Text Generation
Transformers
PyTorch
Safetensors
Arabic
English
llama
conversational
text-generation-inference
Inference Endpoints

Adding `safetensors` variant of this model

#2
model-01-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7a4e0d39323b535f9e5df9ae7da26fc22ad32aab1d56319c72a8814328eb2d2
3
+ size 3892349240
model-02-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ecc001886e2661e1d5472971ccce7c99c5d1df5c885b854ca7d571334882efd
3
+ size 4194405952
model-03-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7e03baa7703b5fa42ed1f7c65806d9e045d675232c0280c320573ec006507ae8
3
+ size 3892382352
model-04-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e0c32f870d1fc482e0127ea61ad35fcff5c1e3cc12b4715e72887404925b614
3
+ size 3892382352
model-05-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ff15f508b4803492b7fdaa5fe0333874f0965a2084c4856cab58d0794fb04e8
3
+ size 4194405968
model-06-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:644a31e2c2338a65e0799ba404410a85b4e0a5bb0fc77af9db2a431d2b32d993
3
+ size 3892382376
model-07-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:260b4d855ffb2b2a643db2e0b56ae7d8c2e2e1300ebe6e8a6ee76c1a19c639e1
3
+ size 3892382368
model-08-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a936cd6500d7fbc62373953c3466a9b161f8e4244184cec6ed4141da5fae3f02
3
+ size 4194405976
model-09-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d759a9801a6b2612099a3e9f640d41a2024cff973e3c5888e9582942d873b83
3
+ size 3892382376
model-10-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:82cfc1a462eb1df3edb3a4d1e570c81c60e85411574fa93c450160de1bbe5be9
3
+ size 3892382368
model-11-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c8e206281c1ac7e78f2bf2a68232e6abae212e6eb2f4d6cde7fc599fcc675b07
3
+ size 4194405976
model-12-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91c0ab0711c9eefe6c25858b59b2a75454f8401e40facb889dcc54d42224f803
3
+ size 3892382376
model-13-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a4a27b21b8d177bdda23dfb39734a02aa9ee0fc88e21b0fea80f2cab2c1e6bb
3
+ size 3892382368
model-14-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b9a3da88f8ea2645370821d71b387e34a1fc8052febbf4ea41890474ffe1ca9
3
+ size 4194405976
model-15-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c6a400355dcfc030246ea5e978c8a02a57f8d1dcccb4237f0f1187d41cbb0ea8
3
+ size 3892382376
model-16-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34b3e9f61716fc2bb04ca58611d91afab54e8ddf76cf21cd0552bda58343c942
3
+ size 3892382368
model-17-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bfdabd8defc002213f8fb41b88d13e45feb42595625e0d4764b2cd0cf87f0466
3
+ size 4194405976
model-18-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57c47174532f926ad2a48e9ac40be9f21e0f0224f2bc6c00e9098dbdae0140e4
3
+ size 3892382376
model-19-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91390e1a7bd99370bc9e1c9c24fb11596852b5c853f1fc15fa54dbd0bd0bd0b4
3
+ size 3892382368
model-20-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0d18ba07ad359cfc95544decc98b85d3e3500dc8369bd3c0a25eaf4cd694fe6
3
+ size 4194405976
model-21-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:678a925ce8c6b08641a78251658fcb51f41cc690744a93b24cda1d04f47a1446
3
+ size 3892382376
model-22-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0b8ffd3967280f1e6690bcd40157e4dee8d6b196fdfe32d33299303f3365cae
3
+ size 3892382368
model-23-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:576ee81f9f3e3e57e9a5c5c51d0375d8a9eeadfc2a4aebf07ee623858251bfd3
3
+ size 4194405976
model-24-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc3b6edd7729faad7e941099c523b9ac2da44fdecc6cc89f6b8cb551dcbdd880
3
+ size 3892382376
model-25-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ebdaff3d2bb373dda80227235b7965aa19a2c1b9e1ad644a15be0aa0893518a
3
+ size 3892382368
model-26-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8dd2570443f7c2859e150b58c9c29b73c6ffbd43a22c58a6a8f2b5778b4b4afb
3
+ size 4194405976
model-27-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8706ec18cdc9c6cc68cdd7b2d7e04531c586a24173caa09dba8604d487699a1e
3
+ size 3892382376
model-28-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:876b216c9d9c609a27f54864cc30e6f414c4491c581ca3af590f3dbdad123cab
3
+ size 3892382368
model-29-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6159524dd54a423069c411506782c251002e1e798b023020b5b32f227cf5c96
3
+ size 4194405976
model-30-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0300e356802264a50d50594793197b39cf466b228277778387709ab56a0812e5
3
+ size 3892382376
model-31-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4be735f4225d8ebb74bf867cad5ae238362c89d5b54b75175f37a2e08d5a0c03
3
+ size 3892382368
model-32-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0cb12c5bb32bc0e0e532c71645fd07bd7cbcb7215f23746f0ec5f2faa63d16ca
3
+ size 4194405976
model-33-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e6a6d6a87c7204af371dfbc0e617b22fa5f936eff4e1baecc9de62a425301a0
3
+ size 3892382376
model-34-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6111f29d61e630667154bc67ebe540fabdb182758b2746f97926be36559b338b
3
+ size 3892382368
model-35-of-35.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:202ae56f54667a072ae30dbf09d61e9e3cd8305b804c3d4048b6c1f65db13254
3
+ size 3120645960
model.safetensors.index.json ADDED
@@ -0,0 +1,808 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {},
3
+ "weight_map": {
4
+ "model.embed_tokens.weight": "model-01-of-35.safetensors",
5
+ "model.layers.0.self_attn.q_proj.weight": "model-01-of-35.safetensors",
6
+ "model.layers.0.self_attn.k_proj.weight": "model-01-of-35.safetensors",
7
+ "model.layers.0.self_attn.v_proj.weight": "model-01-of-35.safetensors",
8
+ "model.layers.0.self_attn.o_proj.weight": "model-01-of-35.safetensors",
9
+ "model.layers.0.self_attn.rotary_emb.inv_freq": "model-01-of-35.safetensors",
10
+ "model.layers.0.mlp.gate_proj.weight": "model-01-of-35.safetensors",
11
+ "model.layers.0.mlp.up_proj.weight": "model-01-of-35.safetensors",
12
+ "model.layers.0.mlp.down_proj.weight": "model-01-of-35.safetensors",
13
+ "model.layers.0.input_layernorm.weight": "model-01-of-35.safetensors",
14
+ "model.layers.0.post_attention_layernorm.weight": "model-01-of-35.safetensors",
15
+ "model.layers.1.self_attn.q_proj.weight": "model-01-of-35.safetensors",
16
+ "model.layers.1.self_attn.k_proj.weight": "model-01-of-35.safetensors",
17
+ "model.layers.1.self_attn.v_proj.weight": "model-01-of-35.safetensors",
18
+ "model.layers.1.self_attn.o_proj.weight": "model-01-of-35.safetensors",
19
+ "model.layers.1.self_attn.rotary_emb.inv_freq": "model-01-of-35.safetensors",
20
+ "model.layers.1.mlp.gate_proj.weight": "model-01-of-35.safetensors",
21
+ "model.layers.1.mlp.up_proj.weight": "model-01-of-35.safetensors",
22
+ "model.layers.1.mlp.down_proj.weight": "model-02-of-35.safetensors",
23
+ "model.layers.1.input_layernorm.weight": "model-02-of-35.safetensors",
24
+ "model.layers.1.post_attention_layernorm.weight": "model-02-of-35.safetensors",
25
+ "model.layers.2.self_attn.q_proj.weight": "model-02-of-35.safetensors",
26
+ "model.layers.2.self_attn.k_proj.weight": "model-02-of-35.safetensors",
27
+ "model.layers.2.self_attn.v_proj.weight": "model-02-of-35.safetensors",
28
+ "model.layers.2.self_attn.o_proj.weight": "model-02-of-35.safetensors",
29
+ "model.layers.2.self_attn.rotary_emb.inv_freq": "model-02-of-35.safetensors",
30
