jokerbit commited on
Commit
7a8b324
·
verified ·
1 Parent(s): abb78d2

Upload folder using huggingface_hub

Browse files
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ sample.png filter=lfs diff=lfs merge=lfs -text
.gitignore ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ **/.cache
2
+ **/__pycache__
3
+ **/*.egg-info
4
+ *.safetensors
5
+ **/.venv
6
+ .venv
7
+ .git
8
+ *.png
9
+ *.jpeg
README.md ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # flux-schnell-edge-inference
2
+
3
+ This holds the baseline for the FLUX Schnel NVIDIA GeForce RTX 4090 contest, which can be forked freely and optimized
4
+
5
+ Some recommendations are as follows:
6
+ - Installing dependencies should be done in `pyproject.toml`, including git dependencies
7
+ - HuggingFace models should be specified in the `models` array in the `pyproject.toml` file, and will be downloaded before benchmarking
8
+ - The pipeline does **not** have internet access so all dependencies and models must be included in the `pyproject.toml`
9
+ - Compiled models should be hosted on HuggingFace and included in the `models` array in the `pyproject.toml` (rather than compiling during loading). Loading time matters far more than file sizes
10
+ - Avoid changing `src/main.py`, as that includes mostly protocol logic. Most changes should be in `models` and `src/pipeline.py`
11
+ - Ensure the entire repository (excluding dependencies and HuggingFace models) is under 16MB
12
+
13
+ For testing, you need a docker container with pytorch and ubuntu 22.04.
14
+ You can download your listed dependencies with `uv`, installed with:
15
+ ```bash
16
+ pipx ensurepath
17
+ pipx install uv
18
+ ```
19
+ You can then relock with `uv lock`, and then run with `uv run start_inference`
pyproject.toml ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [build-system]
2
+ requires = ["setuptools >= 75.0"]
3
+ build-backend = "setuptools.build_meta"
4
+
5
+ [project]
6
+ name = "flux-schnell-edge-inference"
7
+ description = "An edge-maxxing model submission for the 4090 Flux contest"
8
+ requires-python = ">=3.10,<3.13"
9
+ version = "8"
10
+ dependencies = [
11
+ "diffusers==0.31.0",
12
+ "transformers==4.46.2",
13
+ "accelerate==1.1.0",
14
+ "omegaconf==2.3.0",
15
+ "torch==2.5.1",
16
+ "protobuf==5.28.3",
17
+ "sentencepiece==0.2.0",
18
+ "hf_transfer==0.1.8",
19
+ "edge-maxxing-pipelines @ git+https://github.com/womboai/edge-maxxing@7c760ac54f6052803dadb3ade8ebfc9679a94589#subdirectory=pipelines",
20
+ "torchao>=0.6.1",
21
+ "ipython>=8.29.0",
22
+ "setuptools >= 75.0",
23
+ "optimum-quanto>=0.2.6",
24
+ ]
25
+
26
+ [[tool.edge-maxxing.models]]
27
+ repository = "jokerbit/flux.1-schnell-Robert-int8wo"
28
+ revision = "5ef0012f11a863e5111ec56540302a023bc8587b"
29
+
30
+ [[tool.edge-maxxing.models]]
31
+ repository = "madebyollin/taef1"
32
+ revision = "2d552378e58c9c94201075708d7de4e1163b2689"
33
+
34
+ [project.scripts]
35
+ start_inference = "main:main"
src/.pipeline.py.swp ADDED
Binary file (16.4 kB). View file
 
src/main.py ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from io import BytesIO
2
+ from multiprocessing.connection import Listener
3
+ from os import chmod, remove
4
+ from os.path import abspath, exists
5
+ from pathlib import Path
6
+
7
+ from PIL.JpegImagePlugin import JpegImageFile
8
+ from pipelines.models import TextToImageRequest
9
+
10
+ from pipeline import load_pipeline, infer
11
+
12
+ SOCKET = abspath(Path(__file__).parent.parent / "inferences.sock")
13
+
14
+
15
+ def main():
16
+ print(f"Loading pipeline")
17
+ pipeline = load_pipeline()
18
+
19
+ print(f"Pipeline loaded, creating socket at '{SOCKET}'")
20
+
21
+ if exists(SOCKET):
22
+ remove(SOCKET)
23
+
24
+ with Listener(SOCKET) as listener:
25
+ chmod(SOCKET, 0o777)
26
+
27
+ print(f"Awaiting connections")
28
+ with listener.accept() as connection:
29
+ print(f"Connected")
30
+
31
+ while True:
32
+ try:
33
+ request = TextToImageRequest.model_validate_json(connection.recv_bytes().decode("utf-8"))
34
+ except EOFError:
35
+ print(f"Inference socket exiting")
36
+
37
+ return
38
+
39
+ image = infer(request, pipeline)
40
+
41
+ data = BytesIO()
42
+ image.save(data, format=JpegImageFile.format)
43
+
44
+ packet = data.getvalue()
45
+
46
+ connection.send_bytes(packet)
47
+
48
+
49
+ if __name__ == '__main__':
50
+ main()
src/pipeline.py ADDED
@@ -0,0 +1,108 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import gc
2
+ import os
3
+ from typing import TypeAlias
4
+
5
+ import torch
6
+ from PIL.Image import Image
7
+ from diffusers import FluxPipeline, FluxTransformer2DModel, AutoencoderKL, AutoencoderTiny
8
+ from huggingface_hub.constants import HF_HUB_CACHE
9
+ from pipelines.models import TextToImageRequest
10
+ from torch import Generator
11
+ from torchao.quantization import quantize_, int8_weight_only
12
+ from transformers import T5EncoderModel, CLIPTextModel
13
+
14
+ torch.backends.cudnn.benchmark = True
15
+ torch.backends.cuda.matmul.allow_tf32 = True
16
+ Pipeline: TypeAlias = FluxPipeline
17
+
18
+ CHECKPOINT = "jokerbit/flux.1-schnell-Robert-int8wo"
19
+ REVISION = "5ef0012f11a863e5111ec56540302a023bc8587b"
20
+
21
+ TinyVAE = "madebyollin/taef1"
22
+ TinyVAE_REV = "2d552378e58c9c94201075708d7de4e1163b2689"
23
+
24
+
25
+ def load_pipeline() -> Pipeline:
26
+ text_encoder = CLIPTextModel.from_pretrained(
27
+ CHECKPOINT,
28
+ revision=REVISION,
29
+ subfolder="text_encoder",
30
+ local_files_only=True,
31
+ torch_dtype=torch.bfloat16,
32
+ )
33
+
34
+ text_encoder_2 = T5EncoderModel.from_pretrained(
35
+ CHECKPOINT,
36
+ revision=REVISION,
37
+ subfolder="text_encoder_2",
38
+ local_files_only=True,
39
+ torch_dtype=torch.bfloat16,
40
+ )
41
+
42
+ vae = AutoencoderTiny.from_pretrained(
43
+ TinyVAE,
44
+ revision=TinyVAE_REV,
45
+ local_files_only=True,
46
+ torch_dtype=torch.bfloat16,
47
+ )
48
+
49
+ path = os.path.join(HF_HUB_CACHE, "models--jokerbit--flux.1-schnell-Robert-int8wo/snapshots/5ef0012f11a863e5111ec56540302a023bc8587b/transformer")
50
+
51
+ transformer = FluxTransformer2DModel.from_pretrained(
52
+ path,
53
+ torch_dtype=torch.bfloat16,
54
+ use_safetensors=False,
55
+ )
56
+
57
+ pipeline = FluxPipeline.from_pretrained(
58
+ CHECKPOINT,
59
+ revision=REVISION,
60
+ local_files_only=True,
61
+ text_encoder=text_encoder,
62
+ text_encoder_2=text_encoder_2,
63
+ transformer=transformer,
64
+ vae=vae,
65
+ torch_dtype=torch.bfloat16,
66
+ ).to("cuda")
67
+ pipeline.text_encoder_2.to(memory_format=torch.channels_last)
68
+ pipeline.transformer.to(memory_format=torch.channels_last)
69
+ pipeline.vae.to(memory_format=torch.channels_last)
70
+ for _ in range(2):
71
+ pipeline("cat", num_inference_steps=4)
72
+
73
+ return pipeline
74
+
75
+ def infer(request: TextToImageRequest, pipeline: Pipeline) -> Image:
76
+ gc.collect()
77
+ torch.cuda.empty_cache()
78
+ torch.cuda.reset_peak_memory_stats()
79
+
80
+ generator = Generator(pipeline.device).manual_seed(request.seed)
81
+
82
+ return pipeline(
83
+ request.prompt,
84
+ generator=generator,
85
+ guidance_scale=0.0,
86
+ num_inference_steps=4,
87
+ max_sequence_length=256,
88
+ height=request.height,
89
+ width=request.width,
90
+ ).images[0]
91
+
92
+
93
+ if __name__ == "__main__":
94
+ from time import perf_counter
95
+ PROMPT = 'martyr, semiconformity, peregrination, quip, twineless, emotionless, tawa, depickle'
96
+ request = TextToImageRequest(prompt=PROMPT,
97
+ height=None,
98
+ width=None,
99
+ seed=666)
100
+ start_time = perf_counter()
101
+ pipe_ = load_pipeline()
102
+ stop_time = perf_counter()
103
+ print(f"Pipeline is loaded in {stop_time - start_time}s")
104
+ for _ in range(4):
105
+ start_time = perf_counter()
106
+ infer(request, pipe_)
107
+ stop_time = perf_counter()
108
+ print(f"Request in {stop_time - start_time}s")
t5_xxl/.gitattributes ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ *.7z filter=lfs diff=lfs merge=lfs -text
2
+ *.arrow filter=lfs diff=lfs merge=lfs -text
3
+ *.bin filter=lfs diff=lfs merge=lfs -text
4
+ *.bz2 filter=lfs diff=lfs merge=lfs -text
5
+ *.ckpt filter=lfs diff=lfs merge=lfs -text
6
+ *.ftz filter=lfs diff=lfs merge=lfs -text
7
+ *.gz filter=lfs diff=lfs merge=lfs -text
8
+ *.h5 filter=lfs diff=lfs merge=lfs -text
9
+ *.joblib filter=lfs diff=lfs merge=lfs -text
10
+ *.lfs.* filter=lfs diff=lfs merge=lfs -text
11
+ *.mlmodel filter=lfs diff=lfs merge=lfs -text
12
+ *.model filter=lfs diff=lfs merge=lfs -text
13
+ *.msgpack filter=lfs diff=lfs merge=lfs -text
14
+ *.npy filter=lfs diff=lfs merge=lfs -text
15
+ *.npz filter=lfs diff=lfs merge=lfs -text
16
+ *.onnx filter=lfs diff=lfs merge=lfs -text
17
+ *.ot filter=lfs diff=lfs merge=lfs -text
18
+ *.parquet filter=lfs diff=lfs merge=lfs -text
19
+ *.pb filter=lfs diff=lfs merge=lfs -text
20
+ *.pickle filter=lfs diff=lfs merge=lfs -text
21
+ *.pkl filter=lfs diff=lfs merge=lfs -text
22
+ *.pt filter=lfs diff=lfs merge=lfs -text
23
+ *.pth filter=lfs diff=lfs merge=lfs -text
24
+ *.rar filter=lfs diff=lfs merge=lfs -text
25
+ *.safetensors filter=lfs diff=lfs merge=lfs -text
26
+ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
27
+ *.tar.* filter=lfs diff=lfs merge=lfs -text
28
+ *.tar filter=lfs diff=lfs merge=lfs -text
29
+ *.tflite filter=lfs diff=lfs merge=lfs -text
30
+ *.tgz filter=lfs diff=lfs merge=lfs -text
31
+ *.wasm filter=lfs diff=lfs merge=lfs -text
32
+ *.xz filter=lfs diff=lfs merge=lfs -text
33
+ *.zip filter=lfs diff=lfs merge=lfs -text
34
+ *.zst filter=lfs diff=lfs merge=lfs -text
35
+ *tfevents* filter=lfs diff=lfs merge=lfs -text
t5_xxl/README.md ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ ---
4
+
5
+ Quantized T5-XXL of FLUX.1[schnell] using HuggingFace [optimum-quanto](https://github.com/huggingface/optimum-quanto).
6
+
7
+ ### Quantize
8
+
9
+ ```py
10
+ import torch
11
+ from transformers import T5EncoderModel
12
+ from optimum.quanto import (
13
+ QuantizedTransformersModel,
14
+ qfloat8_e4m3fn,
15
+ qfloat8_e5m2,
16
+ qint8,
17
+ qint4,
18
+ )
19
+
20
+ REPO_NAME = "black-forest-labs/FLUX.1-schnell"
21
+ TEXT_ENCODER = "text_encoder_2"
22
+
23
+ model = T5EncoderModel.from_pretrained(
24
+ REPO_NAME, subfolder=TEXT_ENCODER, torch_dtype=torch.bfloat16
25
+ )
26
+ qmodel = QuantizedTransformersModel.quantize(
27
+ model,
28
+ weights=qfloat8_e4m3fn,
29
+ )
30
+ qmodel.save_pretrained("./t5_xxl/qfloat8_e4m3fn")
31
+ ```
32
+
33
+ ### Load
34
+
35
+ Currently `QuantizedTransformersModel` [does not support](https://github.com/huggingface/optimum-quanto/blob/601dc193ce0ed381c479fde54a81ba546bdf64d1/optimum/quanto/models/transformers_models.py#L151) load a quantized model from huggingface hub.
