CharlieFRuan
commited on
Commit
•
995cb17
1
Parent(s):
8e129b2
Upload folder using huggingface_hub
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- .gitattributes +1 -0
- mlc-chat-config.json +83 -0
- ndarray-cache.json +0 -0
- params_shard_0.bin +3 -0
- params_shard_1.bin +3 -0
- params_shard_10.bin +3 -0
- params_shard_100.bin +3 -0
- params_shard_101.bin +3 -0
- params_shard_102.bin +3 -0
- params_shard_103.bin +3 -0
- params_shard_104.bin +3 -0
- params_shard_105.bin +3 -0
- params_shard_106.bin +3 -0
- params_shard_107.bin +3 -0
- params_shard_108.bin +3 -0
- params_shard_109.bin +3 -0
- params_shard_11.bin +3 -0
- params_shard_110.bin +3 -0
- params_shard_111.bin +3 -0
- params_shard_112.bin +3 -0
- params_shard_113.bin +3 -0
- params_shard_114.bin +3 -0
- params_shard_115.bin +3 -0
- params_shard_116.bin +3 -0
- params_shard_117.bin +3 -0
- params_shard_118.bin +3 -0
- params_shard_119.bin +3 -0
- params_shard_12.bin +3 -0
- params_shard_120.bin +3 -0
- params_shard_121.bin +3 -0
- params_shard_122.bin +3 -0
- params_shard_123.bin +3 -0
- params_shard_124.bin +3 -0
- params_shard_125.bin +3 -0
- params_shard_126.bin +3 -0
- params_shard_127.bin +3 -0
- params_shard_128.bin +3 -0
- params_shard_129.bin +3 -0
- params_shard_13.bin +3 -0
- params_shard_130.bin +3 -0
- params_shard_131.bin +3 -0
- params_shard_132.bin +3 -0
- params_shard_133.bin +3 -0
- params_shard_134.bin +3 -0
- params_shard_135.bin +3 -0
- params_shard_136.bin +3 -0
- params_shard_137.bin +3 -0
- params_shard_138.bin +3 -0
- params_shard_139.bin +3 -0
- params_shard_14.bin +3 -0
.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 |
+
tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
mlc-chat-config.json
ADDED
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"version": "0.1.0",
|
3 |
+
"model_type": "gemma2",
|
4 |
+
"quantization": "q0f16",
|
5 |
+
"model_config": {
|
6 |
+
"hidden_size": 4608,
|
7 |
+
"intermediate_size": 36864,
|
8 |
+
"attention_bias": false,
|
9 |
+
"num_attention_heads": 32,
|
10 |
+
"num_key_value_heads": 16,
|
11 |
+
"head_dim": 128,
|
12 |
+
"num_hidden_layers": 46,
|
13 |
+
"rms_norm_eps": 1e-06,
|
14 |
+
"vocab_size": 256000,
|
15 |
+
"hidden_activation": "gelu_pytorch_tanh",
|
16 |
+
"position_embedding_base": 10000.0,
|
17 |
+
"context_window_size": 4096,
|
18 |
+
"prefill_chunk_size": 2048,
|
19 |
+
"tensor_parallel_shards": 1,
|
20 |
+
"max_batch_size": 80,
|
21 |
+
"attn_logit_softcapping": 50.0,
|
22 |
+
"final_logit_softcapping": 30.0,
|
23 |
+
"query_pre_attn_scalar": 144,
|
24 |
+
"sliding_window": 4096
|
25 |
+
},
|
26 |
+
"vocab_size": 256000,
|
27 |
+
"context_window_size": 4096,
|
28 |
+
"sliding_window_size": -1,
|
29 |
+
"prefill_chunk_size": 2048,
|
30 |
+
"attention_sink_size": -1,
|
31 |
+
"tensor_parallel_shards": 1,
|
32 |
+
"temperature": 1.0,
|
33 |
+
"presence_penalty": 0.0,
|
34 |
+
"frequency_penalty": 0.0,
|
35 |
+
"repetition_penalty": 1.0,
|
36 |
+
"top_p": 1.0,
|
37 |
+
"tokenizer_files": [
|
38 |
+
"tokenizer.model",
|
39 |
+
"tokenizer.json",
|
40 |
+
"tokenizer_config.json"
|
41 |
+
],
|
42 |
+
"tokenizer_info": {
|
43 |
+
"token_postproc_method": "byte_fallback",
|
44 |
+
"prepend_space_in_encode": false,
|
45 |
+
"strip_space_in_decode": false
|
46 |
+
},
|
47 |
+
"conv_template": {
|
48 |
+
"name": "gemma_instruction",
|
49 |
+
"system_template": "{system_message}",
|
50 |
+
"system_message": "",
|
51 |
+
"system_prefix_token_ids": [
|
52 |
+
2
|
53 |
+
],
|
54 |
+
"add_role_after_system_message": true,
|
55 |
+
"roles": {
|
56 |
+
"user": "<start_of_turn>user",
|
57 |
+
"assistant": "<start_of_turn>model"
|
58 |
+
},
|
59 |
+
"role_templates": {
|
60 |
+
"user": "{user_message}",
|
61 |
+
"assistant": "{assistant_message}",
|
62 |
+
"tool": "{tool_message}"
|
63 |
+
},
|
64 |
+
"messages": [],
|
65 |
+
"seps": [
|
66 |
+
"<end_of_turn>\n"
|
67 |
+
],
|
68 |
+
"role_content_sep": "\n",
|
69 |
+
"role_empty_sep": "\n",