+ "model.layers.2.mlp.gate_proj.weight": "model-02-of-35.safetensors",
31
+ "model.layers.2.mlp.up_proj.weight": "model-02-of-35.safetensors",
32
+ "model.layers.2.mlp.down_proj.weight": "model-02-of-35.safetensors",
33
+ "model.layers.2.input_layernorm.weight": "model-02-of-35.safetensors",
34
+ "model.layers.2.post_attention_layernorm.weight": "model-02-of-35.safetensors",
35
+ "model.layers.3.self_attn.q_proj.weight": "model-02-of-35.safetensors",
36
+ "model.layers.3.self_attn.k_proj.weight": "model-02-of-35.safetensors",
37
+ "model.layers.3.self_attn.v_proj.weight": "model-02-of-35.safetensors",
38
+ "model.layers.3.self_attn.o_proj.weight": "model-02-of-35.safetensors",
39
+ "model.layers.3.self_attn.rotary_emb.inv_freq": "model-02-of-35.safetensors",
40
+ "model.layers.3.mlp.gate_proj.weight": "model-02-of-35.safetensors",
41
+ "model.layers.3.mlp.up_proj.weight": "model-02-of-35.safetensors",
42
+ "model.layers.3.mlp.down_proj.weight": "model-02-of-35.safetensors",
43
+ "model.layers.3.input_layernorm.weight": "model-02-of-35.safetensors",
44
+ "model.layers.3.post_attention_layernorm.weight": "model-02-of-35.safetensors",
45
+ "model.layers.4.self_attn.q_proj.weight": "model-02-of-35.safetensors",
46
+ "model.layers.4.self_attn.k_proj.weight": "model-02-of-35.safetensors",
47
+ "model.layers.4.self_attn.v_proj.weight": "model-02-of-35.safetensors",
48
+ "model.layers.4.self_attn.o_proj.weight": "model-02-of-35.safetensors",
49
+ "model.layers.4.self_attn.rotary_emb.inv_freq": "model-02-of-35.safetensors",
50
+ "model.layers.4.mlp.gate_proj.weight": "model-03-of-35.safetensors",
51
+ "model.layers.4.mlp.up_proj.weight": "model-03-of-35.safetensors",
52
+ "model.layers.4.mlp.down_proj.weight": "model-03-of-35.safetensors",
53
+ "model.layers.4.input_layernorm.weight": "model-03-of-35.safetensors",
54
+ "model.layers.4.post_attention_layernorm.weight": "model-03-of-35.safetensors",
55
+ "model.layers.5.self_attn.q_proj.weight": "model-03-of-35.safetensors",
56
+ "model.layers.5.self_attn.k_proj.weight": "model-03-of-35.safetensors",
57
+ "model.layers.5.self_attn.v_proj.weight": "model-03-of-35.safetensors",
58
+ "model.layers.5.self_attn.o_proj.weight": "model-03-of-35.safetensors",
59
+ "model.layers.5.self_attn.rotary_emb.inv_freq": "model-03-of-35.safetensors",
60
+ "model.layers.5.mlp.gate_proj.weight": "model-03-of-35.safetensors",
61
+ "model.layers.5.mlp.up_proj.weight": "model-03-of-35.safetensors",
62
+ "model.layers.5.mlp.down_proj.weight": "model-03-of-35.safetensors",
63
+ "model.layers.5.input_layernorm.weight": "model-03-of-35.safetensors",
64
+ "model.layers.5.post_attention_layernorm.weight": "model-03-of-35.safetensors",
65
+ "model.layers.6.self_attn.q_proj.weight": "model-03-of-35.safetensors",
66
+ "model.layers.6.self_attn.k_proj.weight": "model-03-of-35.safetensors",
67
+ "model.layers.6.self_attn.v_proj.weight": "model-03-of-35.safetensors",
68
+ "model.layers.6.self_attn.o_proj.weight": "model-03-of-35.safetensors",
69
+ "model.layers.6.self_attn.rotary_emb.inv_freq": "model-03-of-35.safetensors",
70
+ "model.layers.6.mlp.gate_proj.weight": "model-03-of-35.safetensors",
71
+ "model.layers.6.mlp.up_proj.weight": "model-04-of-35.safetensors",
72
+ "model.layers.6.mlp.down_proj.weight": "model-04-of-35.safetensors",
73
+ "model.layers.6.input_layernorm.weight": "model-04-of-35.safetensors",
74
+ "model.layers.6.post_attention_layernorm.weight": "model-04-of-35.safetensors",
75
+ "model.layers.7.self_attn.q_proj.weight": "model-04-of-35.safetensors",
76
+ "model.layers.7.self_attn.k_proj.weight": "model-04-of-35.safetensors",
77
+ "model.layers.7.self_attn.v_proj.weight": "model-04-of-35.safetensors",
78
+ "model.layers.7.self_attn.o_proj.weight": "model-04-of-35.safetensors",
79
+ "model.layers.7.self_attn.rotary_emb.inv_freq": "model-04-of-35.safetensors",
80
+ "model.layers.7.mlp.gate_proj.weight": "model-04-of-35.safetensors",
81
+ "model.layers.7.mlp.up_proj.weight": "model-04-of-35.safetensors",
82
+ "model.layers.7.mlp.down_proj.weight": "model-04-of-35.safetensors",
83
+ "model.layers.7.input_layernorm.weight": "model-04-of-35.safetensors",
84
+ "model.layers.7.post_attention_layernorm.weight": "model-04-of-35.safetensors",
85
+ "model.layers.8.self_attn.q_proj.weight": "model-04-of-35.safetensors",
86
+ "model.layers.8.self_attn.k_proj.weight": "model-04-of-35.safetensors",
87
+ "model.layers.8.self_attn.v_proj.weight": "model-04-of-35.safetensors",
88
+ "model.layers.8.self_attn.o_proj.weight": "model-04-of-35.safetensors",
89
+ "model.layers.8.self_attn.rotary_emb.inv_freq": "model-04-of-35.safetensors",
90
+ "model.layers.8.mlp.gate_proj.weight": "model-04-of-35.safetensors",
91
+ "model.layers.8.mlp.up_proj.weight": "model-04-of-35.safetensors",
92
+ "model.layers.8.mlp.down_proj.weight": "model-05-of-35.safetensors",
93
+ "model.layers.8.input_layernorm.weight": "model-05-of-35.safetensors",
94
+ "model.layers.8.post_attention_layernorm.weight": "model-05-of-35.safetensors",
95
+ "model.layers.9.self_attn.q_proj.weight": "model-05-of-35.safetensors",
96
+ "model.layers.9.self_attn.k_proj.weight": "model-05-of-35.safetensors",
97
+ "model.layers.9.self_attn.v_proj.weight": "model-05-of-35.safetensors",
98
+ "model.layers.9.self_attn.o_proj.weight": "model-05-of-35.safetensors",
99
+ "model.layers.9.self_attn.rotary_emb.inv_freq": "model-05-of-35.safetensors",
100
+ "model.layers.9.mlp.gate_proj.weight": "model-05-of-35.safetensors",
101
+ "model.layers.9.mlp.up_proj.weight": "model-05-of-35.safetensors",
102
+ "model.layers.9.mlp.down_proj.weight": "model-05-of-35.safetensors",
103
+ "model.layers.9.input_layernorm.weight": "model-05-of-35.safetensors",
104
+ "model.layers.9.post_attention_layernorm.weight": "model-05-of-35.safetensors",
105
+ "model.layers.10.self_attn.q_proj.weight": "model-05-of-35.safetensors",
106
+ "model.layers.10.self_attn.k_proj.weight": "model-05-of-35.safetensors",
107
+ "model.layers.10.self_attn.v_proj.weight": "model-05-of-35.safetensors",
108
+ "model.layers.10.self_attn.o_proj.weight": "model-05-of-35.safetensors",
109
+ "model.layers.10.self_attn.rotary_emb.inv_freq": "model-05-of-35.safetensors",
110
+ "model.layers.10.mlp.gate_proj.weight": "model-05-of-35.safetensors",
111
+ "model.layers.10.mlp.up_proj.weight": "model-05-of-35.safetensors",
112
+ "model.layers.10.mlp.down_proj.weight": "model-05-of-35.safetensors",
113
+ "model.layers.10.input_layernorm.weight": "model-05-of-35.safetensors",
114
+ "model.layers.10.post_attention_layernorm.weight": "model-05-of-35.safetensors",
115
+ "model.layers.11.self_attn.q_proj.weight": "model-05-of-35.safetensors",
116
+ "model.layers.11.self_attn.k_proj.weight": "model-05-of-35.safetensors",
117
+ "model.layers.11.self_attn.v_proj.weight": "model-05-of-35.safetensors",
118
+ "model.layers.11.self_attn.o_proj.weight": "model-05-of-35.safetensors",
119
+ "model.layers.11.self_attn.rotary_emb.inv_freq": "model-05-of-35.safetensors",
120
+ "model.layers.11.mlp.gate_proj.weight": "model-06-of-35.safetensors",
121
+ "model.layers.11.mlp.up_proj.weight": "model-06-of-35.safetensors",
122
+ "model.layers.11.mlp.down_proj.weight": "model-06-of-35.safetensors",
123
+ "model.layers.11.input_layernorm.weight": "model-06-of-35.safetensors",
124
+ "model.layers.11.post_attention_layernorm.weight": "model-06-of-35.safetensors",
125
+ "model.layers.12.self_attn.q_proj.weight": "model-06-of-35.safetensors",
126
