faxenoff commited on
Commit
0807aba
·
verified ·
1 Parent(s): e669757

Add compiled engines: TRT win sm_120 + OV (cpu/igpu/npu) + TVM win vulkan + tokenizer

Browse files
.gitattributes CHANGED
@@ -33,3 +33,9 @@ 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
+ code-daemon-embed-v1-l_b256_s128_win_x64_tvm_vulkan.dll filter=lfs diff=lfs merge=lfs -text
37
+ code-daemon-embed-v1-l_win_x64_trt_sm_120.engine filter=lfs diff=lfs merge=lfs -text
38
+ code-daemon-embed-v1-m_b128_s64_win_x64_tvm_vulkan.dll filter=lfs diff=lfs merge=lfs -text
39
+ code-daemon-embed-v1-m_win_x64_trt_sm_120.engine filter=lfs diff=lfs merge=lfs -text
40
+ code-daemon-embed-v1-s_b64_s40_win_x64_tvm_vulkan.dll filter=lfs diff=lfs merge=lfs -text
41
+ code-daemon-embed-v1-s_win_x64_trt_sm_120.engine filter=lfs diff=lfs merge=lfs -text
code-daemon-embed-v1-l_b256_s128_win_x64_tvm_vulkan.dll ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d12c6e51baa8755d1cc6e394e2221633591f72e0569ee14e236df5d29163e3c
3
+ size 307570176
code-daemon-embed-v1-l_ov_cpu_int8_b256_s128.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7ef4c484617bd28af132c6e204b2066776555ef7f4d02df538b0514d4e20051
3
+ size 54376048
code-daemon-embed-v1-l_ov_cpu_int8_b256_s128.xml ADDED
The diff for this file is too large to render. See raw diff
 
code-daemon-embed-v1-l_ov_igpu_lnl_int8_b256_s128.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d4f2b2431ae840d89bcf335f4c376f6e3a571050721ea4d431719cddc0b4667
3
+ size 40429525
code-daemon-embed-v1-l_ov_igpu_lnl_int8_b256_s128.xml ADDED
The diff for this file is too large to render. See raw diff
 
code-daemon-embed-v1-l_ov_npu_fp16_b16_s128.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d4f2b2431ae840d89bcf335f4c376f6e3a571050721ea4d431719cddc0b4667
3
+ size 40429525
code-daemon-embed-v1-l_ov_npu_fp16_b16_s128.xml ADDED
The diff for this file is too large to render. See raw diff
 
code-daemon-embed-v1-l_win_x64_trt_sm_120.engine ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d2572f0bdb12e263c4ff32244a90a5935c96ddb428addb029eddc2e5ff0b0cb5
3
+ size 82864268
code-daemon-embed-v1-m_b128_s64_win_x64_tvm_vulkan.dll ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8b208a0d6b173d34f8af5abad8654da7f345c0bce8ed4acc7d6bca5a349fafdf
3
+ size 232067584
code-daemon-embed-v1-m_ov_cpu_int8_b128_s64.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b52c06c810371009957e204de25d772c7bcf32fa3b84351b607cd812c372213b
3
+ size 54376048
code-daemon-embed-v1-m_ov_cpu_int8_b128_s64.xml ADDED
The diff for this file is too large to render. See raw diff
 
code-daemon-embed-v1-m_ov_igpu_lnl_int8_b128_s64.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d4f2b2431ae840d89bcf335f4c376f6e3a571050721ea4d431719cddc0b4667
3
+ size 40429525
code-daemon-embed-v1-m_ov_igpu_lnl_int8_b128_s64.xml ADDED
The diff for this file is too large to render. See raw diff
 
code-daemon-embed-v1-m_ov_npu_fp16_b16_s64.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d4f2b2431ae840d89bcf335f4c376f6e3a571050721ea4d431719cddc0b4667
3
+ size 40429525
code-daemon-embed-v1-m_ov_npu_fp16_b16_s64.xml ADDED
The diff for this file is too large to render. See raw diff
 
code-daemon-embed-v1-m_win_x64_trt_sm_120.engine ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b8121bc2476c71c0b442a48e1d371a7a0ba14a1daaf82497eff933056c94872
3
+ size 83350636
code-daemon-embed-v1-s_b64_s40_win_x64_tvm_vulkan.dll ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4825220718089ffc1a8b6e56f269c73f4b171f44ff9dc43665dc9930e8c2827
3
+ size 214774272
code-daemon-embed-v1-s_ov_cpu_int8_b64_s40.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33f49c7fb524a45acad9fa5876555ae2665bbf10b1b5056a504330101b0cd2da
3
+ size 54376048
code-daemon-embed-v1-s_ov_cpu_int8_b64_s40.xml ADDED
The diff for this file is too large to render. See raw diff
 
code-daemon-embed-v1-s_ov_igpu_lnl_int8_b64_s40.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d4f2b2431ae840d89bcf335f4c376f6e3a571050721ea4d431719cddc0b4667
3
+ size 40429525
code-daemon-embed-v1-s_ov_igpu_lnl_int8_b64_s40.xml ADDED
The diff for this file is too large to render. See raw diff
 
code-daemon-embed-v1-s_ov_npu_fp16_b16_s40.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d4f2b2431ae840d89bcf335f4c376f6e3a571050721ea4d431719cddc0b4667
3
+ size 40429525
code-daemon-embed-v1-s_ov_npu_fp16_b16_s40.xml ADDED
The diff for this file is too large to render. See raw diff
 
code-daemon-embed-v1-s_win_x64_trt_sm_120.engine ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:591ea466c30eadb27f7c7497cec3f36893331fca8a21b448d12153ebee3e7e76
3
+ size 83705740
config.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_cross_attention": false,
3
+ "attention_probs_dropout_prob": 0.1,
4
+ "bos_token_id": 0,
5
+ "classifier_dropout": null,
6
+ "eos_token_id": 2,
7
+ "hidden_act": "gelu",
8
+ "hidden_dropout_prob": 0.1,
9
+ "hidden_size": 768,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 3072,
12
+ "is_decoder": false,
13
+ "layer_norm_eps": 1e-05,
14
+ "max_position_embeddings": 514,
15
+ "model_type": "xlm-roberta",
16
+ "num_attention_heads": 12,
17
+ "num_hidden_layers": 4,
18
+ "pad_token_id": 1,
19
+ "tie_word_embeddings": true,
20
+ "transformers_version": "5.1.0",
21
+ "type_vocab_size": 1,
22
+ "use_cache": true,
23
+ "vocab_size": 32000
24
+ }
manifest.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "compiled_at": "2026-06-22T13:50:43Z",
3
+ "targets": [
4
+ "gpu-vulkan"
5
+ ],
6
+ "quantization": "int8",
7
+ "max_tokens": 128,
8
+ "dimension": 768,
9
+ "model_id": "code-daemon-embed-v1"
10
+ }
sentencepiece.bpe.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d06858a769b0232f511fad2a3d2bcb3bd36b96789a7c517e2657cd00b920cab9
3
+ size 752712
tokenizer_config.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "tokenizer_class": "XLMRobertaTokenizer",
3
+ "model_max_length": 512,
4
+ "bos_token": "<s>",
5
+ "eos_token": "</s>",
6
+ "unk_token": "<unk>",
7
+ "pad_token": "<pad>"
8
+ }