7 @@ To compress your model even more, 4-bit integer quantization may be used. To ach

**TODO (Digant): a8w4dq eager mode support [#335](https://github.com/pytorch/torchchat/issues/335) **
```
python3 export.py --checkpoint-path ${MODEL_PATH} -d fp32 --quantize "{'linear:a8w4dq': {'groupsize' : 7} }" [ --output-pte-path ${MODEL_OUT}/${MODEL_NAME}_8da4w.pte | ...dso... ]
python3 export.py --checkpoint-path ${MODEL_PATH} -d fp32 --quantize '{"linear:a8w4dq": {"groupsize" : 8} }' [ --output-pte-path ${MODEL_OUT}/${MODEL_NAME}_8da4w.pte | ...dso... ]
```

Now you can run your model with the same command as before:
Expand All @@ -237,7 +237,7 @@ python3 generate.py [ --pte-path ${MODEL_OUT}/${MODEL_NAME}_a8w4dq.pte | ...dso.
```

## 4-bit Integer Linear Quantization with GPTQ (gptq)
Compression offers smaller memory footprints (to fit on memory-constrained accelerators and mobile/edge devices) and reduced memory bandwidth (for better performance), but often at the price of quality degradation. GPTQ 4-bit integer quantization may be used to reduce the quality impact. To achieve good accuracy, we recommend the use of groupwise quantization where (small to mid-sized) groups of int4 weights share a scale.
Compression offers smaller memory footprints (to fit on memory-constrained accelerators and mobile/edge devices) and reduced memory bandwidth (for better performance), but often at the price of quality degradation. GPTQ 4-bit integer quantization may be used to reduce the quality impact. To achieve good accuracy, we recommend the use of groupwise quantization where (small to mid-sized) groups of int4 weights share a scale.

**TODO (Jerry): GPTQ quantization documentation [#336](https://github.com/pytorch/torchchat/issues/336) **

Expand All @@ -247,7 +247,7 @@ python3 generate.py [--compile] --checkpoint-path ${MODEL_PATH} --prompt "Hello,
```

```
python3 export.py --checkpoint-path ${MODEL_PATH} -d fp32 --quantize "{'linear:gptq': {'groupsize' : 32} }" [ --output-pte-path ${MODEL_OUT}/${MODEL_NAME}_gptq.pte | ...dso... ]
python3 export.py --checkpoint-path ${MODEL_PATH} -d fp32 --quantize '{"linear:gptq": {"groupsize" : 32} }' [ --output-pte-path ${MODEL_OUT}/${MODEL_NAME}_gptq.pte | ...dso... ]
```
Now you can run your model with the same command as before:

Expand All @@ -257,7 +257,7 @@ python3 generate.py [ --pte-path ${MODEL_OUT}/${MODEL_NAME}_gptq.pte | ...dso...

## 4-bit Integer Linear Quantization with HQQ (hqq)

Compression offers smaller memory footprints (to fit on memory-constrained accelerators and mobile/edge devices) and reduced memory bandwidth (for better performance), but often at the price of quality degradation. GPTQ 4-bit integer quantization may be used to reduce the quality impact, but at the cost of significant additional computation time. HQQ Quantization balances performance, accuracy, and runtime, we recommend the use of groupwise quantization where (small to mid-sized) groups of int4 weights share a scale.
Compression offers smaller memory footprints (to fit on memory-constrained accelerators and mobile/edge devices) and reduced memory bandwidth (for better performance), but often at the price of quality degradation. GPTQ 4-bit integer quantization may be used to reduce the quality impact, but at the cost of significant additional computation time. HQQ Quantization balances performance, accuracy, and runtime, we recommend the use of groupwise quantization where (small to mid-sized) groups of int4 weights share a scale.

**TODO (Zhengxu): HQQ quantization documentation [#337](https://github.com/pytorch/torchchat/issues/336) **

Expand All @@ -267,7 +267,7 @@ python3 generate.py [--compile] --checkpoint-path ${MODEL_PATH} --prompt "Hello,
```

```
python3 export.py --checkpoint-path ${MODEL_PATH} -d fp32 --quantize "{'linear:hqq': {'groupsize' : 32} }" [ --output-pte-path ${MODEL_OUT}/${MODEL_NAME}_hqq.pte | ...dso... ]
python3 export.py --checkpoint-path ${MODEL_PATH} -d fp32 --quantize '{"linear:hqq": {"groupsize" : 32} }' [ --output-pte-path ${MODEL_OUT}/${MODEL_NAME}_hqq.pte | ...dso... ]
```
Now you can run your model with the same command as before:

Expand Down