+ "model.layers.12.self_attn.k_proj.weight": "model-06-of-35.safetensors",
127
+ "model.layers.12.self_attn.v_proj.weight": "model-06-of-35.safetensors",
128
+ "model.layers.12.self_attn.o_proj.weight": "model-06-of-35.safetensors",
129
+ "model.layers.12.self_attn.rotary_emb.inv_freq": "model-06-of-35.safetensors",
130
+ "model.layers.12.mlp.gate_proj.weight": "model-06-of-35.safetensors",
131
+ "model.layers.12.mlp.up_proj.weight": "model-06-of-35.safetensors",
132
+ "model.layers.12.mlp.down_proj.weight": "model-06-of-35.safetensors",
133
+ "model.layers.12.input_layernorm.weight": "model-06-of-35.safetensors",
134
+ "model.layers.12.post_attention_layernorm.weight": "model-06-of-35.safetensors",
135
+ "model.layers.13.self_attn.q_proj.weight": "model-06-of-35.safetensors",
136
+ "model.layers.13.self_attn.k_proj.weight": "model-06-of-35.safetensors",
137
+ "model.layers.13.self_attn.v_proj.weight": "model-06-of-35.safetensors",
138
+ "model.layers.13.self_attn.o_proj.weight": "model-06-of-35.safetensors",
139
+ "model.layers.13.self_attn.rotary_emb.inv_freq": "model-06-of-35.safetensors",
140
+ "model.layers.13.mlp.gate_proj.weight": "model-06-of-35.safetensors",
141
+ "model.layers.13.mlp.up_proj.weight": "model-07-of-35.safetensors",
142
+ "model.layers.13.mlp.down_proj.weight": "model-07-of-35.safetensors",
143
+ "model.layers.13.input_layernorm.weight": "model-07-of-35.safetensors",
144
+ "model.layers.13.post_attention_layernorm.weight": "model-07-of-35.safetensors",
145
+ "model.layers.14.self_attn.q_proj.weight": "model-07-of-35.safetensors",
146
+ "model.layers.14.self_attn.k_proj.weight": "model-07-of-35.safetensors",
147
+ "model.layers.14.self_attn.v_proj.weight": "model-07-of-35.safetensors",
148
+ "model.layers.14.self_attn.o_proj.weight": "model-07-of-35.safetensors",
149
+ "model.layers.14.self_attn.rotary_emb.inv_freq": "model-07-of-35.safetensors",
150
+ "model.layers.14.mlp.gate_proj.weight": "model-07-of-35.safetensors",
151
+ "model.layers.14.mlp.up_proj.weight": "model-07-of-35.safetensors",
152
+ "model.layers.14.mlp.down_proj.weight": "model-07-of-35.safetensors",
153
+ "model.layers.14.input_layernorm.weight": "model-07-of-35.safetensors",
154
+ "model.layers.14.post_attention_layernorm.weight": "model-07-of-35.safetensors",
155
+ "model.layers.15.self_attn.q_proj.weight": "model-07-of-35.safetensors",
156
+ "model.layers.15.self_attn.k_proj.weight": "model-07-of-35.safetensors",
157
+ "model.layers.15.self_attn.v_proj.weight": "model-07-of-35.safetensors",
158
+ "model.layers.15.self_attn.o_proj.weight": "model-07-of-35.safetensors",
159
+ "model.layers.15.self_attn.rotary_emb.inv_freq": "model-07-of-35.safetensors",
160
+ "model.layers.15.mlp.gate_proj.weight": "model-07-of-35.safetensors",
161
+ "model.layers.15.mlp.up_proj.weight": "model-07-of-35.safetensors",
162
+ "model.layers.15.mlp.down_proj.weight": "model-08-of-35.safetensors",
163
+ "model.layers.15.input_layernorm.weight": "model-08-of-35.safetensors",
164
+ "model.layers.15.post_attention_layernorm.weight": "model-08-of-35.safetensors",
165
+ "model.layers.16.self_attn.q_proj.weight": "model-08-of-35.safetensors",
166
+ "model.layers.16.self_attn.k_proj.weight": "model-08-of-35.safetensors",
167
+ "model.layers.16.self_attn.v_proj.weight": "model-08-of-35.safetensors",
168
+ "model.layers.16.self_attn.o_proj.weight": "model-08-of-35.safetensors",
169
+ "model.layers.16.self_attn.rotary_emb.inv_freq": "model-08-of-35.safetensors",
170
+ "model.layers.16.mlp.gate_proj.weight": "model-08-of-35.safetensors",
171
+ "model.layers.16.mlp.up_proj.weight": "model-08-of-35.safetensors",
172
+ "model.layers.16.mlp.down_proj.weight": "model-08-of-35.safetensors",
173
+ "model.layers.16.input_layernorm.weight": "model-08-of-35.safetensors",
174
+ "model.layers.16.post_attention_layernorm.weight": "model-08-of-35.safetensors",
175
+ "model.layers.17.self_attn.q_proj.weight": "model-08-of-35.safetensors",
176
+ "model.layers.17.self_attn.k_proj.weight": "model-08-of-35.safetensors",
177
+ "model.layers.17.self_attn.v_proj.weight": "model-08-of-35.safetensors",
178
+ "model.layers.17.self_attn.o_proj.weight": "model-08-of-35.safetensors",
179
+ "model.layers.17.self_attn.rotary_emb.inv_freq": "model-08-of-35.safetensors",
180
+ "model.layers.17.mlp.gate_proj.weight": "model-08-of-35.safetensors",
181
+ "model.layers.17.mlp.up_proj.weight": "model-08-of-35.safetensors",
182
+ "model.layers.17.mlp.down_proj.weight": "model-08-of-35.safetensors",
183
+ "model.layers.17.input_layernorm.weight": "model-08-of-35.safetensors",
184
+ "model.layers.17.post_attention_layernorm.weight": "model-08-of-35.safetensors",
185
+ "model.layers.18.self_attn.q_proj.weight": "model-08-of-35.safetensors",
186
+ "model.layers.18.self_attn.k_proj.weight": "model-08-of-35.safetensors",
187
+ "model.layers.18.self_attn.v_proj.weight": "model-08-of-35.safetensors",
188
+ "model.layers.18.self_attn.o_proj.weight": "model-08-of-35.safetensors",
189
+ "model.layers.18.self_attn.rotary_emb.inv_freq": "model-08-of-35.safetensors",
190
+ "model.layers.18.mlp.gate_proj.weight": "model-09-of-35.safetensors",
191
+ "model.layers.18.mlp.up_proj.weight": "model-09-of-35.safetensors",
192
+ "model.layers.18.mlp.down_proj.weight": "model-09-of-35.safetensors",
193
+ "model.layers.18.input_layernorm.weight": "model-09-of-35.safetensors",
194
+ "model.layers.18.post_attention_layernorm.weight": "model-09-of-35.safetensors",
195
+ "model.layers.19.self_attn.q_proj.weight": "model-09-of-35.safetensors",
196
+ "model.layers.19.self_attn.k_proj.weight": "model-09-of-35.safetensors",
197
+ "model.layers.19.self_attn.v_proj.weight": "model-09-of-35.safetensors",
198
+ "model.layers.19.self_attn.o_proj.weight": "model-09-of-35.safetensors",
199
+ "model.layers.19.self_attn.rotary_emb.inv_freq": "model-09-of-35.safetensors",
200
+ "model.layers.19.mlp.gate_proj.weight": "model-09-of-35.safetensors",
201
+ "model.layers.19.mlp.up_proj.weight": "model-09-of-35.safetensors",
202
+ "model.layers.19.mlp.down_proj.weight": "model-09-of-35.safetensors",
203
+ "model.layers.19.input_layernorm.weight": "model-09-of-35.safetensors",
204
+ "model.layers.19.post_attention_layernorm.weight": "model-09-of-35.safetensors",
205
+ "model.layers.20.self_attn.q_proj.weight": "model-09-of-35.safetensors",
206
+ "model.layers.20.self_attn.k_proj.weight": "model-09-of-35.safetensors",
207
+ "model.layers.20.self_attn.v_proj.weight": "model-09-of-35.safetensors",
208
+ "model.layers.20.self_attn.o_proj.weight": "model-09-of-35.safetensors",
209
+ "model.layers.20.self_attn.rotary_emb.inv_freq": "model-09-of-35.safetensors",
210
+ "model.layers.20.mlp.gate_proj.weight": "model-09-of-35.safetensors",
211
+ "model.layers.20.mlp.up_proj.weight": "model-10-of-35.safetensors",
212
+ "model.layers.20.mlp.down_proj.weight": "model-10-of-35.safetensors",
213
+ "model.layers.20.input_layernorm.weight": "model-10-of-35.safetensors",
214
+ "model.layers.20.post_attention_layernorm.weight": "model-10-of-35.safetensors",
215
+ "model.layers.21.self_attn.q_proj.weight": "model-10-of-35.safetensors",
216
+ "model.layers.21.self_attn.k_proj.weight": "model-10-of-35.safetensors",
217
+ "model.layers.21.self_attn.v_proj.weight": "model-10-of-35.safetensors",
218
+ "model.layers.21.self_attn.o_proj.weight": "model-10-of-35.safetensors",
219
+ "model.layers.21.self_attn.rotary_emb.inv_freq": "model-10-of-35.safetensors",
220
+ "model.layers.21.mlp.gate_proj.weight": "model-10-of-35.safetensors",