36
+
37
+ ```py
38
+ from transformers import T5EncoderModel, AutoModelForTextEncoding
39
+ from optimum.quanto import QuantizedTransformersModel
40
+
41
+ MODEL_PATH = "./t5_xxl/qfloat8_e4m3fn"
42
+
43
+ class QuantizedModelForTextEncoding(QuantizedTransformersModel):
44
+ auto_class = AutoModelForTextEncoding
45
+
46
+ qmodel = QuantizedModelForTextEncoding.from_pretrained(
47
+ "./t5_xxl/qint8",
48
+ )
49
+ ```
t5_xxl/qfloat8_e4m3fn/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "black-forest-labs/FLUX.1-schnell",
3
+ "architectures": [
4
+ "T5EncoderModel"
5
+ ],
6
+ "classifier_dropout": 0.0,
7
+ "d_ff": 10240,
8
+ "d_kv": 64,
9
+ "d_model": 4096,
10
+ "decoder_start_token_id": 0,
11
+ "dense_act_fn": "gelu_new",
12
+ "dropout_rate": 0.1,
13
+ "eos_token_id": 1,
14
+ "feed_forward_proj": "gated-gelu",
15
+ "initializer_factor": 1.0,
16
+ "is_encoder_decoder": true,
17
+ "is_gated_act": true,
18
+ "layer_norm_epsilon": 1e-06,
19
+ "model_type": "t5",
20
+ "num_decoder_layers": 24,
21
+ "num_heads": 64,
22
+ "num_layers": 24,
23
+ "output_past": true,
24
+ "pad_token_id": 0,
25
+ "relative_attention_max_distance": 128,
26
+ "relative_attention_num_buckets": 32,
27
+ "tie_word_embeddings": false,
28
+ "torch_dtype": "bfloat16",
29
+ "transformers_version": "4.43.4",
30
+ "use_cache": true,
31
+ "vocab_size": 32128
32
+ }
t5_xxl/qfloat8_e4m3fn/quanto_qmap.json ADDED
@@ -0,0 +1,674 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "encoder.block.0.layer.0.SelfAttention.q": {
3
+ "weights": "qfloat8_e4m3fn",
4
+ "activations": "none"
5
+ },
6
+ "encoder.block.0.layer.0.SelfAttention.k": {
7
+ "weights": "qfloat8_e4m3fn",
8
+ "activations": "none"
9
+ },
10
+ "encoder.block.0.layer.0.SelfAttention.v": {
11
+ "weights": "qfloat8_e4m3fn",
12
+ "activations": "none"
13
+ },
14
+ "encoder.block.0.layer.0.SelfAttention.o": {
15
+ "weights": "qfloat8_e4m3fn",
16
+ "activations": "none"
17
+ },
18
+ "encoder.block.0.layer.1.DenseReluDense.wi_0": {
19
+ "weights": "qfloat8_e4m3fn",
20
+ "activations": "none"
21
+ },
22
+ "encoder.block.0.layer.1.DenseReluDense.wi_1": {
23
+ "weights": "qfloat8_e4m3fn",
24
+ "activations": "none"
25
+ },
26
+ "encoder.block.0.layer.1.DenseReluDense.wo": {
27
+ "weights": "qfloat8_e4m3fn",
28
+ "activations": "none"
29
+ },
30
+ "encoder.block.1.layer.0.SelfAttention.q": {
31
+ "weights": "qfloat8_e4m3fn",
32
+ "activations": "none"
33
+ },
34
+ "encoder.block.1.layer.0.SelfAttention.k": {
35
+ "weights": "qfloat8_e4m3fn",
36
+ "activations": "none"
37
+ },
38
+ "encoder.block.1.layer.0.SelfAttention.v": {
39
+ "weights": "qfloat8_e4m3fn",
40
+ "activations": "none"
41
+ },
42
+ "encoder.block.1.layer.0.SelfAttention.o": {
43
+ "weights": "qfloat8_e4m3fn",
44
+ "activations": "none"
45
+ },
46
+ "encoder.block.1.layer.1.DenseReluDense.wi_0": {
47
+ "weights": "qfloat8_e4m3fn",
48
+ "activations": "none"
49
+ },
50
+ "encoder.block.1.layer.1.DenseReluDense.wi_1": {
51
+ "weights": "qfloat8_e4m3fn",
52
+ "activations": "none"
53
+ },
54
+ "encoder.block.1.layer.1.DenseReluDense.wo": {
55
+ "weights": "qfloat8_e4m3fn",
56
+ "activations": "none"
57
+ },
58
+ "encoder.block.2.layer.0.SelfAttention.q": {
59
+ "weights": "qfloat8_e4m3fn",
60
+ "activations": "none"
61
+ },
62
+ "encoder.block.2.layer.0.SelfAttention.k": {
63
+ "weights": "qfloat8_e4m3fn",
64
+ "activations": "none"
65
+ },
66
+ "encoder.block.2.layer.0.SelfAttention.v": {
67
+ "weights": "qfloat8_e4m3fn",
68
+ "activations": "none"
69
+ },
70
+ "encoder.block.2.layer.0.SelfAttention.o": {
71
+ "weights": "qfloat8_e4m3fn",
72
+ "activations": "none"
73
+ },
74
+ "encoder.block.2.layer.1.DenseReluDense.wi_0": {
75
+ "weights": "qfloat8_e4m3fn",
76
+ "activations": "none"
77
+ },
78
+ "encoder.block.2.layer.1.DenseReluDense.wi_1": {
79
+ "weights": "qfloat8_e4m3fn",
80
+ "activations": "none"
81
+ },
82
+ "encoder.block.2.layer.1.DenseReluDense.wo": {
83
+ "weights": "qfloat8_e4m3fn",
84
+ "activations": "none"
85
+ },
86
+ "encoder.block.3.layer.0.SelfAttention.q": {
87
+ "weights": "qfloat8_e4m3fn",
88
+ "activations": "none"
89
+ },
90
+ "encoder.block.3.layer.0.SelfAttention.k": {
91
+ "weights": "qfloat8_e4m3fn",
92
+ "activations": "none"
93
+ },
94
+ "encoder.block.3.layer.0.SelfAttention.v": {
95
+ "weights": "qfloat8_e4m3fn",
96
+ "activations": "none"
97
+ },
98
+ "encoder.block.3.layer.0.SelfAttention.o": {
99
+ "weights": "qfloat8_e4m3fn",
100
+ "activations": "none"
101
+ },
102
+ "encoder.block.3.layer.1.DenseReluDense.wi_0": {
103
+ "weights": "qfloat8_e4m3fn",
104
+ "activations": "none"
105
+ },
106
+ "encoder.block.3.layer.1.DenseReluDense.wi_1": {
107
+ "weights": "qfloat8_e4m3fn",
108
+ "activations": "none"
109
+ },
110
+ "encoder.block.3.layer.1.DenseReluDense.wo": {
111
+ "weights": "qfloat8_e4m3fn",
112
+ "activations": "none"
113
+ },
114
+ "encoder.block.4.layer.0.SelfAttention.q": {
115
+ "weights": "qfloat8_e4m3fn",
116
+ "activations": "none"
117
+ },
118
+ "encoder.block.4.layer.0.SelfAttention.k": {
119
+ "weights": "qfloat8_e4m3fn",
120
+ "activations": "none"
121
+ },
122
+ "encoder.block.4.layer.0.SelfAttention.v": {
123
+ "weights": "qfloat8_e4m3fn",
124
+ "activations": "none"
125
+ },
126
+ "encoder.block.4.layer.0.SelfAttention.o": {
127
+ "weights": "qfloat8_e4m3fn",
128
+ "activations": "none"
129
+ },
130
+ "encoder.block.4.layer.1.DenseReluDense.wi_0": {
131
+ "weights": "qfloat8_e4m3fn",
132
+ "activations": "none"
133
+ },
134
+ "encoder.block.4.layer.1.DenseReluDense.wi_1": {
135
+ "weights": "qfloat8_e4m3fn",
136
+ "activations": "none"
137
+ },
138
+ "encoder.block.4.layer.1.DenseReluDense.wo": {
139
+ "weights": "qfloat8_e4m3fn",
140
+ "activations": "none"
141
+ },
142
+ "encoder.block.5.layer.0.SelfAttention.q": {
143
+ "weights": "qfloat8_e4m3fn",
144
+ "activations": "none"
145
+ },
146
+ "encoder.block.5.layer.0.SelfAttention.k": {
147
+ "weights": "qfloat8_e4m3fn",
148
+ "activations": "none"
149
+ },
150
+ "encoder.block.5.layer.0.SelfAttention.v": {
151
+ "weights": "qfloat8_e4m3fn",
152
+ "activations": "none"
153
+ },
154
+ "encoder.block.5.layer.0.SelfAttention.o": {
155
+ "weights": "qfloat8_e4m3fn",
156
+ "activations": "none"
157
+ },
158
+ "encoder.block.5.layer.1.DenseReluDense.wi_0": {
159
+ "weights": "qfloat8_e4m3fn",
160
+ "activations": "none"
161
+ },
162
+ "encoder.block.5.layer.1.DenseReluDense.wi_1": {
163
+ "weights": "qfloat8_e4m3fn",
164
+ "activations": "none"
165
+ },
166
+ "encoder.block.5.layer.1.DenseReluDense.wo": {
167
+ "weights": "qfloat8_e4m3fn",
168
+ "activations": "none"
169
+ },
170
+ "encoder.block.6.layer.0.SelfAttention.q": {
171
+ "weights": "qfloat8_e4m3fn",
172
+ "activations": "none"
173
+ },
174
+ "encoder.block.6.layer.0.SelfAttention.k": {
175
+ "weights": "qfloat8_e4m3fn",
176
+ "activations": "none"
177
+ },
178
+ "encoder.block.6.layer.0.SelfAttention.v": {
179
+ "weights": "qfloat8_e4m3fn",
180
+ "activations": "none"
181
+ },
182
+ "encoder.block.6.layer.0.SelfAttention.o": {
183
+ "weights": "qfloat8_e4m3fn",
184
+ "activations": "none"
185
+ },
186
+ "encoder.block.6.layer.1.DenseReluDense.wi_0": {
187
+ "weights": "qfloat8_e4m3fn",
188
+ "activations": "none"
189
+ },
190
+ "encoder.block.6.layer.1.DenseReluDense.wi_1": {
191
+ "weights": "qfloat8_e4m3fn",
192
+ "activations": "none"
193
+ },
194
+ "encoder.block.6.layer.1.DenseReluDense.wo": {
195
+ "weights": "qfloat8_e4m3fn",
196
+ "activations": "none"
197
+ },
198
+ "encoder.block.7.layer.0.SelfAttention.q": {
199
+ "weights": "qfloat8_e4m3fn",
200
+ "activations": "none"
201
+ },
202
+ "encoder.block.7.layer.0.SelfAttention.k": {
203
+ "weights": "qfloat8_e4m3fn",
204
+ "activations": "none"
205
+ },
206
+ "encoder.block.7.layer.0.SelfAttention.v": {
207
+ "weights": "qfloat8_e4m3fn",
208
+ "activations": "none"
209
+ },
210
+ "encoder.block.7.layer.0.SelfAttention.o": {
211
+ "weights": "qfloat8_e4m3fn",
212
+ "activations": "none"
213
+ },
214
+ "encoder.block.7.layer.1.DenseReluDense.wi_0": {
215
+ "weights": "qfloat8_e4m3fn",
216
+ "activations": "none"
217
+ },
218
+ "encoder.block.7.layer.1.DenseReluDense.wi_1": {
219
+ "weights": "qfloat8_e4m3fn",
220
+ "activations": "none"
221
+ },
222
+ "encoder.block.7.layer.1.DenseReluDense.wo": {
223
+ "weights": "qfloat8_e4m3fn",
224
+ "activations": "none"
225
+ },
226
+ "encoder.block.8.layer.0.SelfAttention.q": {
227
+ "weights": "qfloat8_e4m3fn",
228
+ "activations": "none"
229
+ },
230
+ "encoder.block.8.layer.0.SelfAttention.k": {
231
+ "weights": "qfloat8_e4m3fn",
232
+ "activations": "none"
233
+ },
234
+ "encoder.block.8.layer.0.SelfAttention.v": {
235
+ "weights": "qfloat8_e4m3fn",
236
+ "activations": "none"
237
+ },
238
+ "encoder.block.8.layer.0.SelfAttention.o": {
239
+ "weights": "qfloat8_e4m3fn",
240
+ "activations": "none"
241
+ },
242
+ "encoder.block.8.layer.1.DenseReluDense.wi_0": {
243
+ "weights": "qfloat8_e4m3fn",
244
+ "activations": "none"
245
+ },
246
+ "encoder.block.8.layer.1.DenseReluDense.wi_1": {
247
+ "weights": "qfloat8_e4m3fn",
248
+ "activations": "none"
249
+ },
250
+ "encoder.block.8.layer.1.DenseReluDense.wo": {
251
+ "weights": "qfloat8_e4m3fn",
252
+ "activations": "none"
253
+ },
254
+ "encoder.block.9.layer.0.SelfAttention.q": {
255
+ "weights": "qfloat8_e4m3fn",
256
+ "activations": "none"
257
+ },
258
+ "encoder.block.9.layer.0.SelfAttention.k": {
259
+ "weights": "qfloat8_e4m3fn",
260
+ "activations": "none"
261
+ },
262
+ "encoder.block.9.layer.0.SelfAttention.v": {
263
+ "weights": "qfloat8_e4m3fn",
264
+ "activations": "none"
265
+ },
266
+ "encoder.block.9.layer.0.SelfAttention.o": {
267
+ "weights": "qfloat8_e4m3fn",
268
+ "activations": "none"
269
+ },
270
+ "encoder.block.9.layer.1.DenseReluDense.wi_0": {
271
+ "weights": "qfloat8_e4m3fn",
272
+ "activations": "none"
273
+ },
274
+ "encoder.block.9.layer.1.DenseReluDense.wi_1": {
275
+ "weights": "qfloat8_e4m3fn",
276
+ "activations": "none"
277
+ },
278
+ "encoder.block.9.layer.1.DenseReluDense.wo": {
279
+ "weights": "qfloat8_e4m3fn",
280
+ "activations": "none"
281
+ },
282
+ "encoder.block.10.layer.0.SelfAttention.q": {
283
+ "weights": "qfloat8_e4m3fn",
284
+ "activations": "none"
285
+ },
286
+ "encoder.block.10.layer.0.SelfAttention.k": {
287
+ "weights": "qfloat8_e4m3fn",
288
+ "activations": "none"
289
+ },
290
+ "encoder.block.10.layer.0.SelfAttention.v": {
291
+ "weights": "qfloat8_e4m3fn",
292
+ "activations": "none"
293
+ },
294
+ "encoder.block.10.layer.0.SelfAttention.o": {
295
+ "weights": "qfloat8_e4m3fn",
296
+ "activations": "none"
297
+ },
298
+ "encoder.block.10.layer.1.DenseReluDense.wi_0": {
299
+ "weights": "qfloat8_e4m3fn",
300
+ "activations": "none"
301
+ },
302
+ "encoder.block.10.layer.1.DenseReluDense.wi_1": {
303
+ "weights": "qfloat8_e4m3fn",
304
+ "activations": "none"
305
+ },
306
+ "encoder.block.10.layer.1.DenseReluDense.wo": {
307
+ "weights": "qfloat8_e4m3fn",
308
+ "activations": "none"
309
+ },
310
+ "encoder.block.11.layer.0.SelfAttention.q": {
311
+ "weights": "qfloat8_e4m3fn",
312
+ "activations": "none"
313
+ },
314
+ "encoder.block.11.layer.0.SelfAttention.k": {
315
+ "weights": "qfloat8_e4m3fn",
316
+ "activations": "none"
317
+ },
318
+ "encoder.block.11.layer.0.SelfAttention.v": {
319
+ "weights": "qfloat8_e4m3fn",
320
+ "activations": "none"
321
+ },
322
+ "encoder.block.11.layer.0.SelfAttention.o": {
323
+ "weights": "qfloat8_e4m3fn",
324
+ "activations": "none"
325
+ },
326
+ "encoder.block.11.layer.1.DenseReluDense.wi_0": {
327
+ "weights": "qfloat8_e4m3fn",
328
+ "activations": "none"
329
+ },
330
+ "encoder.block.11.layer.1.DenseReluDense.wi_1": {
331
+ "weights": "qfloat8_e4m3fn",
332
+ "activations": "none"
333
+ },
334
+ "encoder.block.11.layer.1.DenseReluDense.wo": {
335
+ "weights": "qfloat8_e4m3fn",
336
+ "activations": "none"
337
+ },
338
+ "encoder.block.12.layer.0.SelfAttention.q": {
339
+ "weights": "qfloat8_e4m3fn",
340
+ "activations": "none"
341
+ },
342
+ "encoder.block.12.layer.0.SelfAttention.k": {
343
+ "weights": "qfloat8_e4m3fn",
344
+ "activations": "none"
345
+ },
346
+ "encoder.block.12.layer.0.SelfAttention.v": {
347