|
70 |
+
"stop_str": [
|
71 |
+
"<end_of_turn>"
|
72 |
+
],
|
73 |
+
"stop_token_ids": [
|
74 |
+
1,
|
75 |
+
107
|
76 |
+
],
|
77 |
+
"function_string": "",
|
78 |
+
"use_function_calling": false
|
79 |
+
},
|
80 |
+
"pad_token_id": 0,
|
81 |
+
"bos_token_id": 2,
|
82 |
+
"eos_token_id": 1
|
83 |
+
}
|
ndarray-cache.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
params_shard_0.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ecaa755837342c5069038bb5054525e02744d3ecc35b5109f2be6ec3d152df1f
|
3 |
+
size 2359296000
|
params_shard_1.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:14d671c57e249ef013bb2136f6accdc8742e7367ecfa7c8b9cca83bce7708998
|
3 |
+
size 339738624
|
params_shard_10.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:bb54ac3fe9d341753e0cbcdc9c2528fc54652f80517ae6c4000cbd456f68f797
|
3 |
+
size 37748736
|
params_shard_100.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:076d647c503ae058200793c5e32daa9083874ead355f0fe14f6f8b016a37d953
|
3 |
+
size 37748736
|
params_shard_101.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:62aa92ed97cb853ed03827dc749df47cdebcdb55ab0aa9a0047f0eafad75c3a6
|
3 |
+
size 339738624
|
params_shard_102.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a8efe171b6009827ecda19248f8401ce81301772dce8cccf44e6a91f40e62664
|
3 |
+
size 679477248
|
params_shard_103.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4dd429f2c7befe252c4bd828f02ad65f841910e757e729748467b02e6f006305
|
3 |
+
size 75497472
|
params_shard_104.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5881e82b721ff399ad908d854acf53539c648171e6720a3a7eb28fb256aacb6f
|
3 |
+
size 37748736
|
params_shard_105.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:85deca34e0f233eed6cb439415ee549f2ea2c3c7e6cb45553e0d9d7f49271cb0
|
3 |
+
size 339738624
|
params_shard_106.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8366b8d0314da4d8a8519993076bba671d18479b65cfa1a73cf7969c60ea4ffa
|
3 |
+
size 679477248
|
params_shard_107.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f53bf4a48f761e9fe0f4da8b24dfef5eb03105a4a5b3df1e6a4a07aa8c39d866
|
3 |
+
size 75497472
|
params_shard_108.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9f3356eca8e15400fdab4abb1689a8800e1626d59430f3c8fcd7062d834cb3e8
|
3 |
+
size 37748736
|
params_shard_109.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b851241961ed31bd317ffb8d15eaa101c026b5c0ffe7748473e4ff23150861ea
|
3 |
+
size 339738624
|
params_shard_11.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6a73643dd56d9f775aab47dabb8c293a0030bc5b8160e9e67d78db516d72a5ce
|
3 |
+
size 339738624
|
params_shard_110.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:27138550c3d349472b3971d230aff44cce5ed4378fa1c3adafccee199982c8cc
|
3 |
+
size 679477248
|
params_shard_111.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:be7903fc48dfc16e5763c392d941097a2fa18547ce783cbf6a930e53a2660226
|
3 |
+
size 75497472
|
params_shard_112.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:123854a67f40a398121c4baf172b95337945a4b917e5a7018b8e00b99c19eb2e
|
3 |
+
size 37748736
|
params_shard_113.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:99269b8152eeb6b36175297e2e3603a517fc57d307a7e85658a308074cafafde
|
3 |
+
size 75497472
|
params_shard_114.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:29361e843cb11a33ea8e587cc46677536831bf4ca932c559b65bcada4f9c8f59
|
3 |
+
size 37748736
|
params_shard_115.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6677f0a55a895af5246354bf61936e077d155dae0d5cdc3fbe75e4718131bcc2
|
3 |
+
size 339738624
|
params_shard_116.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2ff826c0245003fdd46e2cbbbfe7980c0ad7d5a11adeadd5e7bc8c83e4ac5d7f
|
3 |
+
size 679477248
|
params_shard_117.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:eba4e94dc47467252d84fa115c081ac7a71f173e7a17de900fc22dfe6db065fe
|
3 |
+
size 339738624
|