221
+ "model.layers.21.mlp.up_proj.weight": "model-10-of-35.safetensors",
222
+ "model.layers.21.mlp.down_proj.weight": "model-10-of-35.safetensors",
223
+ "model.layers.21.input_layernorm.weight": "model-10-of-35.safetensors",
224
+ "model.layers.21.post_attention_layernorm.weight": "model-10-of-35.safetensors",
225
+ "model.layers.22.self_attn.q_proj.weight": "model-10-of-35.safetensors",
226
+ "model.layers.22.self_attn.k_proj.weight": "model-10-of-35.safetensors",
227
+ "model.layers.22.self_attn.v_proj.weight": "model-10-of-35.safetensors",
228
+ "model.layers.22.self_attn.o_proj.weight": "model-10-of-35.safetensors",
229
+ "model.layers.22.self_attn.rotary_emb.inv_freq": "model-10-of-35.safetensors",
230
+ "model.layers.22.mlp.gate_proj.weight": "model-10-of-35.safetensors",
231
+ "model.layers.22.mlp.up_proj.weight": "model-10-of-35.safetensors",
232
+ "model.layers.22.mlp.down_proj.weight": "model-11-of-35.safetensors",
233
+ "model.layers.22.input_layernorm.weight": "model-11-of-35.safetensors",
234
+ "model.layers.22.post_attention_layernorm.weight": "model-11-of-35.safetensors",
235
+ "model.layers.23.self_attn.q_proj.weight": "model-11-of-35.safetensors",
236
+ "model.layers.23.self_attn.k_proj.weight": "model-11-of-35.safetensors",
237
+ "model.layers.23.self_attn.v_proj.weight": "model-11-of-35.safetensors",
238
+ "model.layers.23.self_attn.o_proj.weight": "model-11-of-35.safetensors",
239
+ "model.layers.23.self_attn.rotary_emb.inv_freq": "model-11-of-35.safetensors",
240
+ "model.layers.23.mlp.gate_proj.weight": "model-11-of-35.safetensors",
241
+ "model.layers.23.mlp.up_proj.weight": "model-11-of-35.safetensors",
242
+ "model.layers.23.mlp.down_proj.weight": "model-11-of-35.safetensors",
243
+ "model.layers.23.input_layernorm.weight": "model-11-of-35.safetensors",
244
+ "model.layers.23.post_attention_layernorm.weight": "model-11-of-35.safetensors",
245
+ "model.layers.24.self_attn.q_proj.weight": "model-11-of-35.safetensors",
246
+ "model.layers.24.self_attn.k_proj.weight": "model-11-of-35.safetensors",
247
+ "model.layers.24.self_attn.v_proj.weight": "model-11-of-35.safetensors",
248
+ "model.layers.24.self_attn.o_proj.weight": "model-11-of-35.safetensors",
249
+ "model.layers.24.self_attn.rotary_emb.inv_freq": "model-11-of-35.safetensors",
250
+ "model.layers.24.mlp.gate_proj.weight": "model-11-of-35.safetensors",
251
+ "model.layers.24.mlp.up_proj.weight": "model-11-of-35.safetensors",
252
+ "model.layers.24.mlp.down_proj.weight": "model-11-of-35.safetensors",
253
+ "model.layers.24.input_layernorm.weight": "model-11-of-35.safetensors",
254
+ "model.layers.24.post_attention_layernorm.weight": "model-11-of-35.safetensors",
255
+ "model.layers.25.self_attn.q_proj.weight": "model-11-of-35.safetensors",
256
+ "model.layers.25.self_attn.k_proj.weight": "model-11-of-35.safetensors",
257
+ "model.layers.25.self_attn.v_proj.weight": "model-11-of-35.safetensors",
258
+ "model.layers.25.self_attn.o_proj.weight": "model-11-of-35.safetensors",
259
+ "model.layers.25.self_attn.rotary_emb.inv_freq": "model-11-of-35.safetensors",
260
+ "model.layers.25.mlp.gate_proj.weight": "model-12-of-35.safetensors",
261
+ "model.layers.25.mlp.up_proj.weight": "model-12-of-35.safetensors",
262
+ "model.layers.25.mlp.down_proj.weight": "model-12-of-35.safetensors",
263
+ "model.layers.25.input_layernorm.weight": "model-12-of-35.safetensors",
264
+ "model.layers.25.post_attention_layernorm.weight": "model-12-of-35.safetensors",
265
+ "model.layers.26.self_attn.q_proj.weight": "model-12-of-35.safetensors",
266
+ "model.layers.26.self_attn.k_proj.weight": "model-12-of-35.safetensors",
267
+ "model.layers.26.self_attn.v_proj.weight": "model-12-of-35.safetensors",
268
+ "model.layers.26.self_attn.o_proj.weight": "model-12-of-35.safetensors",
269
+ "model.layers.26.self_attn.rotary_emb.inv_freq": "model-12-of-35.safetensors",
270
+ "model.layers.26.mlp.gate_proj.weight": "model-12-of-35.safetensors",
271
+ "model.layers.26.mlp.up_proj.weight": "model-12-of-35.safetensors",
272
+ "model.layers.26.mlp.down_proj.weight": "model-12-of-35.safetensors",
273
+ "model.layers.26.input_layernorm.weight": "model-12-of-35.safetensors",
274
+ "model.layers.26.post_attention_layernorm.weight": "model-12-of-35.safetensors",
275
+ "model.layers.27.self_attn.q_proj.weight": "model-12-of-35.safetensors",
276
+ "model.layers.27.self_attn.k_proj.weight": "model-12-of-35.safetensors",
277
+ "model.layers.27.self_attn.v_proj.weight": "model-12-of-35.safetensors",
278
+ "model.layers.27.self_attn.o_proj.weight": "model-12-of-35.safetensors",
279
+ "model.layers.27.self_attn.rotary_emb.inv_freq": "model-12-of-35.safetensors",
280
+ "model.layers.27.mlp.gate_proj.weight": "model-12-of-35.safetensors",
281
+ "model.layers.27.mlp.up_proj.weight": "model-13-of-35.safetensors",
282
+ "model.layers.27.mlp.down_proj.weight": "model-13-of-35.safetensors",
283
+ "model.layers.27.input_layernorm.weight": "model-13-of-35.safetensors",
284
+ "model.layers.27.post_attention_layernorm.weight": "model-13-of-35.safetensors",
285
+ "model.layers.28.self_attn.q_proj.weight": "model-13-of-35.safetensors",
286
+ "model.layers.28.self_attn.k_proj.weight": "model-13-of-35.safetensors",
287
+ "model.layers.28.self_attn.v_proj.weight": "model-13-of-35.safetensors",
288
+ "model.layers.28.self_attn.o_proj.weight": "model-13-of-35.safetensors",
289
+ "model.layers.28.self_attn.rotary_emb.inv_freq": "model-13-of-35.safetensors",
290
+ "model.layers.28.mlp.gate_proj.weight": "model-13-of-35.safetensors",
291
+ "model.layers.28.mlp.up_proj.weight": "model-13-of-35.safetensors",
292
+ "model.layers.28.mlp.down_proj.weight": "model-13-of-35.safetensors",
293
+ "model.layers.28.input_layernorm.weight": "model-13-of-35.safetensors",
294
+ "model.layers.28.post_attention_layernorm.weight": "model-13-of-35.safetensors",
295
+ "model.layers.29.self_attn.q_proj.weight": "model-13-of-35.safetensors",
296
+ "model.layers.29.self_attn.k_proj.weight": "model-13-of-35.safetensors",
297
+ "model.layers.29.self_attn.v_proj.weight": "model-13-of-35.safetensors",
298
+ "model.layers.29.self_attn.o_proj.weight": "model-13-of-35.safetensors",
299
+ "model.layers.29.self_attn.rotary_emb.inv_freq": "model-13-of-35.safetensors",
300
+ "model.layers.29.mlp.gate_proj.weight": "model-13-of-35.safetensors",
301
+ "model.layers.29.mlp.up_proj.weight": "model-13-of-35.safetensors",
302
+ "model.layers.29.mlp.down_proj.weight": "model-14-of-35.safetensors",
303
+ "model.layers.29.input_layernorm.weight": "model-14-of-35.safetensors",
304
+ "model.layers.29.post_attention_layernorm.weight": "model-14-of-35.safetensors",
305
+ "model.layers.30.self_attn.q_proj.weight": "model-14-of-35.safetensors",
306
+ "model.layers.30.self_attn.k_proj.weight": "model-14-of-35.safetensors",
307
+ "model.layers.30.self_attn.v_proj.weight": "model-14-of-35.safetensors",
308
+ "model.layers.30.self_attn.o_proj.weight": "model-14-of-35.safetensors",
309
+ "model.layers.30.self_attn.rotary_emb.inv_freq": "model-14-of-35.safetensors",
310
+ "model.layers.30.mlp.gate_proj.weight": "model-14-of-35.safetensors",
311
+ "model.layers.30.mlp.up_proj.weight": "model-14-of-35.safetensors",
312
+ "model.layers.30.mlp.down_proj.weight": "model-14-of-35.safetensors",
313
+ "model.layers.30.input_layernorm.weight": "model-14-of-35.safetensors",
314
+ "model.layers.30.post_attention_layernorm.weight": "model-14-of-35.safetensors",