+ "weights": "qfloat8_e4m3fn",
348
+ "activations": "none"
349
+ },
350
+ "encoder.block.12.layer.0.SelfAttention.o": {
351
+ "weights": "qfloat8_e4m3fn",
352
+ "activations": "none"
353
+ },
354
+ "encoder.block.12.layer.1.DenseReluDense.wi_0": {
355
+ "weights": "qfloat8_e4m3fn",
356
+ "activations": "none"
357
+ },
358
+ "encoder.block.12.layer.1.DenseReluDense.wi_1": {
359
+ "weights": "qfloat8_e4m3fn",
360
+ "activations": "none"
361
+ },
362
+ "encoder.block.12.layer.1.DenseReluDense.wo": {
363
+ "weights": "qfloat8_e4m3fn",
364
+ "activations": "none"
365
+ },
366
+ "encoder.block.13.layer.0.SelfAttention.q": {
367
+ "weights": "qfloat8_e4m3fn",
368
+ "activations": "none"
369
+ },
370
+ "encoder.block.13.layer.0.SelfAttention.k": {
371
+ "weights": "qfloat8_e4m3fn",
372
+ "activations": "none"
373
+ },
374
+ "encoder.block.13.layer.0.SelfAttention.v": {
375
+ "weights": "qfloat8_e4m3fn",
376
+ "activations": "none"
377
+ },
378
+ "encoder.block.13.layer.0.SelfAttention.o": {
379
+ "weights": "qfloat8_e4m3fn",
380
+ "activations": "none"
381
+ },
382
+ "encoder.block.13.layer.1.DenseReluDense.wi_0": {
383
+ "weights": "qfloat8_e4m3fn",
384
+ "activations": "none"
385
+ },
386
+ "encoder.block.13.layer.1.DenseReluDense.wi_1": {
387
+ "weights": "qfloat8_e4m3fn",
388
+ "activations": "none"
389
+ },
390
+ "encoder.block.13.layer.1.DenseReluDense.wo": {
391
+ "weights": "qfloat8_e4m3fn",
392
+ "activations": "none"
393
+ },
394
+ "encoder.block.14.layer.0.SelfAttention.q": {
395
+ "weights": "qfloat8_e4m3fn",
396
+ "activations": "none"
397
+ },
398
+ "encoder.block.14.layer.0.SelfAttention.k": {
399
+ "weights": "qfloat8_e4m3fn",
400
+ "activations": "none"
401
+ },
402
+ "encoder.block.14.layer.0.SelfAttention.v": {
403
+ "weights": "qfloat8_e4m3fn",
404
+ "activations": "none"
405
+ },
406
+ "encoder.block.14.layer.0.SelfAttention.o": {
407
+ "weights": "qfloat8_e4m3fn",
408
+ "activations": "none"
409
+ },
410
+ "encoder.block.14.layer.1.DenseReluDense.wi_0": {
411
+ "weights": "qfloat8_e4m3fn",
412
+ "activations": "none"
413
+ },
414
+ "encoder.block.14.layer.1.DenseReluDense.wi_1": {
415
+ "weights": "qfloat8_e4m3fn",
416
+ "activations": "none"
417
+ },
418
+ "encoder.block.14.layer.1.DenseReluDense.wo": {
419
+ "weights": "qfloat8_e4m3fn",
420
+ "activations": "none"
421
+ },
422
+ "encoder.block.15.layer.0.SelfAttention.q": {
423
+ "weights": "qfloat8_e4m3fn",
424
+ "activations": "none"
425
+ },
426
+ "encoder.block.15.layer.0.SelfAttention.k": {
427
+ "weights": "qfloat8_e4m3fn",
428
+ "activations": "none"
429
+ },
430
+ "encoder.block.15.layer.0.SelfAttention.v": {
431
+ "weights": "qfloat8_e4m3fn",
432
+ "activations": "none"
433
+ },
434
+ "encoder.block.15.layer.0.SelfAttention.o": {
435
+ "weights": "qfloat8_e4m3fn",
436
+ "activations": "none"
437
+ },
438
+ "encoder.block.15.layer.1.DenseReluDense.wi_0": {
439
+ "weights": "qfloat8_e4m3fn",
440
+ "activations": "none"
441
+ },
442
+ "encoder.block.15.layer.1.DenseReluDense.wi_1": {
443
+ "weights": "qfloat8_e4m3fn",
444
+ "activations": "none"
445
+ },
446
+ "encoder.block.15.layer.1.DenseReluDense.wo": {
447
+ "weights": "qfloat8_e4m3fn",
448
+ "activations": "none"
449
+ },
450
+ "encoder.block.16.layer.0.SelfAttention.q": {
451
+ "weights": "qfloat8_e4m3fn",
452
+ "activations": "none"
453
+ },
454
+ "encoder.block.16.layer.0.SelfAttention.k": {
455
+ "weights": "qfloat8_e4m3fn",
456
+ "activations": "none"
457
+ },
458
+ "encoder.block.16.layer.0.SelfAttention.v": {
459
+ "weights": "qfloat8_e4m3fn",
460
+ "activations": "none"
461
+ },
462
+ "encoder.block.16.layer.0.SelfAttention.o": {
463
+ "weights": "qfloat8_e4m3fn",
464
+ "activations": "none"
465
+ },
466
+ "encoder.block.16.layer.1.DenseReluDense.wi_0": {
467
+ "weights": "qfloat8_e4m3fn",
468
+ "activations": "none"
469
+ },
470
+ "encoder.block.16.layer.1.DenseReluDense.wi_1": {
471
+ "weights": "qfloat8_e4m3fn",
472
+ "activations": "none"
473
+ },
474
+ "encoder.block.16.layer.1.DenseReluDense.wo": {
475
+ "weights": "qfloat8_e4m3fn",
476
+ "activations": "none"
477
+ },
478
+ "encoder.block.17.layer.0.SelfAttention.q": {
479
+ "weights": "qfloat8_e4m3fn",
480
+ "activations": "none"
481
+ },
482
+ "encoder.block.17.layer.0.SelfAttention.k": {
483
+ "weights": "qfloat8_e4m3fn",
484
+ "activations": "none"
485
+ },
486
+ "encoder.block.17.layer.0.SelfAttention.v": {
487
+ "weights": "qfloat8_e4m3fn",
488
+ "activations": "none"
489
+ },
490
+ "encoder.block.17.layer.0.SelfAttention.o": {
491
+ "weights": "qfloat8_e4m3fn",
492
+ "activations": "none"
493
+ },
494
+ "encoder.block.17.layer.1.DenseReluDense.wi_0": {
495
+ "weights": "qfloat8_e4m3fn",
496
+ "activations": "none"
497
+ },
498
+ "encoder.block.17.layer.1.DenseReluDense.wi_1": {
499
+ "weights": "qfloat8_e4m3fn",
500
+ "activations": "none"
501
+ },
502
+ "encoder.block.17.layer.1.DenseReluDense.wo": {
503
+ "weights": "qfloat8_e4m3fn",
504
+ "activations": "none"
505
+ },
506
+ "encoder.block.18.layer.0.SelfAttention.q": {
507
+ "weights": "qfloat8_e4m3fn",
508
+ "activations": "none"
509
+ },
510
+ "encoder.block.18.layer.0.SelfAttention.k": {
511
+ "weights": "qfloat8_e4m3fn",
512
+ "activations": "none"
513
+ },
514
+ "encoder.block.18.layer.0.SelfAttention.v": {
515
+ "weights": "qfloat8_e4m3fn",
516
+ "activations": "none"
517
+ },
518
+ "encoder.block.18.layer.0.SelfAttention.o": {
519
+ "weights": "qfloat8_e4m3fn",
520
+ "activations": "none"
521
+ },
522
+ "encoder.block.18.layer.1.DenseReluDense.wi_0": {
523
+ "weights": "qfloat8_e4m3fn",
524
+ "activations": "none"
525
+ },
526
+ "encoder.block.18.layer.1.DenseReluDense.wi_1": {
527
+ "weights": "qfloat8_e4m3fn",
528
+ "activations": "none"
529
+ },
530
+ "encoder.block.18.layer.1.DenseReluDense.wo": {
531
+ "weights": "qfloat8_e4m3fn",
532
+ "activations": "none"
533
+ },
534
+ "encoder.block.19.layer.0.SelfAttention.q": {
535
+ "weights": "qfloat8_e4m3fn",
536
+ "activations": "none"
537
+ },
538
+ "encoder.block.19.layer.0.SelfAttention.k": {
539
+ "weights": "qfloat8_e4m3fn",
540
+ "activations": "none"
541
+ },
542
+ "encoder.block.19.layer.0.SelfAttention.v": {
543
+ "weights": "qfloat8_e4m3fn",
544
+ "activations": "none"
545
+ },
546
+ "encoder.block.19.layer.0.SelfAttention.o": {
547
+ "weights": "qfloat8_e4m3fn",
548
+ "activations": "none"
549
+ },
550
+ "encoder.block.19.layer.1.DenseReluDense.wi_0": {
551
+ "weights": "qfloat8_e4m3fn",
552
+ "activations": "none"
553
+ },
554
+ "encoder.block.19.layer.1.DenseReluDense.wi_1": {
555
+ "weights": "qfloat8_e4m3fn",
556
+ "activations": "none"
557
+ },
558
+ "encoder.block.19.layer.1.DenseReluDense.wo": {
559
+ "weights": "qfloat8_e4m3fn",
560
+ "activations": "none"
561
+ },
562
+ "encoder.block.20.layer.0.SelfAttention.q": {
563
+ "weights": "qfloat8_e4m3fn",
564
+ "activations": "none"
565
+ },
566
+ "encoder.block.20.layer.0.SelfAttention.k": {
567
+ "weights": "qfloat8_e4m3fn",
568
+ "activations": "none"
569
+ },
570
+ "encoder.block.20.layer.0.SelfAttention.v": {
571
+ "weights": "qfloat8_e4m3fn",
572
+ "activations": "none"
573
+ },
574
+ "encoder.block.20.layer.0.SelfAttention.o": {
575
+ "weights": "qfloat8_e4m3fn",
576
+ "activations": "none"
577
+ },
578
+ "encoder.block.20.layer.1.DenseReluDense.wi_0": {
579
+ "weights": "qfloat8_e4m3fn",
580
+ "activations": "none"
581
+ },
582
+ "encoder.block.20.layer.1.DenseReluDense.wi_1": {
583
+ "weights": "qfloat8_e4m3fn",
584
+ "activations": "none"
585
+ },
586
+ "encoder.block.20.layer.1.DenseReluDense.wo": {
587
+ "weights": "qfloat8_e4m3fn",
588
+ "activations": "none"
589
+ },
590
+ "encoder.block.21.layer.0.SelfAttention.q": {
591
+ "weights": "qfloat8_e4m3fn",
592
+ "activations": "none"
593
+ },
594
+ "encoder.block.21.layer.0.SelfAttention.k": {
595
+ "weights": "qfloat8_e4m3fn",
596
+ "activations": "none"
597
+ },
598
+ "encoder.block.21.layer.0.SelfAttention.v": {
599
+ "weights": "qfloat8_e4m3fn",
600
+ "activations": "none"
601
+ },
602
+ "encoder.block.21.layer.0.SelfAttention.o": {
603
+ "weights": "qfloat8_e4m3fn",
604
+ "activations": "none"
605
+ },
606
+ "encoder.block.21.layer.1.DenseReluDense.wi_0": {
607
+ "weights": "qfloat8_e4m3fn",
608
+ "activations": "none"
609
+ },
610
+ "encoder.block.21.layer.1.DenseReluDense.wi_1": {
611
+ "weights": "qfloat8_e4m3fn",
612
+ "activations": "none"
613
+ },
614
+ "encoder.block.21.layer.1.DenseReluDense.wo": {
615
+ "weights": "qfloat8_e4m3fn",
616
+ "activations": "none"
617
+ },
618
+ "encoder.block.22.layer.0.SelfAttention.q": {
619
+ "weights": "qfloat8_e4m3fn",
620
+ "activations": "none"
621
+ },
622
+ "encoder.block.22.layer.0.SelfAttention.k": {
623
+ "weights": "qfloat8_e4m3fn",
624
+ "activations": "none"
625
+ },
626
+ "encoder.block.22.layer.0.SelfAttention.v": {
627
+ "weights": "qfloat8_e4m3fn",
628
+ "activations": "none"
629
+ },
630
+ "encoder.block.22.layer.0.SelfAttention.o": {
631
+ "weights": "qfloat8_e4m3fn",
632
+ "activations": "none"
633
+ },
634
+ "encoder.block.22.layer.1.DenseReluDense.wi_0": {
635
+ "weights": "qfloat8_e4m3fn",
636
+ "activations": "none"
637
+ },
638
+ "encoder.block.22.layer.1.DenseReluDense.wi_1": {
639
+ "weights": "qfloat8_e4m3fn",
640
+ "activations": "none"
641
+ },
642
+ "encoder.block.22.layer.1.DenseReluDense.wo": {
643
+ "weights": "qfloat8_e4m3fn",
644
+ "activations": "none"
645
+ },
646
+ "encoder.block.23.layer.0.SelfAttention.q": {
647
+ "weights": "qfloat8_e4m3fn",
648
+ "activations": "none"
649
+ },
650
+ "encoder.block.23.layer.0.SelfAttention.k": {
651
+ "weights": "qfloat8_e4m3fn",
652
+ "activations": "none"
653
+ },
654
+ "encoder.block.23.layer.0.SelfAttention.v": {
655
+ "weights": "qfloat8_e4m3fn",
656
+ "activations": "none"
657
+ },
658
+ "encoder.block.23.layer.0.SelfAttention.o": {
659
+ "weights": "qfloat8_e4m3fn",
660
+ "activations": "none"
661
+ },
662
+ "encoder.block.23.layer.1.DenseReluDense.wi_0": {
663
+ "weights": "qfloat8_e4m3fn",
664
+ "activations": "none"
665
+ },
666
+ "encoder.block.23.layer.1.DenseReluDense.wi_1": {
667
+ "weights": "qfloat8_e4m3fn",
668
+ "activations": "none"
669
+ },
670
+ "encoder.block.23.layer.1.DenseReluDense.wo": {
671
+ "weights": "qfloat8_e4m3fn",
672
+ "activations": "none"
673
+ }
674
+ }
t5_xxl/qfloat8_e5m2/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "black-forest-labs/FLUX.1-schnell",
3
+ "architectures": [
4
+ "T5EncoderModel"
5
+ ],
6
+ "classifier_dropout": 0.0,
7
+ "d_ff": 10240,
8
+ "d_kv": 64,
9
+ "d_model": 4096,
10
+ "decoder_start_token_id": 0,
11
+ "dense_act_fn": "gelu_new",
12
+ "dropout_rate": 0.1,
13
+ "eos_token_id": 1,
14
+ "feed_forward_proj": "gated-gelu",
15
+ "initializer_factor": 1.0,
16
+ "is_encoder_decoder": true,
17
+ "is_gated_act": true,
18
+ "layer_norm_epsilon": 1e-06,
19
+ "model_type": "t5",
20
+ "num_decoder_layers": 24,
21
+ "num_heads": 64,
22
+ "num_layers": 24,
23
+ "output_past": true,
24
+ "pad_token_id": 0,
25
+ "relative_attention_max_distance": 128,
26
+ "relative_attention_num_buckets": 32,
27
+ "tie_word_embeddings": false,
28
+ "torch_dtype": "bfloat16",
29
+ "transformers_version": "4.43.4",
30
+ "use_cache": true,
31
+ "vocab_size": 32128
32
+ }
t5_xxl/qfloat8_e5m2/quanto_qmap.json ADDED
@@ -0,0 +1,674 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "encoder.block.0.layer.0.SelfAttention.q": {
3
+ "weights": "qfloat8_e5m2",
4
+ "activations": "none"
5
+ },
6
+ "encoder.block.0.layer.0.SelfAttention.k": {
7
+ "weights": "qfloat8_e5m2",
8
+ "activations": "none"
9
+ },
10
+ "encoder.block.0.layer.0.SelfAttention.v": {
11
+ "weights": "qfloat8_e5m2",
12
+ "activations": "none"
13
+ },
14
+ "encoder.block.0.layer.0.SelfAttention.o": {
15
+ "weights": "qfloat8_e5m2",
16
+ "activations": "none"
17
+ },
18
+ "encoder.block.0.layer.1.DenseReluDense.wi_0": {
19
+ "weights": "qfloat8_e5m2",
20
+ "activations": "none"
21
+ },
22
+ "encoder.block.0.layer.1.DenseReluDense.wi_1": {
23
+ "weights": "qfloat8_e5m2",
24
+ "activations": "none"
25
+ },
26
+ "encoder.block.0.layer.1.DenseReluDense.wo": {
27
+ "weights": "qfloat8_e5m2",
28
+ "activations": "none"
29
+ },
30
+ "encoder.block.1.layer.0.SelfAttention.q": {
