Mirrors
/
hf_text-generation-inference
mirror of https://github.com/huggingface/text-generation-inference.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
hf_text-generation-inference/server/text_generation_server/models
History
Daniël de Kok 4594e6faba Add support for Marlin-quantized models 
This change adds support for Marlin-quantized models. Marlin is an
FP16xINT4 matmul kernel, which provides good speedups decoding batches
of 16-32 tokens. It supports quantized models with symmetric
quantization, groupsize -1 or 128, and 4-bit.

Tested with:

- Llama 2
- Llama 3
- Phi 3
 		2024-06-06 13:16:52 +02:00
..
custom_modeling Add support for Marlin-quantized models 2024-06-06 13:16:52 +02:00
__init__.py Add support for Marlin-quantized models 2024-06-06 13:16:52 +02:00
bloom.py MI300 compatibility (#1764) 2024-05-17 15:30:47 +02:00
causal_lm.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
flash_causal_lm.py feat: move allocation logic to rust (#1835) 2024-06-05 12:18:38 +02:00
flash_cohere.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
flash_dbrx.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
flash_gemma.py Fix (flash) Gemma prefix and enable tests 2024-05-27 09:58:06 +02:00
flash_gpt2.py Add support for Marlin-quantized models 2024-06-06 13:16:52 +02:00
flash_llama.py Add support for exl2 quantization 2024-05-30 11:28:05 +02:00
flash_mistral.py feat: move allocation logic to rust (#1835) 2024-06-05 12:18:38 +02:00
flash_mixtral.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
flash_neox.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
flash_phi.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
flash_qwen2.py feat: move allocation logic to rust (#1835) 2024-06-05 12:18:38 +02:00
flash_rw.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
flash_santacoder.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
flash_starcoder2.py feat: move allocation logic to rust (#1835) 2024-06-05 12:18:38 +02:00
galactica.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
globals.py Purely refactors paged/attention into `layers/attention` and make hardware differences more obvious with 1 file per hardware. (#1986) 2024-05-31 17:57:01 +02:00
gpt_neox.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
idefics.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
idefics2.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
idefics_causal_lm.py Adding Llava-Next (Llava 1.6) with full support. (#1709) 2024-04-09 21:32:00 +02:00
llava_next.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
mamba.py Fix TunableOp bug (#1920) 2024-05-17 18:21:51 +02:00
model.py Use the generation config. (#1808) 2024-04-25 19:41:50 +02:00
mpt.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
opt.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
pali_gemma.py Pali gemma modeling (#1895) 2024-05-16 06:58:47 +02:00
phi.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
rw.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
santacoder.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
seq2seq_lm.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
t5.py MLPSpeculator. (#1865) 2024-05-14 12:33:18 +02:00
types.py chore: add pre-commit (#1569) 2024-02-16 11:58:58 +01:00
vlm_causal_lm.py feat: move allocation logic to rust (#1835) 2024-06-05 12:18:38 +02:00