315
+ "model.layers.31.self_attn.q_proj.weight": "model-14-of-35.safetensors",
316
+ "model.layers.31.self_attn.k_proj.weight": "model-14-of-35.safetensors",
317
+ "model.layers.31.self_attn.v_proj.weight": "model-14-of-35.safetensors",
318
+ "model.layers.31.self_attn.o_proj.weight": "model-14-of-35.safetensors",
319
+ "model.layers.31.self_attn.rotary_emb.inv_freq": "model-14-of-35.safetensors",
320
+ "model.layers.31.mlp.gate_proj.weight": "model-14-of-35.safetensors",
321
+ "model.layers.31.mlp.up_proj.weight": "model-14-of-35.safetensors",
322
+ "model.layers.31.mlp.down_proj.weight": "model-14-of-35.safetensors",
323
+ "model.layers.31.input_layernorm.weight": "model-14-of-35.safetensors",
324
+ "model.layers.31.post_attention_layernorm.weight": "model-14-of-35.safetensors",
325
+ "model.layers.32.self_attn.q_proj.weight": "model-14-of-35.safetensors",
326
+ "model.layers.32.self_attn.k_proj.weight": "model-14-of-35.safetensors",
327
+ "model.layers.32.self_attn.v_proj.weight": "model-14-of-35.safetensors",
328
+ "model.layers.32.self_attn.o_proj.weight": "model-14-of-35.safetensors",
329
+ "model.layers.32.self_attn.rotary_emb.inv_freq": "model-14-of-35.safetensors",
330
+ "model.layers.32.mlp.gate_proj.weight": "model-15-of-35.safetensors",
331
+ "model.layers.32.mlp.up_proj.weight": "model-15-of-35.safetensors",
332
+ "model.layers.32.mlp.down_proj.weight": "model-15-of-35.safetensors",
333
+ "model.layers.32.input_layernorm.weight": "model-15-of-35.safetensors",
334
+ "model.layers.32.post_attention_layernorm.weight": "model-15-of-35.safetensors",
335
+ "model.layers.33.self_attn.q_proj.weight": "model-15-of-35.safetensors",
336
+ "model.layers.33.self_attn.k_proj.weight": "model-15-of-35.safetensors",
337
+ "model.layers.33.self_attn.v_proj.weight": "model-15-of-35.safetensors",
338
+ "model.layers.33.self_attn.o_proj.weight": "model-15-of-35.safetensors",
339
+ "model.layers.33.self_attn.rotary_emb.inv_freq": "model-15-of-35.safetensors",
340
+ "model.layers.33.mlp.gate_proj.weight": "model-15-of-35.safetensors",
341
+ "model.layers.33.mlp.up_proj.weight": "model-15-of-35.safetensors",
342
+ "model.layers.33.mlp.down_proj.weight": "model-15-of-35.safetensors",
343
+ "model.layers.33.input_layernorm.weight": "model-15-of-35.safetensors",
344
+ "model.layers.33.post_attention_layernorm.weight": "model-15-of-35.safetensors",
345
+ "model.layers.34.self_attn.q_proj.weight": "model-15-of-35.safetensors",
346
+ "model.layers.34.self_attn.k_proj.weight": "model-15-of-35.safetensors",
347
+ "model.layers.34.self_attn.v_proj.weight": "model-15-of-35.safetensors",
348
+ "model.layers.34.self_attn.o_proj.weight": "model-15-of-35.safetensors",
349
+ "model.layers.34.self_attn.rotary_emb.inv_freq": "model-15-of-35.safetensors",
350
+ "model.layers.34.mlp.gate_proj.weight": "model-15-of-35.safetensors",
351
+ "model.layers.34.mlp.up_proj.weight": "model-16-of-35.safetensors",
352
+ "model.layers.34.mlp.down_proj.weight": "model-16-of-35.safetensors",
353
+ "model.layers.34.input_layernorm.weight": "model-16-of-35.safetensors",
354
+ "model.layers.34.post_attention_layernorm.weight": "model-16-of-35.safetensors",
355
+ "model.layers.35.self_attn.q_proj.weight": "model-16-of-35.safetensors",
356
+ "model.layers.35.self_attn.k_proj.weight": "model-16-of-35.safetensors",
357
+ "model.layers.35.self_attn.v_proj.weight": "model-16-of-35.safetensors",
358
+ "model.layers.35.self_attn.o_proj.weight": "model-16-of-35.safetensors",
359
+ "model.layers.35.self_attn.rotary_emb.inv_freq": "model-16-of-35.safetensors",
360
+ "model.layers.35.mlp.gate_proj.weight": "model-16-of-35.safetensors",
361
+ "model.layers.35.mlp.up_proj.weight": "model-16-of-35.safetensors",
362
+ "model.layers.35.mlp.down_proj.weight": "model-16-of-35.safetensors",
363
+ "model.layers.35.input_layernorm.weight": "model-16-of-35.safetensors",
364
+ "model.layers.35.post_attention_layernorm.weight": "model-16-of-35.safetensors",
365
+ "model.layers.36.self_attn.q_proj.weight": "model-16-of-35.safetensors",
366
+ "model.layers.36.self_attn.k_proj.weight": "model-16-of-35.safetensors",
367
+ "model.layers.36.self_attn.v_proj.weight": "model-16-of-35.safetensors",
368
+ "model.layers.36.self_attn.o_proj.weight": "model-16-of-35.safetensors",
369
+ "model.layers.36.self_attn.rotary_emb.inv_freq": "model-16-of-35.safetensors",
370
+ "model.layers.36.mlp.gate_proj.weight": "model-16-of-35.safetensors",
371
+ "model.layers.36.mlp.up_proj.weight": "model-16-of-35.safetensors",
372
+ "model.layers.36.mlp.down_proj.weight": "model-17-of-35.safetensors",
373
+ "model.layers.36.input_layernorm.weight": "model-17-of-35.safetensors",
374
+ "model.layers.36.post_attention_layernorm.weight": "model-17-of-35.safetensors",
375
+ "model.layers.37.self_attn.q_proj.weight": "model-17-of-35.safetensors",
376
+ "model.layers.37.self_attn.k_proj.weight": "model-17-of-35.safetensors",
377
+ "model.layers.37.self_attn.v_proj.weight": "model-17-of-35.safetensors",
378
+ "model.layers.37.self_attn.o_proj.weight": "model-17-of-35.safetensors",
379
+ "model.layers.37.self_attn.rotary_emb.inv_freq": "model-17-of-35.safetensors",
380
+ "model.layers.37.mlp.gate_proj.weight": "model-17-of-35.safetensors",
381
+ "model.layers.37.mlp.up_proj.weight": "model-17-of-35.safetensors",
382
+ "model.layers.37.mlp.down_proj.weight": "model-17-of-35.safetensors",
383
+ "model.layers.37.input_layernorm.weight": "model-17-of-35.safetensors",
384
+ "model.layers.37.post_attention_layernorm.weight": "model-17-of-35.safetensors",
385
+ "model.layers.38.self_attn.q_proj.weight": "model-17-of-35.safetensors",
386
+ "model.layers.38.self_attn.k_proj.weight": "model-17-of-35.safetensors",
387
+ "model.layers.38.self_attn.v_proj.weight": "model-17-of-35.safetensors",
388
+ "model.layers.38.self_attn.o_proj.weight": "model-17-of-35.safetensors",
389
+ "model.layers.38.self_attn.rotary_emb.inv_freq": "model-17-of-35.safetensors",
390
+ "model.layers.38.mlp.gate_proj.weight": "model-17-of-35.safetensors",
391
+ "model.layers.38.mlp.up_proj.weight": "model-17-of-35.safetensors",
392
+ "model.layers.38.mlp.down_proj.weight": "model-17-of-35.safetensors",
393
+ "model.layers.38.input_layernorm.weight": "model-17-of-35.safetensors",
394
+ "model.layers.38.post_attention_layernorm.weight": "model-17-of-35.safetensors",
395
+ "model.layers.39.self_attn.q_proj.weight": "model-17-of-35.safetensors",
396
+ "model.layers.39.self_attn.k_proj.weight": "model-17-of-35.safetensors",
397
+ "model.layers.39.self_attn.v_proj.weight": "model-17-of-35.safetensors",
398
+ "model.layers.39.self_attn.o_proj.weight": "model-17-of-35.safetensors",
399
+ "model.layers.39.self_attn.rotary_emb.inv_freq": "model-17-of-35.safetensors",
400
+ "model.layers.39.mlp.gate_proj.weight": "model-18-of-35.safetensors",
401
+ "model.layers.39.mlp.up_proj.weight": "model-18-of-35.safetensors",
402
+ "model.layers.39.mlp.down_proj.weight": "model-18-of-35.safetensors",
403
+ "model.layers.39.input_layernorm.weight": "model-18-of-35.safetensors",
404
+ "model.layers.39.post_attention_layernorm.weight": "model-18-of-35.safetensors",
405
+ "model.layers.40.self_attn.q_proj.weight": "model-18-of-35.safetensors",
406
+ "model.layers.40.self_attn.k_proj.weight": "model-18-of-35.safetensors",
407
+ "model.layers.40.self_attn.v_proj.weight": "model-18-of-35.safetensors",
408
+ "model.layers.40.self_attn.o_proj.weight": "model-18-of-35.safetensors",
409