31
+ "weights": "qfloat8_e5m2",
32
+ "activations": "none"
33
+ },
34
+ "encoder.block.1.layer.0.SelfAttention.k": {
35
+ "weights": "qfloat8_e5m2",
36
+ "activations": "none"
37
+ },
38
+ "encoder.block.1.layer.0.SelfAttention.v": {
39
+ "weights": "qfloat8_e5m2",
40
+ "activations": "none"
41
+ },
42
+ "encoder.block.1.layer.0.SelfAttention.o": {
43
+ "weights": "qfloat8_e5m2",
44
+ "activations": "none"
45
+ },
46
+ "encoder.block.1.layer.1.DenseReluDense.wi_0": {
47
+ "weights": "qfloat8_e5m2",
48
+ "activations": "none"
49
+ },
50
+ "encoder.block.1.layer.1.DenseReluDense.wi_1": {
51
+ "weights": "qfloat8_e5m2",
52
+ "activations": "none"
53
+ },
54
+ "encoder.block.1.layer.1.DenseReluDense.wo": {
55
+ "weights": "qfloat8_e5m2",
56
+ "activations": "none"
57
+ },
58
+ "encoder.block.2.layer.0.SelfAttention.q": {
59
+ "weights": "qfloat8_e5m2",
60
+ "activations": "none"
61
+ },
62
+ "encoder.block.2.layer.0.SelfAttention.k": {
63
+ "weights": "qfloat8_e5m2",
64
+ "activations": "none"
65
+ },
66
+ "encoder.block.2.layer.0.SelfAttention.v": {
67
+ "weights": "qfloat8_e5m2",
68
+ "activations": "none"
69
+ },
70
+ "encoder.block.2.layer.0.SelfAttention.o": {
71
+ "weights": "qfloat8_e5m2",
72
+ "activations": "none"
73
+ },
74
+ "encoder.block.2.layer.1.DenseReluDense.wi_0": {
75
+ "weights": "qfloat8_e5m2",
76
+ "activations": "none"
77
+ },
78
+ "encoder.block.2.layer.1.DenseReluDense.wi_1": {
79
+ "weights": "qfloat8_e5m2",
80
+ "activations": "none"
81
+ },
82
+ "encoder.block.2.layer.1.DenseReluDense.wo": {
83
+ "weights": "qfloat8_e5m2",
84
+ "activations": "none"
85
+ },
86
+ "encoder.block.3.layer.0.SelfAttention.q": {
87
+ "weights": "qfloat8_e5m2",
88
+ "activations": "none"
89
+ },
90
+ "encoder.block.3.layer.0.SelfAttention.k": {
91
+ "weights": "qfloat8_e5m2",
92
+ "activations": "none"
93
+ },
94
+ "encoder.block.3.layer.0.SelfAttention.v": {
95
+ "weights": "qfloat8_e5m2",
96
+ "activations": "none"
97
+ },
98
+ "encoder.block.3.layer.0.SelfAttention.o": {
99
+ "weights": "qfloat8_e5m2",
100
+ "activations": "none"
101
+ },
102
+ "encoder.block.3.layer.1.DenseReluDense.wi_0": {
103
+ "weights": "qfloat8_e5m2",
104
+ "activations": "none"
105
+ },
106
+ "encoder.block.3.layer.1.DenseReluDense.wi_1": {
107
+ "weights": "qfloat8_e5m2",
108
+ "activations": "none"
109
+ },
110
+ "encoder.block.3.layer.1.DenseReluDense.wo": {
111
+ "weights": "qfloat8_e5m2",
112
+ "activations": "none"
113
+ },
114
+ "encoder.block.4.layer.0.SelfAttention.q": {
115
+ "weights": "qfloat8_e5m2",
116
+ "activations": "none"
117
+ },
118
+ "encoder.block.4.layer.0.SelfAttention.k": {
119
+ "weights": "qfloat8_e5m2",
120
+ "activations": "none"
121
+ },
122
+ "encoder.block.4.layer.0.SelfAttention.v": {
123
+ "weights": "qfloat8_e5m2",
124
+ "activations": "none"
125
+ },
126
+ "encoder.block.4.layer.0.SelfAttention.o": {
127
+ "weights": "qfloat8_e5m2",
128
+ "activations": "none"
129
+ },
130
+ "encoder.block.4.layer.1.DenseReluDense.wi_0": {
131
+ "weights": "qfloat8_e5m2",
132
+ "activations": "none"
133
+ },
134
+ "encoder.block.4.layer.1.DenseReluDense.wi_1": {
135
+ "weights": "qfloat8_e5m2",
136
+ "activations": "none"
137
+ },
138
+ "encoder.block.4.layer.1.DenseReluDense.wo": {
139
+ "weights": "qfloat8_e5m2",
140
+ "activations": "none"
141
+ },
142
+ "encoder.block.5.layer.0.SelfAttention.q": {
143
+ "weights": "qfloat8_e5m2",
144
+ "activations": "none"
145
+ },
146
+ "encoder.block.5.layer.0.SelfAttention.k": {
147
+ "weights": "qfloat8_e5m2",
148
+ "activations": "none"
149
+ },
150
+ "encoder.block.5.layer.0.SelfAttention.v": {
151
+ "weights": "qfloat8_e5m2",
152
+ "activations": "none"
153
+ },
154
+ "encoder.block.5.layer.0.SelfAttention.o": {
155
+ "weights": "qfloat8_e5m2",
156
+ "activations": "none"
157
+ },
158
+ "encoder.block.5.layer.1.DenseReluDense.wi_0": {
159
+ "weights": "qfloat8_e5m2",
160
+ "activations": "none"
161
+ },
162
+ "encoder.block.5.layer.1.DenseReluDense.wi_1": {
163
+ "weights": "qfloat8_e5m2",
164
+ "activations": "none"
165
+ },
166
+ "encoder.block.5.layer.1.DenseReluDense.wo": {
167
+ "weights": "qfloat8_e5m2",
168
+ "activations": "none"
169
+ },
170
+ "encoder.block.6.layer.0.SelfAttention.q": {
171
+ "weights": "qfloat8_e5m2",
172
+ "activations": "none"
173
+ },
174
+ "encoder.block.6.layer.0.SelfAttention.k": {
175
+ "weights": "qfloat8_e5m2",
176
+ "activations": "none"
177
+ },
178
+ "encoder.block.6.layer.0.SelfAttention.v": {
179
+ "weights": "qfloat8_e5m2",
180
+ "activations": "none"
181
+ },
182
+ "encoder.block.6.layer.0.SelfAttention.o": {
183
+ "weights": "qfloat8_e5m2",
184
+ "activations": "none"
185
+ },
186
+ "encoder.block.6.layer.1.DenseReluDense.wi_0": {
187
+ "weights": "qfloat8_e5m2",
188
+ "activations": "none"
189
+ },
190
+ "encoder.block.6.layer.1.DenseReluDense.wi_1": {
191
+ "weights": "qfloat8_e5m2",
192
+ "activations": "none"
193
+ },
194
+ "encoder.block.6.layer.1.DenseReluDense.wo": {
195
+ "weights": "qfloat8_e5m2",
196
+ "activations": "none"
197
+ },
198
+ "encoder.block.7.layer.0.SelfAttention.q": {
199
+ "weights": "qfloat8_e5m2",
200
+ "activations": "none"
201
+ },
202
+ "encoder.block.7.layer.0.SelfAttention.k": {
203
+ "weights": "qfloat8_e5m2",
204
+ "activations": "none"
205
+ },
206
+ "encoder.block.7.layer.0.SelfAttention.v": {
207
+ "weights": "qfloat8_e5m2",
208
+ "activations": "none"
209
+ },
210
+ "encoder.block.7.layer.0.SelfAttention.o": {
211
+ "weights": "qfloat8_e5m2",
212
+ "activations": "none"
213
+ },
214
+ "encoder.block.7.layer.1.DenseReluDense.wi_0": {
215
+ "weights": "qfloat8_e5m2",
216
+ "activations": "none"
217
+ },
218
+ "encoder.block.7.layer.1.DenseReluDense.wi_1": {
219
+ "weights": "qfloat8_e5m2",
220
+ "activations": "none"
221
+ },
222
+ "encoder.block.7.layer.1.DenseReluDense.wo": {
223
+ "weights": "qfloat8_e5m2",
224
+ "activations": "none"
225
+ },
226
+ "encoder.block.8.layer.0.SelfAttention.q": {
227
+ "weights": "qfloat8_e5m2",
228
+ "activations": "none"
229
+ },
230
+ "encoder.block.8.layer.0.SelfAttention.k": {
231
+ "weights": "qfloat8_e5m2",
232
+ "activations": "none"
233
+ },
234
+ "encoder.block.8.layer.0.SelfAttention.v": {
235
+ "weights": "qfloat8_e5m2",
236
+ "activations": "none"
237
+ },
238
+ "encoder.block.8.layer.0.SelfAttention.o": {
239
+ "weights": "qfloat8_e5m2",
240
+ "activations": "none"
241
+ },
242
+ "encoder.block.8.layer.1.DenseReluDense.wi_0": {
243
+ "weights": "qfloat8_e5m2",
244
+ "activations": "none"
245
+ },
246
+ "encoder.block.8.layer.1.DenseReluDense.wi_1": {
247
+ "weights": "qfloat8_e5m2",
248
+ "activations": "none"
249
+ },
250
+ "encoder.block.8.layer.1.DenseReluDense.wo": {
251
+ "weights": "qfloat8_e5m2",
252
+ "activations": "none"
253
+ },
254
+ "encoder.block.9.layer.0.SelfAttention.q": {
255
+ "weights": "qfloat8_e5m2",
256
+ "activations": "none"
257
+ },
258
+ "encoder.block.9.layer.0.SelfAttention.k": {
259
+ "weights": "qfloat8_e5m2",
260
+ "activations": "none"
261
+ },
262
+ "encoder.block.9.layer.0.SelfAttention.v": {
263
+ "weights": "qfloat8_e5m2",
264
+ "activations": "none"
265
+ },
266
+ "encoder.block.9.layer.0.SelfAttention.o": {
267
+ "weights": "qfloat8_e5m2",
268
+ "activations": "none"
269
+ },
270
+ "encoder.block.9.layer.1.DenseReluDense.wi_0": {
271
+ "weights": "qfloat8_e5m2",
272
+ "activations": "none"
273
+ },
274
+ "encoder.block.9.layer.1.DenseReluDense.wi_1": {
275
+ "weights": "qfloat8_e5m2",
276
+ "activations": "none"
277
+ },
278
+ "encoder.block.9.layer.1.DenseReluDense.wo": {
279
+ "weights": "qfloat8_e5m2",
280
+ "activations": "none"
281
+ },
282
+ "encoder.block.10.layer.0.SelfAttention.q": {
283
+ "weights": "qfloat8_e5m2",
284
+ "activations": "none"
285
+ },
286
+ "encoder.block.10.layer.0.SelfAttention.k": {
287
+ "weights": "qfloat8_e5m2",
288
+ "activations": "none"
289
+ },
290
+ "encoder.block.10.layer.0.SelfAttention.v": {
291
+ "weights": "qfloat8_e5m2",
292
+ "activations": "none"
293
+ },
294
+ "encoder.block.10.layer.0.SelfAttention.o": {
295
+ "weights": "qfloat8_e5m2",
296
+ "activations": "none"
297
+ },
298
+ "encoder.block.10.layer.1.DenseReluDense.wi_0": {
299
+ "weights": "qfloat8_e5m2",
300
+ "activations": "none"
301
+ },
302
+ "encoder.block.10.layer.1.DenseReluDense.wi_1": {
303
+ "weights": "qfloat8_e5m2",
304
+ "activations": "none"
305
+ },
306
+ "encoder.block.10.layer.1.DenseReluDense.wo": {
307
+ "weights": "qfloat8_e5m2",
308
+ "activations": "none"
309
+ },
310
+ "encoder.block.11.layer.0.SelfAttention.q": {
311
+ "weights": "qfloat8_e5m2",
312
+ "activations": "none"
313
+ },
314
+ "encoder.block.11.layer.0.SelfAttention.k": {
315
+ "weights": "qfloat8_e5m2",
316
+ "activations": "none"
317
+ },
318
+ "encoder.block.11.layer.0.SelfAttention.v": {
319
+ "weights": "qfloat8_e5m2",
320
+ "activations": "none"
321
+ },
322
+ "encoder.block.11.layer.0.SelfAttention.o": {
323
+ "weights": "qfloat8_e5m2",
324
+ "activations": "none"
325
+ },
326
+ "encoder.block.11.layer.1.DenseReluDense.wi_0": {
327
+ "weights": "qfloat8_e5m2",
328
+ "activations": "none"
329
+ },
330
+ "encoder.block.11.layer.1.DenseReluDense.wi_1": {
331
+ "weights": "qfloat8_e5m2",
332
+ "activations": "none"
333
+ },
334
+ "encoder.block.11.layer.1.DenseReluDense.wo": {
335
+ "weights": "qfloat8_e5m2",
336
+ "activations": "none"
337
+ },
338
+ "encoder.block.12.layer.0.SelfAttention.q": {
339
+ "weights": "qfloat8_e5m2",
340
+ "activations": "none"
341
+ },
342
+ "encoder.block.12.layer.0.SelfAttention.k": {
343
+ "weights": "qfloat8_e5m2",
344
+ "activations": "none"
345
+ },
346
+ "encoder.block.12.layer.0.SelfAttention.v": {
347
+ "weights": "qfloat8_e5m2",
348
+ "activations": "none"
349
+ },
350
+ "encoder.block.12.layer.0.SelfAttention.o": {
351
+ "weights": "qfloat8_e5m2",
352
+ "activations": "none"
353
+ },
354
+ "encoder.block.12.layer.1.DenseReluDense.wi_0": {
355
+ "weights": "qfloat8_e5m2",
356
+ "activations": "none"
357
+ },
358
+ "encoder.block.12.layer.1.DenseReluDense.wi_1": {
359
+ "weights": "qfloat8_e5m2",
360
+ "activations": "none"
361
+ },
362
+ "encoder.block.12.layer.1.DenseReluDense.wo": {
363
+ "weights": "qfloat8_e5m2",
364
+ "activations": "none"
365
+ },
366
+ "encoder.block.13.layer.0.SelfAttention.q": {
367
+ "weights": "qfloat8_e5m2",
368
+ "activations": "none"
369
+ },
370
+ "encoder.block.13.layer.0.SelfAttention.k": {
371
+ "weights": "qfloat8_e5m2",
372
+ "activations": "none"
373
+ },
374
+ "encoder.block.13.layer.0.SelfAttention.v": {
375
+ "weights": "qfloat8_e5m2",
376
+ "activations": "none"
377
+ },
378
+ "encoder.block.13.layer.0.SelfAttention.o": {
379
+ "weights": "qfloat8_e5m2",
380
+ "activations": "none"
381
+ },
382
+ "encoder.block.13.layer.1.DenseReluDense.wi_0": {
383
+ "weights": "qfloat8_e5m2",
384
+ "activations": "none"
385
+ },
386
+ "encoder.block.13.layer.1.DenseReluDense.wi_1": {
387
+ "weights": "qfloat8_e5m2",
388
+ "activations": "none"
389
+ },
390
+ "encoder.block.13.layer.1.DenseReluDense.wo": {
391
+ "weights": "qfloat8_e5m2",
392
+ "activations": "none"
393
+ },
394
+ "encoder.block.14.layer.0.SelfAttention.q": {
395
+ "weights": "qfloat8_e5m2",
396
+ "activations": "none"
397
+ },
398
+ "encoder.block.14.layer.0.SelfAttention.k": {
399
+ "weights": "qfloat8_e5m2",
400
+ "activations": "none"
401
+ },
402
+ "encoder.block.14.layer.0.SelfAttention.v": {
403
+ "weights": "qfloat8_e5m2",
404
+ "activations": "none"
405
+ },
406
+ "encoder.block.14.layer.0.SelfAttention.o": {
407
+ "weights": "qfloat8_e5m2",
408
+ "activations": "none"
409
+ },
410
+ "encoder.block.14.layer.1.DenseReluDense.wi_0": {
411
+ "weights": "qfloat8_e5m2",
412
+ "activations": "none"
413
+ },
414
+ "encoder.block.14.layer.1.DenseReluDense.wi_1": {
415
+ "weights": "qfloat8_e5m2",
416
+ "activations": "none"
417
+ },
418
+ "encoder.block.14.layer.1.DenseReluDense.wo": {
419
+ "weights": "qfloat8_e5m2",
420
+ "activations": "none"
421
+ },
422
+ "encoder.block.15.layer.0.SelfAttention.q": {
423
+ "weights": "qfloat8_e5m2",
424
+ "activations": "none"
425
+ },
426
+ "encoder.block.15.layer.0.SelfAttention.k": {
427