params_shard_118.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:46939bf5e252db56ced273bcc51e166957617290d078331e22ce7b2e2a4b7276
|
3 |
+
size 679477248
|
params_shard_119.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:133d66efc7353a9d6dae8f51fed094db7b1ffc3409f87ec1fd0a2d60e92b894e
|
3 |
+
size 75497472
|
params_shard_12.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:941f2874ab37ab9b2ec2880deeb1b44290ae1db0be83daa9a6f72803c3c93061
|
3 |
+
size 339738624
|
params_shard_120.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:283512493abe797d500dfdb8e0cf024dbe68e14f8f7473b1215466001e41aadf
|
3 |
+
size 37748736
|
params_shard_121.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:91f05a00ea740d9ff1883c5442ddc2990e05760399ad2d5d17d29a06da50ddea
|
3 |
+
size 339738624
|
params_shard_122.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:473641ba16f348f1d459490b4c33e89776adc6461a71d550e42ffb6cc39ed5bc
|
3 |
+
size 679477248
|
params_shard_123.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4c6cf4aafe569d0216519598cbd728deb94c36106dcab95e2c0f65708922cc7b
|
3 |
+
size 75497472
|
params_shard_124.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:38a77fb0e2002dadf6e1e5343eac729771945c47a9f4895adfc53eef59e227e2
|
3 |
+
size 37748736
|
params_shard_125.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2d3399d324c7e0985ff4256379bd1cab7ce8114b1b48acf01dd57e6c9a51d015
|
3 |
+
size 339738624
|
params_shard_126.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3094b8c41bc6132bf3f7532886d6a9ba0f5d5ab04f383d0a6ee3e263c7d06bf8
|
3 |
+
size 679477248
|
params_shard_127.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d45df021550c221753d1d0628bdc31ec4260fb928ec2f91748c9a879872f14ab
|
3 |
+
size 75497472
|
params_shard_128.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c2b838dd62f14dd5a05c7bb824666da2a88edec3c9369ef4381ae73ef077d046
|
3 |
+
size 37748736
|
params_shard_129.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:dba8c4754964cbc59fa8d2a71d672140dc5371ab1446d159e3567bcc67c234e2
|
3 |
+
size 679477248
|
params_shard_13.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7eadc8293e643c9b9406c4efdcc9203fa1d34ca1ca43cb3879786bb16322b1ff
|
3 |
+
size 679477248
|
params_shard_130.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:efca8ee2862c483f29bac6b05ae9397a73e3739d5d5589fe38379f9ef574d918
|
3 |
+
size 75497472
|
params_shard_131.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:98f78d32035b441a1a32af41d633456837cd39d81f78034077a0f3348d361e43
|
3 |
+
size 37748736
|
params_shard_132.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1d6d5339e8be94409ec67479876b823deba6898ba0f8138f7e8b8971e5293638
|
3 |
+
size 339738624
|
params_shard_133.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4a520ecbcf82789087b8df3c33e1ce4766aa5f4b159081e0cb959db336319a51
|
3 |
+
size 339738624
|
params_shard_134.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d65a166b8ac7a93171b6b9893fe6805a106e78a6968167d25653115e478b7432
|
3 |
+
size 679477248
|
params_shard_135.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:cf66535f3e36f427f73d68064551ca26513417f49ef59f54a9cfb539978e4690
|
3 |
+
size 75497472
|
params_shard_136.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0891e3b6752084b80c3391aeb065d91c20ed31d0314cc0430281f1f17bc74fbd
|
3 |
+
size 37748736
|
params_shard_137.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8dd1d57b1a92b0863baa747133d50c1e53ec6e67e165aea37ab0fc287b16f16c
|
3 |
+
size 339738624
|
params_shard_138.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:62ff833d2cecf7557bf9ec2a8c5c9a0cf6f734e564ec6cc9ac5f9c33ef743a46
|
3 |
+
size 679477248
|
params_shard_139.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2849307db6413641f27e967726dddeee41913f015f64694b4610c97ed0cb46f9
|
3 |
+
size 75497472
|
params_shard_14.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:29efc3743754155f6abc6ac884c813bd76b15d6ea0d58293c1050a8187960def
|
3 |
+
size 75497472
|