+ "model.layers.40.self_attn.rotary_emb.inv_freq": "model-18-of-35.safetensors",
410
+ "model.layers.40.mlp.gate_proj.weight": "model-18-of-35.safetensors",
411
+ "model.layers.40.mlp.up_proj.weight": "model-18-of-35.safetensors",
412
+ "model.layers.40.mlp.down_proj.weight": "model-18-of-35.safetensors",
413
+ "model.layers.40.input_layernorm.weight": "model-18-of-35.safetensors",
414
+ "model.layers.40.post_attention_layernorm.weight": "model-18-of-35.safetensors",
415
+ "model.layers.41.self_attn.q_proj.weight": "model-18-of-35.safetensors",
416
+ "model.layers.41.self_attn.k_proj.weight": "model-18-of-35.safetensors",
417
+ "model.layers.41.self_attn.v_proj.weight": "model-18-of-35.safetensors",
418
+ "model.layers.41.self_attn.o_proj.weight": "model-18-of-35.safetensors",
419
+ "model.layers.41.self_attn.rotary_emb.inv_freq": "model-18-of-35.safetensors",
420
+ "model.layers.41.mlp.gate_proj.weight": "model-18-of-35.safetensors",
421
+ "model.layers.41.mlp.up_proj.weight": "model-19-of-35.safetensors",
422
+ "model.layers.41.mlp.down_proj.weight": "model-19-of-35.safetensors",
423
+ "model.layers.41.input_layernorm.weight": "model-19-of-35.safetensors",
424
+ "model.layers.41.post_attention_layernorm.weight": "model-19-of-35.safetensors",
425
+ "model.layers.42.self_attn.q_proj.weight": "model-19-of-35.safetensors",
426
+ "model.layers.42.self_attn.k_proj.weight": "model-19-of-35.safetensors",
427
+ "model.layers.42.self_attn.v_proj.weight": "model-19-of-35.safetensors",
428
+ "model.layers.42.self_attn.o_proj.weight": "model-19-of-35.safetensors",
429
+ "model.layers.42.self_attn.rotary_emb.inv_freq": "model-19-of-35.safetensors",
430
+ "model.layers.42.mlp.gate_proj.weight": "model-19-of-35.safetensors",
431
+ "model.layers.42.mlp.up_proj.weight": "model-19-of-35.safetensors",
432
+ "model.layers.42.mlp.down_proj.weight": "model-19-of-35.safetensors",
433
+ "model.layers.42.input_layernorm.weight": "model-19-of-35.safetensors",
434
+ "model.layers.42.post_attention_layernorm.weight": "model-19-of-35.safetensors",
435
+ "model.layers.43.self_attn.q_proj.weight": "model-19-of-35.safetensors",
436
+ "model.layers.43.self_attn.k_proj.weight": "model-19-of-35.safetensors",
437
+ "model.layers.43.self_attn.v_proj.weight": "model-19-of-35.safetensors",
438
+ "model.layers.43.self_attn.o_proj.weight": "model-19-of-35.safetensors",
439
+ "model.layers.43.self_attn.rotary_emb.inv_freq": "model-19-of-35.safetensors",
440
+ "model.layers.43.mlp.gate_proj.weight": "model-19-of-35.safetensors",
441
+ "model.layers.43.mlp.up_proj.weight": "model-19-of-35.safetensors",
442
+ "model.layers.43.mlp.down_proj.weight": "model-20-of-35.safetensors",
443
+ "model.layers.43.input_layernorm.weight": "model-20-of-35.safetensors",
444
+ "model.layers.43.post_attention_layernorm.weight": "model-20-of-35.safetensors",
445
+ "model.layers.44.self_attn.q_proj.weight": "model-20-of-35.safetensors",
446
+ "model.layers.44.self_attn.k_proj.weight": "model-20-of-35.safetensors",
447
+ "model.layers.44.self_attn.v_proj.weight": "model-20-of-35.safetensors",
448
+ "model.layers.44.self_attn.o_proj.weight": "model-20-of-35.safetensors",
449
+ "model.layers.44.self_attn.rotary_emb.inv_freq": "model-20-of-35.safetensors",
450
+ "model.layers.44.mlp.gate_proj.weight": "model-20-of-35.safetensors",
451
+ "model.layers.44.mlp.up_proj.weight": "model-20-of-35.safetensors",
452
+ "model.layers.44.mlp.down_proj.weight": "model-20-of-35.safetensors",
453
+ "model.layers.44.input_layernorm.weight": "model-20-of-35.safetensors",
454
+ "model.layers.44.post_attention_layernorm.weight": "model-20-of-35.safetensors",
455
+ "model.layers.45.self_attn.q_proj.weight": "model-20-of-35.safetensors",
456
+ "model.layers.45.self_attn.k_proj.weight": "model-20-of-35.safetensors",
457
+ "model.layers.45.self_attn.v_proj.weight": "model-20-of-35.safetensors",
458
+ "model.layers.45.self_attn.o_proj.weight": "model-20-of-35.safetensors",
459
+ "model.layers.45.self_attn.rotary_emb.inv_freq": "model-20-of-35.safetensors",
460
+ "model.layers.45.mlp.gate_proj.weight": "model-20-of-35.safetensors",
461
+ "model.layers.45.mlp.up_proj.weight": "model-20-of-35.safetensors",
462
+ "model.layers.45.mlp.down_proj.weight": "model-20-of-35.safetensors",
463
+ "model.layers.45.input_layernorm.weight": "model-20-of-35.safetensors",
464
+ "model.layers.45.post_attention_layernorm.weight": "model-20-of-35.safetensors",
465
+ "model.layers.46.self_attn.q_proj.weight": "model-20-of-35.safetensors",
466
+ "model.layers.46.self_attn.k_proj.weight": "model-20-of-35.safetensors",
467
+ "model.layers.46.self_attn.v_proj.weight": "model-20-of-35.safetensors",
468
+ "model.layers.46.self_attn.o_proj.weight": "model-20-of-35.safetensors",
469
+ "model.layers.46.self_attn.rotary_emb.inv_freq": "model-20-of-35.safetensors",
470
+ "model.layers.46.mlp.gate_proj.weight": "model-21-of-35.safetensors",
471
+ "model.layers.46.mlp.up_proj.weight": "model-21-of-35.safetensors",
472
+ "model.layers.46.mlp.down_proj.weight": "model-21-of-35.safetensors",
473
+ "model.layers.46.input_layernorm.weight": "model-21-of-35.safetensors",
474
+ "model.layers.46.post_attention_layernorm.weight": "model-21-of-35.safetensors",
475
+ "model.layers.47.self_attn.q_proj.weight": "model-21-of-35.safetensors",
476
+ "model.layers.47.self_attn.k_proj.weight": "model-21-of-35.safetensors",
477
+ "model.layers.47.self_attn.v_proj.weight": "model-21-of-35.safetensors",
478
+ "model.layers.47.self_attn.o_proj.weight": "model-21-of-35.safetensors",
479
+ "model.layers.47.self_attn.rotary_emb.inv_freq": "model-21-of-35.safetensors",
480
+ "model.layers.47.mlp.gate_proj.weight": "model-21-of-35.safetensors",
481
+ "model.layers.47.mlp.up_proj.weight": "model-21-of-35.safetensors",
482
+ "model.layers.47.mlp.down_proj.weight": "model-21-of-35.safetensors",
483
+ "model.layers.47.input_layernorm.weight": "model-21-of-35.safetensors",
484
+ "model.layers.47.post_attention_layernorm.weight": "model-21-of-35.safetensors",
485
+ "model.layers.48.self_attn.q_proj.weight": "model-21-of-35.safetensors",
486
+ "model.layers.48.self_attn.k_proj.weight": "model-21-of-35.safetensors",
487
+ "model.layers.48.self_attn.v_proj.weight": "model-21-of-35.safetensors",
488
+ "model.layers.48.self_attn.o_proj.weight": "model-21-of-35.safetensors",
489
+ "model.layers.48.self_attn.rotary_emb.inv_freq": "model-21-of-35.safetensors",
490
+ "model.layers.48.mlp.gate_proj.weight": "model-21-of-35.safetensors",
491
+ "model.layers.48.mlp.up_proj.weight": "model-22-of-35.safetensors",
492
+ "model.layers.48.mlp.down_proj.weight": "model-22-of-35.safetensors",
493
+ "model.layers.48.input_layernorm.weight": "model-22-of-35.safetensors",
494
+ "model.layers.48.post_attention_layernorm.weight": "model-22-of-35.safetensors",
495
+ "model.layers.49.self_attn.q_proj.weight": "model-22-of-35.safetensors",
496
+ "model.layers.49.self_attn.k_proj.weight": "model-22-of-35.safetensors",
497
+ "model.layers.49.self_attn.v_proj.weight": "model-22-of-35.safetensors",
498
+ "model.layers.49.self_attn.o_proj.weight": "model-22-of-35.safetensors",
499
+ "model.layers.49.self_attn.rotary_emb.inv_freq": "model-22-of-35.safetensors",
500
+ "model.layers.49.mlp.gate_proj.weight": "model-22-of-35.safetensors",
501
+ "model.layers.49.mlp.up_proj.weight": "model-22-of-35.safetensors",
502
+ "model.layers.49.mlp.down_proj.weight": "model-22-of-35.safetensors",
503
+ "model.layers.49.input_layernorm.weight": "model-22-of-35.safetensors",