+ "weights": "qfloat8_e5m2",
428
+ "activations": "none"
429
+ },
430
+ "encoder.block.15.layer.0.SelfAttention.v": {
431
+ "weights": "qfloat8_e5m2",
432
+ "activations": "none"
433
+ },
434
+ "encoder.block.15.layer.0.SelfAttention.o": {
435
+ "weights": "qfloat8_e5m2",
436
+ "activations": "none"
437
+ },
438
+ "encoder.block.15.layer.1.DenseReluDense.wi_0": {
439
+ "weights": "qfloat8_e5m2",
440
+ "activations": "none"
441
+ },
442
+ "encoder.block.15.layer.1.DenseReluDense.wi_1": {
443
+ "weights": "qfloat8_e5m2",
444
+ "activations": "none"
445
+ },
446
+ "encoder.block.15.layer.1.DenseReluDense.wo": {
447
+ "weights": "qfloat8_e5m2",
448
+ "activations": "none"
449
+ },
450
+ "encoder.block.16.layer.0.SelfAttention.q": {
451
+ "weights": "qfloat8_e5m2",
452
+ "activations": "none"
453
+ },
454
+ "encoder.block.16.layer.0.SelfAttention.k": {
455
+ "weights": "qfloat8_e5m2",
456
+ "activations": "none"
457
+ },
458
+ "encoder.block.16.layer.0.SelfAttention.v": {
459
+ "weights": "qfloat8_e5m2",
460
+ "activations": "none"
461
+ },
462
+ "encoder.block.16.layer.0.SelfAttention.o": {
463
+ "weights": "qfloat8_e5m2",
464
+ "activations": "none"
465
+ },
466
+ "encoder.block.16.layer.1.DenseReluDense.wi_0": {
467
+ "weights": "qfloat8_e5m2",
468
+ "activations": "none"
469
+ },
470
+ "encoder.block.16.layer.1.DenseReluDense.wi_1": {
471
+ "weights": "qfloat8_e5m2",
472
+ "activations": "none"
473
+ },
474
+ "encoder.block.16.layer.1.DenseReluDense.wo": {
475
+ "weights": "qfloat8_e5m2",
476
+ "activations": "none"
477
+ },
478
+ "encoder.block.17.layer.0.SelfAttention.q": {
479
+ "weights": "qfloat8_e5m2",
480
+ "activations": "none"
481
+ },
482
+ "encoder.block.17.layer.0.SelfAttention.k": {
483
+ "weights": "qfloat8_e5m2",
484
+ "activations": "none"
485
+ },
486
+ "encoder.block.17.layer.0.SelfAttention.v": {
487
+ "weights": "qfloat8_e5m2",
488
+ "activations": "none"
489
+ },
490
+ "encoder.block.17.layer.0.SelfAttention.o": {
491
+ "weights": "qfloat8_e5m2",
492
+ "activations": "none"
493
+ },
494
+ "encoder.block.17.layer.1.DenseReluDense.wi_0": {
495
+ "weights": "qfloat8_e5m2",
496
+ "activations": "none"
497
+ },
498
+ "encoder.block.17.layer.1.DenseReluDense.wi_1": {
499
+ "weights": "qfloat8_e5m2",
500
+ "activations": "none"
501
+ },
502
+ "encoder.block.17.layer.1.DenseReluDense.wo": {
503
+ "weights": "qfloat8_e5m2",
504
+ "activations": "none"
505
+ },
506
+ "encoder.block.18.layer.0.SelfAttention.q": {
507
+ "weights": "qfloat8_e5m2",
508
+ "activations": "none"
509
+ },
510
+ "encoder.block.18.layer.0.SelfAttention.k": {
511
+ "weights": "qfloat8_e5m2",
512
+ "activations": "none"
513
+ },
514
+ "encoder.block.18.layer.0.SelfAttention.v": {
515
+ "weights": "qfloat8_e5m2",
516
+ "activations": "none"
517
+ },
518
+ "encoder.block.18.layer.0.SelfAttention.o": {
519
+ "weights": "qfloat8_e5m2",
520
+ "activations": "none"
521
+ },
522
+ "encoder.block.18.layer.1.DenseReluDense.wi_0": {
523
+ "weights": "qfloat8_e5m2",
524
+ "activations": "none"
525
+ },
526
+ "encoder.block.18.layer.1.DenseReluDense.wi_1": {
527
+ "weights": "qfloat8_e5m2",
528
+ "activations": "none"
529
+ },
530
+ "encoder.block.18.layer.1.DenseReluDense.wo": {
531
+ "weights": "qfloat8_e5m2",
532
+ "activations": "none"
533
+ },
534
+ "encoder.block.19.layer.0.SelfAttention.q": {
535
+ "weights": "qfloat8_e5m2",
536
+ "activations": "none"
537
+ },
538
+ "encoder.block.19.layer.0.SelfAttention.k": {
539
+ "weights": "qfloat8_e5m2",
540
+ "activations": "none"
541
+ },
542
+ "encoder.block.19.layer.0.SelfAttention.v": {
543
+ "weights": "qfloat8_e5m2",
544
+ "activations": "none"
545
+ },
546
+ "encoder.block.19.layer.0.SelfAttention.o": {
547
+ "weights": "qfloat8_e5m2",
548
+ "activations": "none"
549
+ },
550
+ "encoder.block.19.layer.1.DenseReluDense.wi_0": {
551
+ "weights": "qfloat8_e5m2",
552
+ "activations": "none"
553
+ },
554
+ "encoder.block.19.layer.1.DenseReluDense.wi_1": {
555
+ "weights": "qfloat8_e5m2",
556
+ "activations": "none"
557
+ },
558
+ "encoder.block.19.layer.1.DenseReluDense.wo": {
559
+ "weights": "qfloat8_e5m2",
560
+ "activations": "none"
561
+ },
562
+ "encoder.block.20.layer.0.SelfAttention.q": {
563
+ "weights": "qfloat8_e5m2",
564
+ "activations": "none"
565
+ },
566
+ "encoder.block.20.layer.0.SelfAttention.k": {
567
+ "weights": "qfloat8_e5m2",
568
+ "activations": "none"
569
+ },
570
+ "encoder.block.20.layer.0.SelfAttention.v": {
571
+ "weights": "qfloat8_e5m2",
572
+ "activations": "none"
573
+ },
574
+ "encoder.block.20.layer.0.SelfAttention.o": {
575
+ "weights": "qfloat8_e5m2",
576
+ "activations": "none"
577
+ },
578
+ "encoder.block.20.layer.1.DenseReluDense.wi_0": {
579
+ "weights": "qfloat8_e5m2",
580
+ "activations": "none"
581
+ },
582
+ "encoder.block.20.layer.1.DenseReluDense.wi_1": {
583
+ "weights": "qfloat8_e5m2",
584
+ "activations": "none"
585
+ },
586
+ "encoder.block.20.layer.1.DenseReluDense.wo": {
587
+ "weights": "qfloat8_e5m2",
588
+ "activations": "none"
589
+ },
590
+ "encoder.block.21.layer.0.SelfAttention.q": {
591
+ "weights": "qfloat8_e5m2",
592
+ "activations": "none"
593
+ },
594
+ "encoder.block.21.layer.0.SelfAttention.k": {
595
+ "weights": "qfloat8_e5m2",
596
+ "activations": "none"
597
+ },
598
+ "encoder.block.21.layer.0.SelfAttention.v": {
599
+ "weights": "qfloat8_e5m2",
600
+ "activations": "none"
601
+ },
602
+ "encoder.block.21.layer.0.SelfAttention.o": {
603
+ "weights": "qfloat8_e5m2",
604
+ "activations": "none"
605
+ },
606
+ "encoder.block.21.layer.1.DenseReluDense.wi_0": {
607
+ "weights": "qfloat8_e5m2",
608
+ "activations": "none"
609
+ },
610
+ "encoder.block.21.layer.1.DenseReluDense.wi_1": {
611
+ "weights": "qfloat8_e5m2",
612
+ "activations": "none"
613
+ },
614
+ "encoder.block.21.layer.1.DenseReluDense.wo": {
615
+ "weights": "qfloat8_e5m2",
616
+ "activations": "none"
617
+ },
618
+ "encoder.block.22.layer.0.SelfAttention.q": {
619
+ "weights": "qfloat8_e5m2",
620
+ "activations": "none"
621
+ },
622
+ "encoder.block.22.layer.0.SelfAttention.k": {
623
+ "weights": "qfloat8_e5m2",
624
+ "activations": "none"
625
+ },
626
+ "encoder.block.22.layer.0.SelfAttention.v": {
627
+ "weights": "qfloat8_e5m2",
628
+ "activations": "none"
629
+ },
630
+ "encoder.block.22.layer.0.SelfAttention.o": {
631
+ "weights": "qfloat8_e5m2",
632
+ "activations": "none"
633
+ },
634
+ "encoder.block.22.layer.1.DenseReluDense.wi_0": {
635
+ "weights": "qfloat8_e5m2",
636
+ "activations": "none"
637
+ },
638
+ "encoder.block.22.layer.1.DenseReluDense.wi_1": {
639
+ "weights": "qfloat8_e5m2",
640
+ "activations": "none"
641
+ },
642
+ "encoder.block.22.layer.1.DenseReluDense.wo": {
643
+ "weights": "qfloat8_e5m2",
644
+ "activations": "none"
645
+ },
646
+ "encoder.block.23.layer.0.SelfAttention.q": {
647
+ "weights": "qfloat8_e5m2",
648
+ "activations": "none"
649
+ },
650
+ "encoder.block.23.layer.0.SelfAttention.k": {
651
+ "weights": "qfloat8_e5m2",
652
+ "activations": "none"
653
+ },
654
+ "encoder.block.23.layer.0.SelfAttention.v": {
655
+ "weights": "qfloat8_e5m2",
656
+ "activations": "none"
657
+ },
658
+ "encoder.block.23.layer.0.SelfAttention.o": {
659
+ "weights": "qfloat8_e5m2",
660
+ "activations": "none"
661
+ },
662
+ "encoder.block.23.layer.1.DenseReluDense.wi_0": {
663
+ "weights": "qfloat8_e5m2",
664
+ "activations": "none"
665
+ },
666
+ "encoder.block.23.layer.1.DenseReluDense.wi_1": {
667
+ "weights": "qfloat8_e5m2",
668
+ "activations": "none"
669
+ },
670
+ "encoder.block.23.layer.1.DenseReluDense.wo": {
671
+ "weights": "qfloat8_e5m2",
672
+ "activations": "none"
673
+ }
674
+ }
t5_xxl/qint4/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "black-forest-labs/FLUX.1-schnell",
3
+ "architectures": [
4
+ "T5EncoderModel"
5
+ ],
6
+ "classifier_dropout": 0.0,
7
+ "d_ff": 10240,
8
+ "d_kv": 64,
9
+ "d_model": 4096,
10
+ "decoder_start_token_id": 0,
11
+ "dense_act_fn": "gelu_new",
12
+ "dropout_rate": 0.1,
13
+ "eos_token_id": 1,
14
+ "feed_forward_proj": "gated-gelu",
15
+ "initializer_factor": 1.0,
16
+ "is_encoder_decoder": true,
17
+ "is_gated_act": true,
18
+ "layer_norm_epsilon": 1e-06,
19
+ "model_type": "t5",
20
+ "num_decoder_layers": 24,
21
+ "num_heads": 64,
22
+ "num_layers": 24,
23
+ "output_past": true,
24
+ "pad_token_id": 0,
25
+ "relative_attention_max_distance": 128,
26
+ "relative_attention_num_buckets": 32,
27
+ "tie_word_embeddings": false,
28
+ "torch_dtype": "bfloat16",
29
+ "transformers_version": "4.43.4",
30
+ "use_cache": true,
31
+ "vocab_size": 32128
32
+ }
t5_xxl/qint4/quanto_qmap.json ADDED
@@ -0,0 +1,674 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "encoder.block.0.layer.0.SelfAttention.q": {
3
+ "weights": "qint4",
4
+ "activations": "none"
5
+ },
6
+ "encoder.block.0.layer.0.SelfAttention.k": {
7
+ "weights": "qint4",
8
+ "activations": "none"
9
+ },
10
+ "encoder.block.0.layer.0.SelfAttention.v": {
11
+ "weights": "qint4",
12
+ "activations": "none"
13
+ },
14
+ "encoder.block.0.layer.0.SelfAttention.o": {
15
+ "weights": "qint4",
16
+ "activations": "none"
17
+ },
18
+ "encoder.block.0.layer.1.DenseReluDense.wi_0": {
19
+ "weights": "qint4",
20
+ "activations": "none"
21
+ },
22
+ "encoder.block.0.layer.1.DenseReluDense.wi_1": {
23
+ "weights": "qint4",
24
+ "activations": "none"
25
+ },
26
+ "encoder.block.0.layer.1.DenseReluDense.wo": {
27
+ "weights": "qint4",
28
+ "activations": "none"
29
+ },
30
+ "encoder.block.1.layer.0.SelfAttention.q": {
31
+ "weights": "qint4",
32
+ "activations": "none"
33
+ },
34
+ "encoder.block.1.layer.0.SelfAttention.k": {
35
+ "weights": "qint4",
36
+ "activations": "none"
37
+ },
38
+ "encoder.block.1.layer.0.SelfAttention.v": {
39
+ "weights": "qint4",
40
+ "activations": "none"
41
+ },
42
+ "encoder.block.1.layer.0.SelfAttention.o": {
43
+ "weights": "qint4",
44
+ "activations": "none"
45
+ },
46
+ "encoder.block.1.layer.1.DenseReluDense.wi_0": {
47
+ "weights": "qint4",
48
+ "activations": "none"
49
+ },
50
+ "encoder.block.1.layer.1.DenseReluDense.wi_1": {
51
+ "weights": "qint4",
52
+ "activations": "none"
53
+ },
54
+ "encoder.block.1.layer.1.DenseReluDense.wo": {
55
+ "weights": "qint4",
56
+ "activations": "none"
57
+ },
58
+ "encoder.block.2.layer.0.SelfAttention.q": {
59
+ "weights": "qint4",
60
+ "activations": "none"
61
+ },
62
+ "encoder.block.2.layer.0.SelfAttention.k": {
63
+ "weights": "qint4",
64
+ "activations": "none"
65
+ },
66
+ "encoder.block.2.layer.0.SelfAttention.v": {
67
+ "weights": "qint4",
68
+ "activations": "none"
69
+ },
70
+ "encoder.block.2.layer.0.SelfAttention.o": {
71
+ "weights": "qint4",
72
+ "activations": "none"
73
+ },
74
+ "encoder.block.2.layer.1.DenseReluDense.wi_0": {
75
+ "weights": "qint4",
76
+ "activations": "none"
77
+ },
78
+ "encoder.block.2.layer.1.DenseReluDense.wi_1": {
79
+ "weights": "qint4",
80
+ "activations": "none"
81
+ },
82
+ "encoder.block.2.layer.1.DenseReluDense.wo": {
83
+ "weights": "qint4",
84
+ "activations": "none"
85
+ },
86
+ "encoder.block.3.layer.0.SelfAttention.q": {
87
+ "weights": "qint4",
88
+ "activations": "none"
89
+ },
90
+ "encoder.block.3.layer.0.SelfAttention.k": {
91
+ "weights": "qint4",
92
+ "activations": "none"
93
+ },
94
+ "encoder.block.3.layer.0.SelfAttention.v": {
95
+ "weights": "qint4",
96
+ "activations": "none"
97
+ },
98
+ "encoder.block.3.layer.0.SelfAttention.o": {
99
+ "weights": "qint4",
100
+ "activations": "none"
101
+ },
102
+ "encoder.block.3.layer.1.DenseReluDense.wi_0": {
103
+ "weights": "qint4",
104
+ "activations": "none"
105
+ },
106
+ "encoder.block.3.layer.1.DenseReluDense.wi_1": {
107
+ "weights": "qint4",
108
+ "activations": "none"
109
+ },
110
+ "encoder.block.3.layer.1.DenseReluDense.wo": {
111
+ "weights": "qint4",
112
+ "activations": "none"
113
+ },
114
+ "encoder.block.4.layer.0.SelfAttention.q": {
115
+ "weights": "qint4",
116
+ "activations": "none"
117
+ },
118
+ "encoder.block.4.layer.0.SelfAttention.k": {
119
+ "weights": "qint4",
120
+ "activations": "none"
121
+ },
122
+ "encoder.block.4.layer.0.SelfAttention.v": {
123
+ "weights": "qint4",
124
+ "activations": "none"
125
+ },
126
+ "encoder.block.4.layer.0.SelfAttention.o": {
