Laurent Mazare c388be93e7 Updated quantized phi model (#2099) ...

* Quantized phi in a separate file.

* Add the quantized phi example + rework the model code.

* Improve the phi model.

* Get some generation out.

* Use the appropriate rope shape.

* Tweak the default prompt.

---------

Co-authored-by: Jane Doe <jane.doe@example.org>

2024-04-21 07:37:07 +02:00

..

clip

Fix clippy lints + minor cleanups. (#1957)

2024-03-28 14:17:46 +01:00

segment_anything

Relax the contiguous check for cuda kernels. (#2000)

2024-04-03 09:02:38 +02:00

stable_diffusion

Improve the handling of matmul with squeezed layouts. (#1998)

2024-04-02 23:17:05 +02:00

whisper

Speaker embeddings computation for metavoice. (#1800)

2024-03-04 14:13:01 +01:00

wuerstchen

Do not implement Module for BatchNorm. (#1513)

2024-01-01 10:13:13 +01:00

bert.rs

Use candle_nn::embedding instead of local copies in a few models. (#1562)

2024-01-10 21:36:27 +01:00

bigcode.rs

Support for attention bias in gemma + refactor things a bit. (#1744)

2024-02-22 09:35:28 +01:00

blip_text.rs

[Wasm] BLIP Example (#1183)

2023-10-26 07:24:02 +01:00

blip.rs

[Wasm] BLIP Example (#1183)

2023-10-26 07:24:02 +01:00

chatglm.rs

Support for attention bias in gemma + refactor things a bit. (#1744)

2024-02-22 09:35:28 +01:00

convmixer.rs

Do not implement Module for BatchNorm. (#1513)

2024-01-01 10:13:13 +01:00

convnext.rs

Add ConvNeXt-V2 and smaller model variants. (#1709)

2024-02-14 10:53:07 +01:00

dinov2.rs

Another fix for squeezing. (#1943)

2024-03-26 17:05:26 +01:00

distilbert.rs

Distibert (#1366)

2023-11-24 15:09:14 +00:00

efficientnet.rs

Do not implement Module for BatchNorm. (#1513)

2024-01-01 10:13:13 +01:00

efficientvit.rs

EfficientVit (MSRA) model (#1783)

2024-03-01 08:53:52 +01:00

encodec.rs

Add a flag to select the dtype used in metavoice. (#1805)

2024-03-05 12:16:00 +01:00

falcon.rs

Make the falcon model cloneable. (#2067)

2024-04-15 09:39:03 +02:00

gemma.rs

Fix for gemma MQA. (#2091)

2024-04-19 21:49:55 +02:00

jina_bert.rs

Fixes for jina-bert. (#1189)

2023-10-26 18:52:30 +01:00

llama2_c_weights.rs

Move the llama2-c model in transformers. (#1205)

2023-10-28 16:51:19 +01:00

llama2_c.rs

Avoid the attention mask where possible. (#1933)

2024-03-25 15:31:04 +01:00

llama.rs

Small cleanups to the llama multi-process example. (#2098)

2024-04-20 22:19:46 +02:00

mamba.rs

Support alternative dtypes for mamba (#2036)

2024-04-10 18:10:01 +02:00

marian.rs

Remove the unused pragma for marian. (#1236)

2023-11-01 20:04:52 +00:00

metavoice.rs

Add some tracing to metavoice. (#1826)

2024-03-09 12:24:11 +01:00

mistral.rs

Use cat for faster MQA computation. (#2043)

2024-04-12 09:15:10 +02:00

mixformer.rs

Faster mask implementation for mixformers. (#2017)

2024-04-05 09:38:26 +02:00

mixtral.rs

Use cat for faster MQA computation. (#2043)

2024-04-12 09:15:10 +02:00

mobileone.rs

Add MobileOne model. (#1595)

2024-01-16 06:34:16 +01:00

mod.rs

Updated quantized phi model (#2099)

2024-04-21 07:37:07 +02:00

moondream.rs

Derive clone and debug traits for Moondream model (#2100)

2024-04-21 07:08:28 +02:00

mpt.rs

Use cat for faster MQA computation. (#2043)