504
+ "model.layers.49.post_attention_layernorm.weight": "model-22-of-35.safetensors",
505
+ "model.layers.50.self_attn.q_proj.weight": "model-22-of-35.safetensors",
506
+ "model.layers.50.self_attn.k_proj.weight": "model-22-of-35.safetensors",
507
+ "model.layers.50.self_attn.v_proj.weight": "model-22-of-35.safetensors",
508
+ "model.layers.50.self_attn.o_proj.weight": "model-22-of-35.safetensors",
509
+ "model.layers.50.self_attn.rotary_emb.inv_freq": "model-22-of-35.safetensors",
510
+ "model.layers.50.mlp.gate_proj.weight": "model-22-of-35.safetensors",
511
+ "model.layers.50.mlp.up_proj.weight": "model-22-of-35.safetensors",
512
+ "model.layers.50.mlp.down_proj.weight": "model-23-of-35.safetensors",
513
+ "model.layers.50.input_layernorm.weight": "model-23-of-35.safetensors",
514
+ "model.layers.50.post_attention_layernorm.weight": "model-23-of-35.safetensors",
515
+ "model.layers.51.self_attn.q_proj.weight": "model-23-of-35.safetensors",
516
+ "model.layers.51.self_attn.k_proj.weight": "model-23-of-35.safetensors",
517
+ "model.layers.51.self_attn.v_proj.weight": "model-23-of-35.safetensors",
518
+ "model.layers.51.self_attn.o_proj.weight": "model-23-of-35.safetensors",
519
+ "model.layers.51.self_attn.rotary_emb.inv_freq": "model-23-of-35.safetensors",
520
+ "model.layers.51.mlp.gate_proj.weight": "model-23-of-35.safetensors",
521
+ "model.layers.51.mlp.up_proj.weight": "model-23-of-35.safetensors",
522
+ "model.layers.51.mlp.down_proj.weight": "model-23-of-35.safetensors",
523
+ "model.layers.51.input_layernorm.weight": "model-23-of-35.safetensors",
524
+ "model.layers.51.post_attention_layernorm.weight": "model-23-of-35.safetensors",
525
+ "model.layers.52.self_attn.q_proj.weight": "model-23-of-35.safetensors",
526
+ "model.layers.52.self_attn.k_proj.weight": "model-23-of-35.safetensors",
527
+ "model.layers.52.self_attn.v_proj.weight": "model-23-of-35.safetensors",
528
+ "model.layers.52.self_attn.o_proj.weight": "model-23-of-35.safetensors",
529
+ "model.layers.52.self_attn.rotary_emb.inv_freq": "model-23-of-35.safetensors",
530
+ "model.layers.52.mlp.gate_proj.weight": "model-23-of-35.safetensors",
531
+ "model.layers.52.mlp.up_proj.weight": "model-23-of-35.safetensors",
532
+ "model.layers.52.mlp.down_proj.weight": "model-23-of-35.safetensors",
533
+ "model.layers.52.input_layernorm.weight": "model-23-of-35.safetensors",
534
+ "model.layers.52.post_attention_layernorm.weight": "model-23-of-35.safetensors",
535
+ "model.layers.53.self_attn.q_proj.weight": "model-23-of-35.safetensors",
536
+ "model.layers.53.self_attn.k_proj.weight": "model-23-of-35.safetensors",
537
+ "model.layers.53.self_attn.v_proj.weight": "model-23-of-35.safetensors",
538
+ "model.layers.53.self_attn.o_proj.weight": "model-23-of-35.safetensors",
539
+ "model.layers.53.self_attn.rotary_emb.inv_freq": "model-23-of-35.safetensors",
540
+ "model.layers.53.mlp.gate_proj.weight": "model-24-of-35.safetensors",
541
+ "model.layers.53.mlp.up_proj.weight": "model-24-of-35.safetensors",
542
+ "model.layers.53.mlp.down_proj.weight": "model-24-of-35.safetensors",
543
+ "model.layers.53.input_layernorm.weight": "model-24-of-35.safetensors",
544
+ "model.layers.53.post_attention_layernorm.weight": "model-24-of-35.safetensors",
545
+ "model.layers.54.self_attn.q_proj.weight": "model-24-of-35.safetensors",
546
+ "model.layers.54.self_attn.k_proj.weight": "model-24-of-35.safetensors",
547
+ "model.layers.54.self_attn.v_proj.weight": "model-24-of-35.safetensors",
548
+ "model.layers.54.self_attn.o_proj.weight": "model-24-of-35.safetensors",
549
+ "model.layers.54.self_attn.rotary_emb.inv_freq": "model-24-of-35.safetensors",
550
+ "model.layers.54.mlp.gate_proj.weight": "model-24-of-35.safetensors",
551
+ "model.layers.54.mlp.up_proj.weight": "model-24-of-35.safetensors",
552
+ "model.layers.54.mlp.down_proj.weight": "model-24-of-35.safetensors",
553
+ "model.layers.54.input_layernorm.weight": "model-24-of-35.safetensors",
554
+ "model.layers.54.post_attention_layernorm.weight": "model-24-of-35.safetensors",
555
+ "model.layers.55.self_attn.q_proj.weight": "model-24-of-35.safetensors",
556
+ "model.layers.55.self_attn.k_proj.weight": "model-24-of-35.safetensors",
557
+ "model.layers.55.self_attn.v_proj.weight": "model-24-of-35.safetensors",
558
+ "model.layers.55.self_attn.o_proj.weight": "model-24-of-35.safetensors",
559
+ "model.layers.55.self_attn.rotary_emb.inv_freq": "model-24-of-35.safetensors",
560
+ "model.layers.55.mlp.gate_proj.weight": "model-24-of-35.safetensors",
561
+ "model.layers.55.mlp.up_proj.weight": "model-25-of-35.safetensors",
562
+ "model.layers.55.mlp.down_proj.weight": "model-25-of-35.safetensors",
563
+ "model.layers.55.input_layernorm.weight": "model-25-of-35.safetensors",
564
+ "model.layers.55.post_attention_layernorm.weight": "model-25-of-35.safetensors",
565
+ "model.layers.56.self_attn.q_proj.weight": "model-25-of-35.safetensors",
566
+ "model.layers.56.self_attn.k_proj.weight": "model-25-of-35.safetensors",
567
+ "model.layers.56.self_attn.v_proj.weight": "model-25-of-35.safetensors",
568
+ "model.layers.56.self_attn.o_proj.weight": "model-25-of-35.safetensors",
569
+ "model.layers.56.self_attn.rotary_emb.inv_freq": "model-25-of-35.safetensors",
570
+ "model.layers.56.mlp.gate_proj.weight": "model-25-of-35.safetensors",
571
+ "model.layers.56.mlp.up_proj.weight": "model-25-of-35.safetensors",
572
+ "model.layers.56.mlp.down_proj.weight": "model-25-of-35.safetensors",
573
+ "model.layers.56.input_layernorm.weight": "model-25-of-35.safetensors",
574
+ "model.layers.56.post_attention_layernorm.weight": "model-25-of-35.safetensors",
575
+ "model.layers.57.self_attn.q_proj.weight": "model-25-of-35.safetensors",
576
+ "model.layers.57.self_attn.k_proj.weight": "model-25-of-35.safetensors",
577
+ "model.layers.57.self_attn.v_proj.weight": "model-25-of-35.safetensors",
578
+ "model.layers.57.self_attn.o_proj.weight": "model-25-of-35.safetensors",
579
+ "model.layers.57.self_attn.rotary_emb.inv_freq": "model-25-of-35.safetensors",
580
+ "model.layers.57.mlp.gate_proj.weight": "model-25-of-35.safetensors",
581
+ "model.layers.57.mlp.up_proj.weight": "model-25-of-35.safetensors",
582
+ "model.layers.57.mlp.down_proj.weight": "model-26-of-35.safetensors",
583
+ "model.layers.57.input_layernorm.weight": "model-26-of-35.safetensors",
584
+ "model.layers.57.post_attention_layernorm.weight": "model-26-of-35.safetensors",
585
+ "model.layers.58.self_attn.q_proj.weight": "model-26-of-35.safetensors",
586
+ "model.layers.58.self_attn.k_proj.weight": "model-26-of-35.safetensors",
587
+ "model.layers.58.self_attn.v_proj.weight": "model-26-of-35.safetensors",
588
+ "model.layers.58.self_attn.o_proj.weight": "model-26-of-35.safetensors",
589
+ "model.layers.58.self_attn.rotary_emb.inv_freq": "model-26-of-35.safetensors",
590
+ "model.layers.58.mlp.gate_proj.weight": "model-26-of-35.safetensors",
591
+ "model.layers.58.mlp.up_proj.weight": "model-26-of-35.safetensors",
592
+ "model.layers.58.mlp.down_proj.weight": "model-26-of-35.safetensors",
593
+ "model.layers.58.input_layernorm.weight": "model-26-of-35.safetensors",
594
+ "model.layers.58.post_attention_layernorm.weight": "model-26-of-35.safetensors",
595
+ "model.layers.59.self_attn.q_proj.weight": "model-26-of-35.safetensors",
596
+ "model.layers.59.self_attn.k_proj.weight": "model-26-of-35.safetensors",
597
+ "model.layers.59.self_attn.v_proj.weight": "model-26-of-35.safetensors",