127
+ "weights": "qint4",
128
+ "activations": "none"
129
+ },
130
+ "encoder.block.4.layer.1.DenseReluDense.wi_0": {
131
+ "weights": "qint4",
132
+ "activations": "none"
133
+ },
134
+ "encoder.block.4.layer.1.DenseReluDense.wi_1": {
135
+ "weights": "qint4",
136
+ "activations": "none"
137
+ },
138
+ "encoder.block.4.layer.1.DenseReluDense.wo": {
139
+ "weights": "qint4",
140
+ "activations": "none"
141
+ },
142
+ "encoder.block.5.layer.0.SelfAttention.q": {
143
+ "weights": "qint4",
144
+ "activations": "none"
145
+ },
146
+ "encoder.block.5.layer.0.SelfAttention.k": {
147
+ "weights": "qint4",
148
+ "activations": "none"
149
+ },
150
+ "encoder.block.5.layer.0.SelfAttention.v": {
151
+ "weights": "qint4",
152
+ "activations": "none"
153
+ },
154
+ "encoder.block.5.layer.0.SelfAttention.o": {
155
+ "weights": "qint4",
156
+ "activations": "none"
157
+ },
158
+ "encoder.block.5.layer.1.DenseReluDense.wi_0": {
159
+ "weights": "qint4",
160
+ "activations": "none"
161
+ },
162
+ "encoder.block.5.layer.1.DenseReluDense.wi_1": {
163
+ "weights": "qint4",
164
+ "activations": "none"
165
+ },
166
+ "encoder.block.5.layer.1.DenseReluDense.wo": {
167
+ "weights": "qint4",
168
+ "activations": "none"
169
+ },
170
+ "encoder.block.6.layer.0.SelfAttention.q": {
171
+ "weights": "qint4",
172
+ "activations": "none"
173
+ },
174
+ "encoder.block.6.layer.0.SelfAttention.k": {
175
+ "weights": "qint4",
176
+ "activations": "none"
177
+ },
178
+ "encoder.block.6.layer.0.SelfAttention.v": {
179
+ "weights": "qint4",
180
+ "activations": "none"
181
+ },
182
+ "encoder.block.6.layer.0.SelfAttention.o": {
183
+ "weights": "qint4",
184
+ "activations": "none"
185
+ },
186
+ "encoder.block.6.layer.1.DenseReluDense.wi_0": {
187
+ "weights": "qint4",
188
+ "activations": "none"
189
+ },
190
+ "encoder.block.6.layer.1.DenseReluDense.wi_1": {
191
+ "weights": "qint4",
192
+ "activations": "none"
193
+ },
194
+ "encoder.block.6.layer.1.DenseReluDense.wo": {
195
+ "weights": "qint4",
196
+ "activations": "none"
197
+ },
198
+ "encoder.block.7.layer.0.SelfAttention.q": {
199
+ "weights": "qint4",
200
+ "activations": "none"
201
+ },
202
+ "encoder.block.7.layer.0.SelfAttention.k": {
203
+ "weights": "qint4",
204
+ "activations": "none"
205
+ },
206
+ "encoder.block.7.layer.0.SelfAttention.v": {
207
+ "weights": "qint4",
208
+ "activations": "none"
209
+ },
210
+ "encoder.block.7.layer.0.SelfAttention.o": {
211
+ "weights": "qint4",
212
+ "activations": "none"
213
+ },
214
+ "encoder.block.7.layer.1.DenseReluDense.wi_0": {
215
+ "weights": "qint4",
216
+ "activations": "none"
217
+ },
218
+ "encoder.block.7.layer.1.DenseReluDense.wi_1": {
219
+ "weights": "qint4",
220
+ "activations": "none"
221
+ },
222
+ "encoder.block.7.layer.1.DenseReluDense.wo": {
223
+ "weights": "qint4",
224
+ "activations": "none"
225
+ },
226
+ "encoder.block.8.layer.0.SelfAttention.q": {
227
+ "weights": "qint4",
228
+ "activations": "none"
229
+ },
230
+ "encoder.block.8.layer.0.SelfAttention.k": {
231
+ "weights": "qint4",
232
+ "activations": "none"
233
+ },
234
+ "encoder.block.8.layer.0.SelfAttention.v": {
235
+ "weights": "qint4",
236
+ "activations": "none"
237
+ },
238
+ "encoder.block.8.layer.0.SelfAttention.o": {
239
+ "weights": "qint4",
240
+ "activations": "none"
241
+ },
242
+ "encoder.block.8.layer.1.DenseReluDense.wi_0": {
243
+ "weights": "qint4",
244
+ "activations": "none"
245
+ },
246
+ "encoder.block.8.layer.1.DenseReluDense.wi_1": {
247
+ "weights": "qint4",
248
+ "activations": "none"
249
+ },
250
+ "encoder.block.8.layer.1.DenseReluDense.wo": {
251
+ "weights": "qint4",
252
+ "activations": "none"
253
+ },
254
+ "encoder.block.9.layer.0.SelfAttention.q": {
255
+ "weights": "qint4",
256
+ "activations": "none"
257
+ },
258
+ "encoder.block.9.layer.0.SelfAttention.k": {
259
+ "weights": "qint4",
260
+ "activations": "none"
261
+ },
262
+ "encoder.block.9.layer.0.SelfAttention.v": {
263
+ "weights": "qint4",
264
+ "activations": "none"
265
+ },
266
+ "encoder.block.9.layer.0.SelfAttention.o": {
267
+ "weights": "qint4",
268
+ "activations": "none"
269
+ },
270
+ "encoder.block.9.layer.1.DenseReluDense.wi_0": {
271
+ "weights": "qint4",
272
+ "activations": "none"
273
+ },
274
+ "encoder.block.9.layer.1.DenseReluDense.wi_1": {
275
+ "weights": "qint4",
276
+ "activations": "none"
277
+ },
278
+ "encoder.block.9.layer.1.DenseReluDense.wo": {
279
+ "weights": "qint4",
280
+ "activations": "none"
281
+ },
282
+ "encoder.block.10.layer.0.SelfAttention.q": {
283
+ "weights": "qint4",
284
+ "activations": "none"
285
+ },
286
+ "encoder.block.10.layer.0.SelfAttention.k": {
287
+ "weights": "qint4",
288
+ "activations": "none"
289
+ },
290
+ "encoder.block.10.layer.0.SelfAttention.v": {
291
+ "weights": "qint4",
292
+ "activations": "none"
293
+ },
294
+ "encoder.block.10.layer.0.SelfAttention.o": {
295
+ "weights": "qint4",
296
+ "activations": "none"
297
+ },
298
+ "encoder.block.10.layer.1.DenseReluDense.wi_0": {
299
+ "weights": "qint4",
300
+ "activations": "none"
301
+ },
302
+ "encoder.block.10.layer.1.DenseReluDense.wi_1": {
303
+ "weights": "qint4",
304
+ "activations": "none"
305
+ },
306
+ "encoder.block.10.layer.1.DenseReluDense.wo": {
307
+ "weights": "qint4",
308
+ "activations": "none"
309
+ },
310
+ "encoder.block.11.layer.0.SelfAttention.q": {
311
+ "weights": "qint4",
312
+ "activations": "none"
313
+ },
314
+ "encoder.block.11.layer.0.SelfAttention.k": {
315
+ "weights": "qint4",
316
+ "activations": "none"
317
+ },
318
+ "encoder.block.11.layer.0.SelfAttention.v": {
319
+ "weights": "qint4",
320
+ "activations": "none"
321
+ },
322
+ "encoder.block.11.layer.0.SelfAttention.o": {
323
+ "weights": "qint4",
324
+ "activations": "none"
325
+ },
326
+ "encoder.block.11.layer.1.DenseReluDense.wi_0": {
327
+ "weights": "qint4",
328
+ "activations": "none"
329
+ },
330
+ "encoder.block.11.layer.1.DenseReluDense.wi_1": {
331
+ "weights": "qint4",
332
+ "activations": "none"
333
+ },
334
+ "encoder.block.11.layer.1.DenseReluDense.wo": {
335
+ "weights": "qint4",
336
+ "activations": "none"
337
+ },
338
+ "encoder.block.12.layer.0.SelfAttention.q": {
339
+ "weights": "qint4",
340
+ "activations": "none"
341
+ },
342
+ "encoder.block.12.layer.0.SelfAttention.k": {
343
+ "weights": "qint4",
344
+ "activations": "none"
345
+ },
346
+ "encoder.block.12.layer.0.SelfAttention.v": {
347
+ "weights": "qint4",
348
+ "activations": "none"
349
+ },
350
+ "encoder.block.12.layer.0.SelfAttention.o": {
351
+ "weights": "qint4",
352
+ "activations": "none"
353
+ },
354
+ "encoder.block.12.layer.1.DenseReluDense.wi_0": {
355
+ "weights": "qint4",
356
+ "activations": "none"
357
+ },
358
+ "encoder.block.12.layer.1.DenseReluDense.wi_1": {
359
+ "weights": "qint4",
360
+ "activations": "none"
361
+ },
362
+ "encoder.block.12.layer.1.DenseReluDense.wo": {
363
+ "weights": "qint4",
364
+ "activations": "none"
365
+ },
366
+ "encoder.block.13.layer.0.SelfAttention.q": {
367
+ "weights": "qint4",
368
+ "activations": "none"
369
+ },
370
+ "encoder.block.13.layer.0.SelfAttention.k": {
371
+ "weights": "qint4",
372
+ "activations": "none"
373
+ },
374
+ "encoder.block.13.layer.0.SelfAttention.v": {
375
+ "weights": "qint4",
376
+ "activations": "none"
377
+ },
378
+ "encoder.block.13.layer.0.SelfAttention.o": {
379
+ "weights": "qint4",
380
+ "activations": "none"
381
+ },
382
+ "encoder.block.13.layer.1.DenseReluDense.wi_0": {
383
+ "weights": "qint4",
384
+ "activations": "none"
385
+ },
386
+ "encoder.block.13.layer.1.DenseReluDense.wi_1": {
387
+ "weights": "qint4",
388
+ "activations": "none"
389
+ },
390
+ "encoder.block.13.layer.1.DenseReluDense.wo": {
391
+ "weights": "qint4",
392
+ "activations": "none"
393
+ },
394
+ "encoder.block.14.layer.0.SelfAttention.q": {
395
+ "weights": "qint4",
396
+ "activations": "none"
397
+ },
398
+ "encoder.block.14.layer.0.SelfAttention.k": {
399
+ "weights": "qint4",
400
+ "activations": "none"
401
+ },
402
+ "encoder.block.14.layer.0.SelfAttention.v": {
403
+ "weights": "qint4",
404
+ "activations": "none"
405
+ },
406
+ "encoder.block.14.layer.0.SelfAttention.o": {
407
+ "weights": "qint4",
408
+ "activations": "none"
409
+ },
410
+ "encoder.block.14.layer.1.DenseReluDense.wi_0": {
411
+ "weights": "qint4",
412
+ "activations": "none"
413
+ },
414
+ "encoder.block.14.layer.1.DenseReluDense.wi_1": {
415
+ "weights": "qint4",
416
+ "activations": "none"
417
+ },
418
+ "encoder.block.14.layer.1.DenseReluDense.wo": {
419
+ "weights": "qint4",
420
+ "activations": "none"
421
+ },
422
+ "encoder.block.15.layer.0.SelfAttention.q": {
423
+ "weights": "qint4",
424
+ "activations": "none"
425
+ },
426
+ "encoder.block.15.layer.0.SelfAttention.k": {
427
+ "weights": "qint4",
428
+ "activations": "none"
429
+ },
430
+ "encoder.block.15.layer.0.SelfAttention.v": {
431
+ "weights": "qint4",
432
+ "activations": "none"
433
+ },
434
+ "encoder.block.15.layer.0.SelfAttention.o": {
435
+ "weights": "qint4",
436
+ "activations": "none"
437
+ },
438
+ "encoder.block.15.layer.1.DenseReluDense.wi_0": {
439
+ "weights": "qint4",
440
+ "activations": "none"
441
+ },
442
+ "encoder.block.15.layer.1.DenseReluDense.wi_1": {
443
+ "weights": "qint4",
444
+ "activations": "none"
445
+ },
446
+ "encoder.block.15.layer.1.DenseReluDense.wo": {
447
+ "weights": "qint4",
448
+ "activations": "none"
449
+ },
450
+ "encoder.block.16.layer.0.SelfAttention.q": {
451
+ "weights": "qint4",
452
+ "activations": "none"
453
+ },
454
+ "encoder.block.16.layer.0.SelfAttention.k": {
455
+ "weights": "qint4",
456
+ "activations": "none"
457
+ },
458
+ "encoder.block.16.layer.0.SelfAttention.v": {
459
+ "weights": "qint4",
460
+ "activations": "none"
461
+ },
462
+ "encoder.block.16.layer.0.SelfAttention.o": {
463
+ "weights": "qint4",
464
+ "activations": "none"
465
+ },
466
+ "encoder.block.16.layer.1.DenseReluDense.wi_0": {
467
+ "weights": "qint4",
468
+ "activations": "none"
469
+ },
470
+ "encoder.block.16.layer.1.DenseReluDense.wi_1": {
471
+ "weights": "qint4",
472
+ "activations": "none"
473
+ },
474
+ "encoder.block.16.layer.1.DenseReluDense.wo": {
475
+ "weights": "qint4",
476
+ "activations": "none"
477
+ },
478
+ "encoder.block.17.layer.0.SelfAttention.q": {
479
+ "weights": "qint4",
480
+ "activations": "none"
481
+ },
482
+ "encoder.block.17.layer.0.SelfAttention.k": {
483
+ "weights": "qint4",
484
+ "activations": "none"
485
+ },
486
+ "encoder.block.17.layer.0.SelfAttention.v": {
487
+ "weights": "qint4",
488
+ "activations": "none"
489
+ },
490
+ "encoder.block.17.layer.0.SelfAttention.o": {
491
+ "weights": "qint4",
492
+ "activations": "none"
493
+ },
494
+ "encoder.block.17.layer.1.DenseReluDense.wi_0": {
495
+ "weights": "qint4",
496
+ "activations": "none"
497
+ },
498
+ "encoder.block.17.layer.1.DenseReluDense.wi_1": {
499
+ "weights": "qint4",
500
+ "activations": "none"
501
+ },
502
+ "encoder.block.17.layer.1.DenseReluDense.wo": {
503
+ "weights": "qint4",
504
+ "activations": "none"
505
+ },
506
+ "encoder.block.18.layer.0.SelfAttention.q": {
507
+ "weights": "qint4",
508
+ "activations": "none"
509
+ },
510
+ "encoder.block.18.layer.0.SelfAttention.k": {
511
+ "weights": "qint4",
512
+ "activations": "none"
513
+ },
514
+ "encoder.block.18.layer.0.SelfAttention.v": {
515
+ "weights": "qint4",
516
+ "activations": "none"
517
+ },
518
+ "encoder.block.18.layer.0.SelfAttention.o": {
519
+ "weights": "qint4",
520
+ "activations": "none"
521
+ },
522
+ "encoder.block.18.layer.1.DenseReluDense.wi_0": {
523
+ "weights": "qint4",
524
+ "activations": "none"
525
+ },
526
+ "encoder.block.18.layer.1.DenseReluDense.wi_1": {
527
+ "weights": "qint4",
528
+ "activations": "none"
529
+ },
530
+ "encoder.block.18.layer.1.DenseReluDense.wo": {
531
+ "weights": "qint4",
532
+ "activations": "none"
533
+ },
534
+ "encoder.block.19.layer.0.SelfAttention.q": {
535
+ "weights": "qint4",
536
+ "activations": "none"
537
+ },
538
+ "encoder.block.19.layer.0.SelfAttention.k": {
539
+ "weights": "qint4",
540
+ "activations": "none"
541
+ },
542
+ "encoder.block.19.layer.0.SelfAttention.v": {
543
+ "weights": "qint4",
544
