dongmao zhang and Michael Goin
87525fab92
[bitsandbytes]: support read bnb pre-quantized model ( #5753 )
...
Co-authored-by: Michael Goin <michael@neuralmagic.com >
2024-07-23 23:45:09 +00:00
Simon Mo
3eda4ec780
support ignore patterns in model loader ( #6673 )
2024-07-22 23:59:42 -07:00
youkaichao
c5201240a4
[misc] only tqdm for first rank ( #6672 )
2024-07-22 21:57:27 -07:00
e519ae097a
add tqdm when loading checkpoint shards ( #6569 )
...
Co-authored-by: tianyi.zhao <tianyi.zhao@transwarp.io >
Co-authored-by: youkaichao <youkaichao@126.com >
2024-07-22 20:48:01 -07:00
youkaichao
ce37be7ba0
[misc][distributed] add seed to dummy weights ( #6491 )
2024-07-16 19:16:34 -07:00
Michael Goin
978aed5300
[Kernel][Attention] Separate Attention.kv_scale into k_scale and v_scale ( #6081 )
2024-07-16 15:31:32 -07:00
Robert Shaw
73030b7dae
[ Misc ] Enable Quantizing All Layers of DeekSeekv2 ( #6423 )
2024-07-14 21:38:42 +00:00
Cyrus Leung
0e9164b40a
[mypy] Enable type checking for test directory ( #5017 )
2024-06-15 04:45:31 +00:00
Dipika Sikka and Michael Goin
5884c2b454
[Misc] Update to comply with the new compressed-tensors config ( #5350 )
...
Co-authored-by: Michael Goin <michael@neuralmagic.com >
2024-06-10 03:49:46 +00:00
chenqianfzh
b9c0605a8e
[Feature][Kernel] Support bitsandbytes quantization and QLoRA ( #4776 )
2024-06-01 14:51:10 -06:00
Robert Shaw and Cody Yu
919770957f
[Bugfix] Fix Mistral v0.3 Weight Loading ( #5005 )
...
Co-authored-by: Cody Yu <hao.yu.cody@gmail.com >
2024-05-24 12:28:27 +00:00
a1242324c9
[Kernel] Initial Activation Quantization Support ( #4525 )
...
Co-authored-by: Varun Sundar Rabindranath <varunsundar08@gmail.com >
Co-authored-by: Varun Sundar Rabindranath <varun@neuralmagic.com >
2024-05-23 21:29:18 +00:00
Mor Zusman and Mor Zusman
f0eecee610
[Bugfix] Fix dummy weight for fp8 ( #4916 )
...
Allow dummy load format for fp8,
torch.uniform_ doesn't support FP8 at the moment
Co-authored-by: Mor Zusman <morz@ai21.com >
2024-05-20 18:44:25 +00:00
Prashant Gupta and Travis Johnson
d6e520e170
[Core] Support offline use of local cache for models ( #4374 )
...
Signed-off-by: Prashant Gupta <prashantgupta@us.ibm.com >
Co-authored-by: Travis Johnson <tjohnson31415@gmail.com >
2024-04-27 09:59:55 -07:00
SangBin Cho and Danny Guinther
a88081bf76
[CI] Disable non-lazy string operation on logging ( #4326 )
...
Co-authored-by: Danny Guinther <dguinther@neuralmagic.com >
2024-04-26 00:16:58 -07:00
Cody Yu
a22cdea371
[Kernel][FP8] Initial support with dynamic per-tensor scaling ( #4118 )
...
Provide an initial support to FP8 computation. This PR is inspired by HuggingFace TGI: huggingface/text-generation-inference#1726
This feature can be enabled with --quantization fp8 or -q fp8 when launching an engine.
Algorithm:
We still load a model checkpoint in FP16/BF16. After the weights are loaded, Fp8LinearMethod calculates the per-tensor scaling factor of weights and quantizes the weights accordingly. The scaling factor will then be stored for future use. Meanwhile, the per-tensor scaling factor for activations is calculated in every forward pass.
Initial Results:
Currently tested Mistral-7B on 1xH100. With prompt length ~5 and decoding length 128:
BF16: 1.47s
FP8: 1.66s
I'll try to use larger models and try to find more performance bottleneck. Meanwhile, you're welcome to try this code.
2024-04-20 04:28:57 +00:00
Antoni Baum
69e1d2fb69
[Core] Refactor model loading code ( #4097 )
2024-04-16 11:34:39 -07:00