2024-04-12 09:15:10 +02:00

persimmon.rs

Add the relu2 and relu6 activations. (#1201)

2023-10-27 20:51:16 +01:00

phi.rs

Use cat for faster MQA computation. (#2043)

2024-04-12 09:15:10 +02:00

quantized_blip_text.rs

Add a quantized blip model. (#1155)

2023-10-22 20:33:25 +01:00

quantized_blip.rs

[Wasm] BLIP Example (#1183)

2023-10-26 07:24:02 +01:00

quantized_llama2_c.rs

Avoid the attention mask where possible. (#1933)

2024-03-25 15:31:04 +01:00

quantized_llama.rs

Use cat for faster MQA computation. (#2043)

2024-04-12 09:15:10 +02:00

quantized_metavoice.rs

Fix the final rmsnorm for quantized-metavoice. (#2021)

2024-04-06 19:35:01 +02:00

quantized_mistral.rs

Use cat for faster MQA computation. (#2043)

2024-04-12 09:15:10 +02:00

quantized_mixformer.rs

Quantized moondream implementation and BOS token (#1980)

2024-04-01 19:37:54 +02:00

quantized_moondream.rs

Quantized moondream implementation and BOS token (#1980)

2024-04-01 19:37:54 +02:00

quantized_mpt.rs

Use cat for faster MQA computation. (#2043)

2024-04-12 09:15:10 +02:00

quantized_phi.rs

Updated quantized phi model (#2099)

2024-04-21 07:37:07 +02:00

quantized_recurrent_gemma.rs

Add a quantized version of recurrent-gemma. (#2054)

2024-04-13 20:07:01 +02:00

quantized_rwkv_v5.rs

add quantized rwkv v5 model (#1743)

2024-02-25 21:43:40 +01:00

quantized_rwkv_v6.rs

add models of rwkv v6 and quantized rwkv v6 (#1781)

2024-03-01 08:37:56 +01:00

quantized_stable_lm.rs

Use cat for faster MQA computation. (#2043)

2024-04-12 09:15:10 +02:00

quantized_t5.rs

Add support to UL2 model family (#1300)

2023-11-09 18:55:09 +01:00

qwen2_moe.rs

Use cat for faster MQA computation. (#2043)

2024-04-12 09:15:10 +02:00

qwen2.rs

Use cat for faster MQA computation. (#2043)

2024-04-12 09:15:10 +02:00

recurrent_gemma.rs

Add a quantized version of recurrent-gemma. (#2054)

2024-04-13 20:07:01 +02:00

repvgg.rs

Add RepVGG model. (#1561)

2024-01-11 07:07:40 +01:00

resnet.rs

Do not implement Module for BatchNorm. (#1513)

2024-01-01 10:13:13 +01:00

rwkv_v5.rs

add quantized rwkv v5 model (#1743)

2024-02-25 21:43:40 +01:00

rwkv_v6.rs

add models of rwkv v6 and quantized rwkv v6 (#1781)

2024-03-01 08:37:56 +01:00

segformer.rs

Add an initial Segformer implementation (#1617)

2024-03-03 16:01:46 +01:00

stable_lm.rs

Use cat for faster MQA computation. (#2043)

2024-04-12 09:15:10 +02:00

starcoder2.rs

Use cat for faster MQA computation. (#2043)

2024-04-12 09:15:10 +02:00

t5.rs

Avoid crashes when running T5 models with F16 tensors on CPU (#2047)

2024-04-13 11:07:28 +02:00

trocr.rs

Support sinusoidal embeddings in trocr. (#1690)

2024-02-10 15:17:51 +01:00

vgg.rs

Allow for different behavior between training and eval (#1213)

2023-10-29 07:53:09 +01:00

vit.rs

Use the repo config for trocr rather than hardcoding it + small tweaks. (#1689)

2024-02-10 13:15:03 +01:00

with_tracing.rs

Use a common with_tracing::RmsNorm in a few models. (#1871)

2024-03-18 21:40:06 +01:00

yi.rs

Use cat for faster MQA computation. (#2043)

2024-04-12 09:15:10 +02:00