+ "activations": "none"
545
+ },
546
+ "encoder.block.19.layer.0.SelfAttention.o": {
547
+ "weights": "qint4",
548
+ "activations": "none"
549
+ },
550
+ "encoder.block.19.layer.1.DenseReluDense.wi_0": {
551
+ "weights": "qint4",
552
+ "activations": "none"
553
+ },
554
+ "encoder.block.19.layer.1.DenseReluDense.wi_1": {
555
+ "weights": "qint4",
556
+ "activations": "none"
557
+ },
558
+ "encoder.block.19.layer.1.DenseReluDense.wo": {
559
+ "weights": "qint4",
560
+ "activations": "none"
561
+ },
562
+ "encoder.block.20.layer.0.SelfAttention.q": {
563
+ "weights": "qint4",
564
+ "activations": "none"
565
+ },
566
+ "encoder.block.20.layer.0.SelfAttention.k": {
567
+ "weights": "qint4",
568
+ "activations": "none"
569
+ },
570
+ "encoder.block.20.layer.0.SelfAttention.v": {
571
+ "weights": "qint4",
572
+ "activations": "none"
573
+ },
574
+ "encoder.block.20.layer.0.SelfAttention.o": {
575
+ "weights": "qint4",
576
+ "activations": "none"
577
+ },
578
+ "encoder.block.20.layer.1.DenseReluDense.wi_0": {
579
+ "weights": "qint4",
580
+ "activations": "none"
581
+ },
582
+ "encoder.block.20.layer.1.DenseReluDense.wi_1": {
583
+ "weights": "qint4",
584
+ "activations": "none"
585
+ },
586
+ "encoder.block.20.layer.1.DenseReluDense.wo": {
587
+ "weights": "qint4",
588
+ "activations": "none"
589
+ },
590
+ "encoder.block.21.layer.0.SelfAttention.q": {
591
+ "weights": "qint4",
592
+ "activations": "none"
593
+ },
594
+ "encoder.block.21.layer.0.SelfAttention.k": {
595
+ "weights": "qint4",
596
+ "activations": "none"
597
+ },
598
+ "encoder.block.21.layer.0.SelfAttention.v": {
599
+ "weights": "qint4",
600
+ "activations": "none"
601
+ },
602
+ "encoder.block.21.layer.0.SelfAttention.o": {
603
+ "weights": "qint4",
604
+ "activations": "none"
605
+ },
606
+ "encoder.block.21.layer.1.DenseReluDense.wi_0": {
607
+ "weights": "qint4",
608
+ "activations": "none"
609
+ },
610
+ "encoder.block.21.layer.1.DenseReluDense.wi_1": {
611
+ "weights": "qint4",
612
+ "activations": "none"
613
+ },
614
+ "encoder.block.21.layer.1.DenseReluDense.wo": {
615
+ "weights": "qint4",
616
+ "activations": "none"
617
+ },
618
+ "encoder.block.22.layer.0.SelfAttention.q": {
619
+ "weights": "qint4",
620
+ "activations": "none"
621
+ },
622
+ "encoder.block.22.layer.0.SelfAttention.k": {
623
+ "weights": "qint4",
624
+ "activations": "none"
625
+ },
626
+ "encoder.block.22.layer.0.SelfAttention.v": {
627
+ "weights": "qint4",
628
+ "activations": "none"
629
+ },
630
+ "encoder.block.22.layer.0.SelfAttention.o": {
631
+ "weights": "qint4",
632
+ "activations": "none"
633
+ },
634
+ "encoder.block.22.layer.1.DenseReluDense.wi_0": {
635
+ "weights": "qint4",
636
+ "activations": "none"
637
+ },
638
+ "encoder.block.22.layer.1.DenseReluDense.wi_1": {
639
+ "weights": "qint4",
640
+ "activations": "none"
641
+ },
642
+ "encoder.block.22.layer.1.DenseReluDense.wo": {
643
+ "weights": "qint4",
644
+ "activations": "none"
645
+ },
646
+ "encoder.block.23.layer.0.SelfAttention.q": {
647
+ "weights": "qint4",
648
+ "activations": "none"
649
+ },
650
+ "encoder.block.23.layer.0.SelfAttention.k": {
651
+ "weights": "qint4",
652
+ "activations": "none"
653
+ },
654
+ "encoder.block.23.layer.0.SelfAttention.v": {
655
+ "weights": "qint4",
656
+ "activations": "none"
657
+ },
658
+ "encoder.block.23.layer.0.SelfAttention.o": {
659
+ "weights": "qint4",
660
+ "activations": "none"
661
+ },
662
+ "encoder.block.23.layer.1.DenseReluDense.wi_0": {
663
+ "weights": "qint4",
664
+ "activations": "none"
665
+ },
666
+ "encoder.block.23.layer.1.DenseReluDense.wi_1": {
667
+ "weights": "qint4",
668
+ "activations": "none"
669
+ },
670
+ "encoder.block.23.layer.1.DenseReluDense.wo": {
671
+ "weights": "qint4",
672
+ "activations": "none"
673
+ }
674
+ }
t5_xxl/qint8/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "black-forest-labs/FLUX.1-schnell",
3
+ "architectures": [
4
+ "T5EncoderModel"
5
+ ],
6
+ "classifier_dropout": 0.0,
7
+ "d_ff": 10240,
8
+ "d_kv": 64,
9
+ "d_model": 4096,
10
+ "decoder_start_token_id": 0,
11
+ "dense_act_fn": "gelu_new",
12
+ "dropout_rate": 0.1,
13
+ "eos_token_id": 1,
14
+ "feed_forward_proj": "gated-gelu",
15
+ "initializer_factor": 1.0,
16
+ "is_encoder_decoder": true,
17
+ "is_gated_act": true,
18
+ "layer_norm_epsilon": 1e-06,
19
+ "model_type": "t5",
20
+ "num_decoder_layers": 24,
21
+ "num_heads": 64,
22
+ "num_layers": 24,
23
+ "output_past": true,
24
+ "pad_token_id": 0,
25
+ "relative_attention_max_distance": 128,
26
+ "relative_attention_num_buckets": 32,
27
+ "tie_word_embeddings": false,
28
+ "torch_dtype": "bfloat16",
29
+ "transformers_version": "4.43.4",
30
+ "use_cache": true,
31
+ "vocab_size": 32128
32
+ }
t5_xxl/qint8/quanto_qmap.json ADDED
@@ -0,0 +1,674 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "encoder.block.0.layer.0.SelfAttention.q": {
3
+ "weights": "qint8",
4
+ "activations": "none"
5
+ },
6
+ "encoder.block.0.layer.0.SelfAttention.k": {
7
+ "weights": "qint8",
8
+ "activations": "none"
9
+ },
10
+ "encoder.block.0.layer.0.SelfAttention.v": {
11
+ "weights": "qint8",
12
+ "activations": "none"
13
+ },
14
+ "encoder.block.0.layer.0.SelfAttention.o": {
15
+ "weights": "qint8",
16
+ "activations": "none"
17
+ },
18
+ "encoder.block.0.layer.1.DenseReluDense.wi_0": {
19
+ "weights": "qint8",
20
+ "activations": "none"
21
+ },
22
+ "encoder.block.0.layer.1.DenseReluDense.wi_1": {
23
+ "weights": "qint8",
24
+ "activations": "none"
25
+ },
26
+ "encoder.block.0.layer.1.DenseReluDense.wo": {
27
+ "weights": "qint8",
28
+ "activations": "none"
29
+ },
30
+ "encoder.block.1.layer.0.SelfAttention.q": {
31
+ "weights": "qint8",
32
+ "activations": "none"
33
+ },
34
+ "encoder.block.1.layer.0.SelfAttention.k": {
35
+ "weights": "qint8",
36
+ "activations": "none"
37
+ },
38
+ "encoder.block.1.layer.0.SelfAttention.v": {
39
+ "weights": "qint8",
40
+ "activations": "none"
41
+ },
42
+ "encoder.block.1.layer.0.SelfAttention.o": {
43
+ "weights": "qint8",
44
+ "activations": "none"
45
+ },
46
+ "encoder.block.1.layer.1.DenseReluDense.wi_0": {
47
+ "weights": "qint8",
48
+ "activations": "none"
49
+ },
50
+ "encoder.block.1.layer.1.DenseReluDense.wi_1": {
51
+ "weights": "qint8",
52
+ "activations": "none"
53
+ },
54
+ "encoder.block.1.layer.1.DenseReluDense.wo": {
55
+ "weights": "qint8",
56
+ "activations": "none"
57
+ },
58
+ "encoder.block.2.layer.0.SelfAttention.q": {
59
+ "weights": "qint8",
60
+ "activations": "none"
61
+ },
62
+ "encoder.block.2.layer.0.SelfAttention.k": {
63
+ "weights": "qint8",
64
+ "activations": "none"
65
+ },
66
+ "encoder.block.2.layer.0.SelfAttention.v": {
67
+ "weights": "qint8",
68
+ "activations": "none"
69
+ },
70
+ "encoder.block.2.layer.0.SelfAttention.o": {
71
+ "weights": "qint8",
72
+ "activations": "none"
73
+ },
74
+ "encoder.block.2.layer.1.DenseReluDense.wi_0": {
75
+ "weights": "qint8",
76
+ "activations": "none"
77
+ },
78
+ "encoder.block.2.layer.1.DenseReluDense.wi_1": {
79
+ "weights": "qint8",
80
+ "activations": "none"
81
+ },
82
+ "encoder.block.2.layer.1.DenseReluDense.wo": {
83
+ "weights": "qint8",
84
+ "activations": "none"
85
+ },
86
+ "encoder.block.3.layer.0.SelfAttention.q": {
87
+ "weights": "qint8",
88
+ "activations": "none"
89
+ },
90
+ "encoder.block.3.layer.0.SelfAttention.k": {
91
+ "weights": "qint8",
92
+ "activations": "none"
93
+ },
94
+ "encoder.block.3.layer.0.SelfAttention.v": {
95
+ "weights": "qint8",
96
+ "activations": "none"
97
+ },
98
+ "encoder.block.3.layer.0.SelfAttention.o": {
99
+ "weights": "qint8",
100
+ "activations": "none"
101
+ },
102
+ "encoder.block.3.layer.1.DenseReluDense.wi_0": {
103
+ "weights": "qint8",
104
+ "activations": "none"
105
+ },
106
+ "encoder.block.3.layer.1.DenseReluDense.wi_1": {
107
+ "weights": "qint8",
108
+ "activations": "none"
109
+ },
110
+ "encoder.block.3.layer.1.DenseReluDense.wo": {
111
+ "weights": "qint8",
112
+ "activations": "none"
113
+ },
114
+ "encoder.block.4.layer.0.SelfAttention.q": {
115
+ "weights": "qint8",
116
+ "activations": "none"
117
+ },
118
+ "encoder.block.4.layer.0.SelfAttention.k": {
119
+ "weights": "qint8",
120
+ "activations": "none"
121
+ },
122
+ "encoder.block.4.layer.0.SelfAttention.v": {
123
+ "weights": "qint8",
124
+ "activations": "none"
125
+ },
126
+ "encoder.block.4.layer.0.SelfAttention.o": {
127
+ "weights": "qint8",
128
+ "activations": "none"
129
+ },
130
+ "encoder.block.4.layer.1.DenseReluDense.wi_0": {
131
+ "weights": "qint8",
132
+ "activations": "none"
133
+ },
134
+ "encoder.block.4.layer.1.DenseReluDense.wi_1": {
135
+ "weights": "qint8",
136
+ "activations": "none"
137
+ },
138
+ "encoder.block.4.layer.1.DenseReluDense.wo": {
139
+ "weights": "qint8",
140
+ "activations": "none"
141
+ },
142
+ "encoder.block.5.layer.0.SelfAttention.q": {
143
+ "weights": "qint8",
144
+ "activations": "none"
145
+ },
146
+ "encoder.block.5.layer.0.SelfAttention.k": {
147
+ "weights": "qint8",
148
+ "activations": "none"
149
+ },
150
+ "encoder.block.5.layer.0.SelfAttention.v": {
151
+ "weights": "qint8",
152
+ "activations": "none"
153
+ },
154
+ "encoder.block.5.layer.0.SelfAttention.o": {
155
+ "weights": "qint8",
156
+ "activations": "none"
157
+ },
158
+ "encoder.block.5.layer.1.DenseReluDense.wi_0": {
159
+ "weights": "qint8",
160
+ "activations": "none"
161
+ },
162
+ "encoder.block.5.layer.1.DenseReluDense.wi_1": {
163
+ "weights": "qint8",
164
+ "activations": "none"
165
+ },
166
+ "encoder.block.5.layer.1.DenseReluDense.wo": {
167
+ "weights": "qint8",
168
+ "activations": "none"
169
+ },
170
+ "encoder.block.6.layer.0.SelfAttention.q": {
171
+ "weights": "qint8",
172
+ "activations": "none"
173
+ },
174
+ "encoder.block.6.layer.0.SelfAttention.k": {
175
+ "weights": "qint8",
176
+ "activations": "none"
177
+ },
178
+ "encoder.block.6.layer.0.SelfAttention.v": {
179
+ "weights": "qint8",
180
+ "activations": "none"
181
+ },
182
+ "encoder.block.6.layer.0.SelfAttention.o": {
183
+ "weights": "qint8",
184
+ "activations": "none"
185
+ },
186
+ "encoder.block.6.layer.1.DenseReluDense.wi_0": {
187
+ "weights": "qint8",
188
+ "activations": "none"
189
+ },
190
+ "encoder.block.6.layer.1.DenseReluDense.wi_1": {
191
+ "weights": "qint8",
192
+ "activations": "none"
193
+ },
194
+ "encoder.block.6.layer.1.DenseReluDense.wo": {
195
+ "weights": "qint8",
196
+ "activations": "none"
197
+ },
198
+ "encoder.block.7.layer.0.SelfAttention.q": {
199
+ "weights": "qint8",
200
+ "activations": "none"
201
+ },
202
+ "encoder.block.7.layer.0.SelfAttention.k": {
203
+ "weights": "qint8",
204
+ "activations": "none"
205
+ },
206
+ "encoder.block.7.layer.0.SelfAttention.v": {
207
+ "weights": "qint8",
208
+ "activations": "none"
209
+ },
210
+ "encoder.block.7.layer.0.SelfAttention.o": {
211
+ "weights": "qint8",
212
+ "activations": "none"
213
+ },
214
+ "encoder.block.7.layer.1.DenseReluDense.wi_0": {
215
+ "weights": "qint8",
216
+ "activations": "none"
217
+ },
218
+ "encoder.block.7.layer.1.DenseReluDense.wi_1": {
219
+ "weights": "qint8",
220
+ "activations": "none"
221
+ },
222
+ "encoder.block.7.layer.1.DenseReluDense.wo": {
223
+ "weights": "qint8",
224
+ "activations": "none"
225
+ },
226
+ "encoder.block.8.layer.0.SelfAttention.q": {
227
+ "weights": "qint8",
228
+ "activations": "none"
229
+ },
230
+ "encoder.block.8.layer.0.SelfAttention.k": {
231
+ "weights": "qint8",
232
+ "activations": "none"
233
+ },
234
+ "encoder.block.8.layer.0.SelfAttention.v": {
235
+ "weights": "qint8",
236
+ "activations": "none"
237
+ },
238
+ "encoder.block.8.layer.0.SelfAttention.o": {
239
+ "weights": "qint8",
240
+ "activations": "none"
241
+ },
242
+ "encoder.block.8.layer.1.DenseReluDense.wi_0": {
243
+ "weights": "qint8",
244
+ "activations": "none"
245
+ },
246
+ "encoder.block.8.layer.1.DenseReluDense.wi_1": {
247
+ "weights": "qint8",
248
+ "activations": "none"
249
+ },
250
+ "encoder.block.8.layer.1.DenseReluDense.wo": {
251
+ "weights": "qint8",
252
+ "activations": "none"
253
+ },
254
+ "encoder.block.9.layer.0.SelfAttention.q": {
255
+ "weights": "qint8",
256
+ "activations": "none"
257
+ },
258
+ "encoder.block.9.layer.0.SelfAttention.k": {
259
+ "weights": "qint8",
260
+ "activations": "none"
261
+ },
262
+ "encoder.block.9.layer.0.SelfAttention.v": {
263
+ "weights": "qint8",
264
+ "activations": "none"
265
+ },
266
+ "encoder.block.9.layer.0.SelfAttention.o": {
267
+ "weights": "qint8",
268
+ "activations": "none"
269
+ },
270
+ "encoder.block.9.layer.1.DenseReluDense.wi_0": {
271
+ "weights": "qint8",
272
+ "activations": "none"
273
+ },
274
+ "encoder.block.9.layer.1.DenseReluDense.wi_1": {
275
+ "weights": "qint8",
276
+ "activations": "none"
277
+ },
278
+ "encoder.block.9.layer.1.DenseReluDense.wo": {
279
+ "weights": "qint8",
280
+ "activations": "none"
281
+ },
282
+ "encoder.block.10.layer.0.SelfAttention.q": {
283
+ "weights": "qint8",
284
+ "activations": "none"
285
+ },
286
+ "encoder.block.10.layer.0.SelfAttention.k": {
287
+ "weights": "qint8",
288
+ "activations": "none"
289
+ },
290
+ "encoder.block.10.layer.0.SelfAttention.v": {
291
+ "weights": "qint8",
292
+ "activations": "none"
293
+ },
294
+ "encoder.block.10.layer.0.SelfAttention.o": {
295
+ "weights": "qint8",
296
+ "activations": "none"
297
+ },
298
+ "encoder.block.10.layer.1.DenseReluDense.wi_0": {
299
+ "weights": "qint8",
300
+ "activations": "none"
301
+ },
302
+ "encoder.block.10.layer.1.DenseReluDense.wi_1": {
303
+ "weights": "qint8",
304
+ "activations": "none"
305
+ },
306
+ "encoder.block.10.layer.1.DenseReluDense.wo": {
307
+ "weights": "qint8",
308
+ "activations": "none"
309
+ },
310
+ "encoder.block.11.layer.0.SelfAttention.q": {
311
+ "weights": "qint8",
312
+ "activations": "none"
313
+ },
314
+ "encoder.block.11.layer.0.SelfAttention.k": {
315
+ "weights": "qint8",
316
+ "activations": "none"
317
+ },
318
+ "encoder.block.11.layer.0.SelfAttention.v": {
319
+ "weights": "qint8",
320
+ "activations": "none"
321
+ },
322
+ "encoder.block.11.layer.0.SelfAttention.o": {
323
+ "weights": "qint8",
324
+ "activations": "none"
325
+ },
326
+ "encoder.block.11.layer.1.DenseReluDense.wi_0": {
327
+ "weights": "qint8",
328
+ "activations": "none"
329
+ },
330
+ "encoder.block.11.layer.1.DenseReluDense.wi_1": {
331
+ "weights": "qint8",
332
+ "activations": "none"
333
+ },
334
+ "encoder.block.11.layer.1.DenseReluDense.wo": {
335
+ "weights": "qint8",
336
+ "activations": "none"
337
+ },
338
+ "encoder.block.12.layer.0.SelfAttention.q": {
339
+ "weights": "qint8",
340
+ "activations": "none"
341
+ },
342
+ "encoder.block.12.layer.0.SelfAttention.k": {
343
+ "weights": "qint8",
344
+ "activations": "none"
345
+ },
346
+ "encoder.block.12.layer.0.SelfAttention.v": {
347
+ "weights": "qint8",
348
+ "activations": "none"
349
+ },
350
+ "encoder.block.12.layer.0.SelfAttention.o": {
351
+ "weights": "qint8",
352
+ "activations": "none"
353
+ },
354
+ "encoder.block.12.layer.1.DenseReluDense.wi_0": {
355
+ "weights": "qint8",
356
+ "activations": "none"
357
+ },
358
+ "encoder.block.12.layer.1.DenseReluDense.wi_1": {
359
+ "weights": "qint8",
360
+ "activations": "none"
361
+ },
362
+ "encoder.block.12.layer.1.DenseReluDense.wo": {
363
+ "weights": "qint8",
364
+ "activations": "none"
365
+ },
366
+ "encoder.block.13.layer.0.SelfAttention.q": {
367
+ "weights": "qint8",
368
+ "activations": "none"
369
+ },
370
+ "encoder.block.13.layer.0.SelfAttention.k": {
371
+ "weights": "qint8",
372
+ "activations": "none"
373
+ },
374
+ "encoder.block.13.layer.0.SelfAttention.v": {
375
+ "weights": "qint8",
376
+ "activations": "none"
377
+ },
378
+ "encoder.block.13.layer.0.SelfAttention.o": {
379
+ "weights": "qint8",
380
+ "activations": "none"
381
+ },
382
+ "encoder.block.13.layer.1.DenseReluDense.wi_0": {
383
+ "weights": "qint8",
384
+ "activations": "none"
385
+ },
386
+ "encoder.block.13.layer.1.DenseReluDense.wi_1": {
387
+ "weights": "qint8",
388
+ "activations": "none"
389
+ },
390
+ "encoder.block.13.layer.1.DenseReluDense.wo": {
391
+ "weights": "qint8",
392
+ "activations": "none"
393
+ },
394
+ "encoder.block.14.layer.0.SelfAttention.q": {
395
+ "weights": "qint8",
396
+ "activations": "none"
397
+ },
398
+ "encoder.block.14.layer.0.SelfAttention.k": {
399
+ "weights": "qint8",
400
+ "activations": "none"
401
+ },
402
+ "encoder.block.14.layer.0.SelfAttention.v": {
403
+ "weights": "qint8",
404
+ "activations": "none"
405
+ },
406
+ "encoder.block.14.layer.0.SelfAttention.o": {
407
+ "weights": "qint8",
408
+ "activations": "none"
409
+ },
410
+ "encoder.block.14.layer.1.DenseReluDense.wi_0": {
411
+ "weights": "qint8",
412
+ "activations": "none"
413
+ },
414
+ "encoder.block.14.layer.1.DenseReluDense.wi_1": {
415
+ "weights": "qint8",
416
+ "activations": "none"
417
+ },
418
+ "encoder.block.14.layer.1.DenseReluDense.wo": {
419
+ "weights": "qint8",
420
+ "activations": "none"
421
+ },
422
+ "encoder.block.15.layer.0.SelfAttention.q": {
423
+ "weights": "qint8",
424
+ "activations": "none"
425
+ },
426
+ "encoder.block.15.layer.0.SelfAttention.k": {
427
+ "weights": "qint8",
428
+ "activations": "none"
429
+ },
430
+ "encoder.block.15.layer.0.SelfAttention.v": {
431
+ "weights": "qint8",
432
+ "activations": "none"
433
+ },
434
+ "encoder.block.15.layer.0.SelfAttention.o": {
435
+ "weights": "qint8",
436
+ "activations": "none"
437
+ },
438
+ "encoder.block.15.layer.1.DenseReluDense.wi_0": {
439
+ "weights": "qint8",
440
+ "activations": "none"
441
+ },
442
+ "encoder.block.15.layer.1.DenseReluDense.wi_1": {
443
+ "weights": "qint8",
444
+ "activations": "none"
445
+ },
446
+ "encoder.block.15.layer.1.DenseReluDense.wo": {
447
+ "weights": "qint8",
448
+ "activations": "none"
449
+ },
450
+ "encoder.block.16.layer.0.SelfAttention.q": {
451
+ "weights": "qint8",
452
+ "activations": "none"
453
+ },
454
+ "encoder.block.16.layer.0.SelfAttention.k": {
455
+ "weights": "qint8",
456
+ "activations": "none"
457
+ },
458
+ "encoder.block.16.layer.0.SelfAttention.v": {
459
+ "weights": "qint8",
460
+ "activations": "none"
461
+ },
462
+ "encoder.block.16.layer.0.SelfAttention.o": {
463
+ "weights": "qint8",
464
+ "activations": "none"
465
+ },
466
+ "encoder.block.16.layer.1.DenseReluDense.wi_0": {
467
+ "weights": "qint8",
468
+ "activations": "none"
469
+ },
470
+ "encoder.block.16.layer.1.DenseReluDense.wi_1": {
471
+ "weights": "qint8",
472
+ "activations": "none"
473
+ },
474
+ "encoder.block.16.layer.1.DenseReluDense.wo": {
475
+ "weights": "qint8",
476
+ "activations": "none"
477
+ },
478
+ "encoder.block.17.layer.0.SelfAttention.q": {
479
+ "weights": "qint8",
480
+ "activations": "none"
481
+ },
482
+ "encoder.block.17.layer.0.SelfAttention.k": {
483
+ "weights": "qint8",
484
+ "activations": "none"
485
+ },
486
+ "encoder.block.17.layer.0.SelfAttention.v": {
487
+ "weights": "qint8",
488
+ "activations": "none"
489
+ },
490
+ "encoder.block.17.layer.0.SelfAttention.o": {
491
+ "weights": "qint8",
492
+ "activations": "none"
493
+ },
494
+ "encoder.block.17.layer.1.DenseReluDense.wi_0": {
495
+ "weights": "qint8",
496
+ "activations": "none"
497
+ },
498
+ "encoder.block.17.layer.1.DenseReluDense.wi_1": {
499
+ "weights": "qint8",
500
+ "activations": "none"
501
+ },
502
+ "encoder.block.17.layer.1.DenseReluDense.wo": {
503
+ "weights": "qint8",
504
+ "activations": "none"
505
+ },
506
+ "encoder.block.18.layer.0.SelfAttention.q": {
507
+ "weights": "qint8",
508
+ "activations": "none"
509
+ },
510
+ "encoder.block.18.layer.0.SelfAttention.k": {
511
+ "weights": "qint8",
512
+ "activations": "none"
513
+ },
514
+ "encoder.block.18.layer.0.SelfAttention.v": {
515
+ "weights": "qint8",
516
+ "activations": "none"
517
+ },
518
+ "encoder.block.18.layer.0.SelfAttention.o": {
519
+ "weights": "qint8",
520
+ "activations": "none"
521
+ },
522
+ "encoder.block.18.layer.1.DenseReluDense.wi_0": {
523
+ "weights": "qint8",
524
+ "activations": "none"
525
+ },
526
+ "encoder.block.18.layer.1.DenseReluDense.wi_1": {
527
+ "weights": "qint8",
528
+ "activations": "none"
529
+ },
530
+ "encoder.block.18.layer.1.DenseReluDense.wo": {
531
+ "weights": "qint8",
532
+ "activations": "none"
533
+ },
534
+ "encoder.block.19.layer.0.SelfAttention.q": {
535
+ "weights": "qint8",
536
+ "activations": "none"
537
+ },
538
+ "encoder.block.19.layer.0.SelfAttention.k": {
539
+ "weights": "qint8",
540
+ "activations": "none"
541
+ },
542
+ "encoder.block.19.layer.0.SelfAttention.v": {
543
+ "weights": "qint8",
544
+ "activations": "none"
545
+ },
546
+ "encoder.block.19.layer.0.SelfAttention.o": {
547
+ "weights": "qint8",
548
+ "activations": "none"
549
+ },
550
+ "encoder.block.19.layer.1.DenseReluDense.wi_0": {
551
+ "weights": "qint8",
552
+ "activations": "none"
553
+ },
554
+ "encoder.block.19.layer.1.DenseReluDense.wi_1": {
555
+ "weights": "qint8",
556
+ "activations": "none"
557
+ },
558
+ "encoder.block.19.layer.1.DenseReluDense.wo": {
559
+ "weights": "qint8",
560
+ "activations": "none"
561
+ },
562
+ "encoder.block.20.layer.0.SelfAttention.q": {
563
+ "weights": "qint8",
564
+ "activations": "none"
565
+ },
566
+ "encoder.block.20.layer.0.SelfAttention.k": {
567
+ "weights": "qint8",
568
+ "activations": "none"
569
+ },
570
+ "encoder.block.20.layer.0.SelfAttention.v": {
571
+ "weights": "qint8",
572
+ "activations": "none"
573
+ },
574
+ "encoder.block.20.layer.0.SelfAttention.o": {
575
+ "weights": "qint8",
576
+ "activations": "none"
577
+ },
578
+ "encoder.block.20.layer.1.DenseReluDense.wi_0": {
579
+ "weights": "qint8",
580
+ "activations": "none"
581
+ },
582
+ "encoder.block.20.layer.1.DenseReluDense.wi_1": {
583
+ "weights": "qint8",
584
+ "activations": "none"
585
+ },
586
+ "encoder.block.20.layer.1.DenseReluDense.wo": {
587
+ "weights": "qint8",
588
+ "activations": "none"
589
+ },
590
+ "encoder.block.21.layer.0.SelfAttention.q": {
591
+ "weights": "qint8",
592
+ "activations": "none"
593
+ },
594
+ "encoder.block.21.layer.0.SelfAttention.k": {
595
+ "weights": "qint8",
596
+ "activations": "none"
597
+ },
598
+ "encoder.block.21.layer.0.SelfAttention.v": {
599
+ "weights": "qint8",
600
+ "activations": "none"
601
+ },
602
+ "encoder.block.21.layer.0.SelfAttention.o": {
603
+ "weights": "qint8",
604
+ "activations": "none"
605
+ },
606
+ "encoder.block.21.layer.1.DenseReluDense.wi_0": {
607
+ "weights": "qint8",
608
+ "activations": "none"
609
+ },
610
+ "encoder.block.21.layer.1.DenseReluDense.wi_1": {
611
+ "weights": "qint8",
612
+ "activations": "none"
613
+ },
614
+ "encoder.block.21.layer.1.DenseReluDense.wo": {
615
+ "weights": "qint8",
616
+ "activations": "none"
617
+ },
618
+ "encoder.block.22.layer.0.SelfAttention.q": {
619
+ "weights": "qint8",
620
+ "activations": "none"
621
+ },
622
+ "encoder.block.22.layer.0.SelfAttention.k": {
623
+ "weights": "qint8",
624
+ "activations": "none"
625
+ },
626
+ "encoder.block.22.layer.0.SelfAttention.v": {
627
+ "weights": "qint8",
628
+ "activations": "none"
629
+ },
630
+ "encoder.block.22.layer.0.SelfAttention.o": {
631
+ "weights": "qint8",
632
+ "activations": "none"
633
+ },
634
+ "encoder.block.22.layer.1.DenseReluDense.wi_0": {
635
+ "weights": "qint8",
636
+ "activations": "none"
637
+ },
638
+ "encoder.block.22.layer.1.DenseReluDense.wi_1": {
639
+ "weights": "qint8",
640
+ "activations": "none"
641
+ },
642
+ "encoder.block.22.layer.1.DenseReluDense.wo": {
643
+ "weights": "qint8",
644
+ "activations": "none"
645
+ },
646
+ "encoder.block.23.layer.0.SelfAttention.q": {
647
+ "weights": "qint8",
648
+ "activations": "none"
649
+ },
650
+ "encoder.block.23.layer.0.SelfAttention.k": {
651
+ "weights": "qint8",
652
+ "activations": "none"
653
+ },
654
+ "encoder.block.23.layer.0.SelfAttention.v": {
655
+ "weights": "qint8",
656
+ "activations": "none"
657
+ },
658
+ "encoder.block.23.layer.0.SelfAttention.o": {
659
+ "weights": "qint8",
660
+ "activations": "none"
661
+ },
662
+ "encoder.block.23.layer.1.DenseReluDense.wi_0": {
663
+ "weights": "qint8",
664
+ "activations": "none"
665
+ },
666
+ "encoder.block.23.layer.1.DenseReluDense.wi_1": {
667
+ "weights": "qint8",
668
+ "activations": "none"
669
+ },
670
+ "encoder.block.23.layer.1.DenseReluDense.wo": {
671
+ "weights": "qint8",
672
+ "activations": "none"
673
+ }
674
+ }
uv.lock ADDED
The diff for this file is too large to render. See raw diff