598
+ "model.layers.59.self_attn.o_proj.weight": "model-26-of-35.safetensors",
599
+ "model.layers.59.self_attn.rotary_emb.inv_freq": "model-26-of-35.safetensors",
600
+ "model.layers.59.mlp.gate_proj.weight": "model-26-of-35.safetensors",
601
+ "model.layers.59.mlp.up_proj.weight": "model-26-of-35.safetensors",
602
+ "model.layers.59.mlp.down_proj.weight": "model-26-of-35.safetensors",
603
+ "model.layers.59.input_layernorm.weight": "model-26-of-35.safetensors",
604
+ "model.layers.59.post_attention_layernorm.weight": "model-26-of-35.safetensors",
605
+ "model.layers.60.self_attn.q_proj.weight": "model-26-of-35.safetensors",
606
+ "model.layers.60.self_attn.k_proj.weight": "model-26-of-35.safetensors",
607
+ "model.layers.60.self_attn.v_proj.weight": "model-26-of-35.safetensors",
608
+ "model.layers.60.self_attn.o_proj.weight": "model-26-of-35.safetensors",
609
+ "model.layers.60.self_attn.rotary_emb.inv_freq": "model-26-of-35.safetensors",
610
+ "model.layers.60.mlp.gate_proj.weight": "model-27-of-35.safetensors",
611
+ "model.layers.60.mlp.up_proj.weight": "model-27-of-35.safetensors",
612
+ "model.layers.60.mlp.down_proj.weight": "model-27-of-35.safetensors",
613
+ "model.layers.60.input_layernorm.weight": "model-27-of-35.safetensors",
614
+ "model.layers.60.post_attention_layernorm.weight": "model-27-of-35.safetensors",
615
+ "model.layers.61.self_attn.q_proj.weight": "model-27-of-35.safetensors",
616
+ "model.layers.61.self_attn.k_proj.weight": "model-27-of-35.safetensors",
617
+ "model.layers.61.self_attn.v_proj.weight": "model-27-of-35.safetensors",
618
+ "model.layers.61.self_attn.o_proj.weight": "model-27-of-35.safetensors",
619
+ "model.layers.61.self_attn.rotary_emb.inv_freq": "model-27-of-35.safetensors",
620
+ "model.layers.61.mlp.gate_proj.weight": "model-27-of-35.safetensors",
621
+ "model.layers.61.mlp.up_proj.weight": "model-27-of-35.safetensors",
622
+ "model.layers.61.mlp.down_proj.weight": "model-27-of-35.safetensors",
623
+ "model.layers.61.input_layernorm.weight": "model-27-of-35.safetensors",
624
+ "model.layers.61.post_attention_layernorm.weight": "model-27-of-35.safetensors",
625
+ "model.layers.62.self_attn.q_proj.weight": "model-27-of-35.safetensors",
626
+ "model.layers.62.self_attn.k_proj.weight": "model-27-of-35.safetensors",
627
+ "model.layers.62.self_attn.v_proj.weight": "model-27-of-35.safetensors",
628
+ "model.layers.62.self_attn.o_proj.weight": "model-27-of-35.safetensors",
629
+ "model.layers.62.self_attn.rotary_emb.inv_freq": "model-27-of-35.safetensors",
630
+ "model.layers.62.mlp.gate_proj.weight": "model-27-of-35.safetensors",
631
+ "model.layers.62.mlp.up_proj.weight": "model-28-of-35.safetensors",
632
+ "model.layers.62.mlp.down_proj.weight": "model-28-of-35.safetensors",
633
+ "model.layers.62.input_layernorm.weight": "model-28-of-35.safetensors",
634
+ "model.layers.62.post_attention_layernorm.weight": "model-28-of-35.safetensors",
635
+ "model.layers.63.self_attn.q_proj.weight": "model-28-of-35.safetensors",
636
+ "model.layers.63.self_attn.k_proj.weight": "model-28-of-35.safetensors",
637
+ "model.layers.63.self_attn.v_proj.weight": "model-28-of-35.safetensors",
638
+ "model.layers.63.self_attn.o_proj.weight": "model-28-of-35.safetensors",
639
+ "model.layers.63.self_attn.rotary_emb.inv_freq": "model-28-of-35.safetensors",
640
+ "model.layers.63.mlp.gate_proj.weight": "model-28-of-35.safetensors",
641
+ "model.layers.63.mlp.up_proj.weight": "model-28-of-35.safetensors",
642
+ "model.layers.63.mlp.down_proj.weight": "model-28-of-35.safetensors",
643
+ "model.layers.63.input_layernorm.weight": "model-28-of-35.safetensors",
644
+ "model.layers.63.post_attention_layernorm.weight": "model-28-of-35.safetensors",
645
+ "model.layers.64.self_attn.q_proj.weight": "model-28-of-35.safetensors",
646
+ "model.layers.64.self_attn.k_proj.weight": "model-28-of-35.safetensors",
647
+ "model.layers.64.self_attn.v_proj.weight": "model-28-of-35.safetensors",
648
+ "model.layers.64.self_attn.o_proj.weight": "model-28-of-35.safetensors",
649
+ "model.layers.64.self_attn.rotary_emb.inv_freq": "model-28-of-35.safetensors",
650
+ "model.layers.64.mlp.gate_proj.weight": "model-28-of-35.safetensors",
651
+ "model.layers.64.mlp.up_proj.weight": "model-28-of-35.safetensors",
652
+ "model.layers.64.mlp.down_proj.weight": "model-29-of-35.safetensors",
653
+ "model.layers.64.input_layernorm.weight": "model-29-of-35.safetensors",
654
+ "model.layers.64.post_attention_layernorm.weight": "model-29-of-35.safetensors",
655
+ "model.layers.65.self_attn.q_proj.weight": "model-29-of-35.safetensors",
656
+ "model.layers.65.self_attn.k_proj.weight": "model-29-of-35.safetensors",
657
+ "model.layers.65.self_attn.v_proj.weight": "model-29-of-35.safetensors",
658
+ "model.layers.65.self_attn.o_proj.weight": "model-29-of-35.safetensors",
659
+ "model.layers.65.self_attn.rotary_emb.inv_freq": "model-29-of-35.safetensors",
660
+ "model.layers.65.mlp.gate_proj.weight": "model-29-of-35.safetensors",
661
+ "model.layers.65.mlp.up_proj.weight": "model-29-of-35.safetensors",
662
+ "model.layers.65.mlp.down_proj.weight": "model-29-of-35.safetensors",
663
+ "model.layers.65.input_layernorm.weight": "model-29-of-35.safetensors",
664
+ "model.layers.65.post_attention_layernorm.weight": "model-29-of-35.safetensors",
665
+ "model.layers.66.self_attn.q_proj.weight": "model-29-of-35.safetensors",
666
+ "model.layers.66.self_attn.k_proj.weight": "model-29-of-35.safetensors",
667
+ "model.layers.66.self_attn.v_proj.weight": "model-29-of-35.safetensors",
668
+ "model.layers.66.self_attn.o_proj.weight": "model-29-of-35.safetensors",
669
+ "model.layers.66.self_attn.rotary_emb.inv_freq": "model-29-of-35.safetensors",
670
+ "model.layers.66.mlp.gate_proj.weight": "model-29-of-35.safetensors",
671
+ "model.layers.66.mlp.up_proj.weight": "model-29-of-35.safetensors",
672
+ "model.layers.66.mlp.down_proj.weight": "model-29-of-35.safetensors",
673
+ "model.layers.66.input_layernorm.weight": "model-29-of-35.safetensors",
674
+ "model.layers.66.post_attention_layernorm.weight": "model-29-of-35.safetensors",
675
+ "model.layers.67.self_attn.q_proj.weight": "model-29-of-35.safetensors",
676
+ "model.layers.67.self_attn.k_proj.weight": "model-29-of-35.safetensors",
677
+ "model.layers.67.self_attn.v_proj.weight": "model-29-of-35.safetensors",
678
+ "model.layers.67.self_attn.o_proj.weight": "model-29-of-35.safetensors",
679
+ "model.layers.67.self_attn.rotary_emb.inv_freq": "model-29-of-35.safetensors",
680
+ "model.layers.67.mlp.gate_proj.weight": "model-30-of-35.safetensors",
681
+ "model.layers.67.mlp.up_proj.weight": "model-30-of-35.safetensors",
682
+ "model.layers.67.mlp.down_proj.weight": "model-30-of-35.safetensors",
683
+ "model.layers.67.input_layernorm.weight": "model-30-of-35.safetensors",
684
+ "model.layers.67.post_attention_layernorm.weight": "model-30-of-35.safetensors",
685
+ "model.layers.68.self_attn.q_proj.weight": "model-30-of-35.safetensors",
686
+ "model.layers.68.self_attn.k_proj.weight": "model-30-of-35.safetensors",
687
+ "model.layers.68.self_attn.v_proj.weight": "model-30-of-35.safetensors",
688
+ "model.layers.68.self_attn.o_proj.weight": "model-30-of-35.safetensors",
689
+ "model.layers.68.self_attn.rotary_emb.inv_freq": "model-30-of-35.safetensors",
690
+ "model.layers.68.mlp.gate_proj.weight": "model-30-of-35.safetensors",
691
+ "model.layers.68.mlp.up_proj.weight": "model-30-of-35.safetensors",
692
+ "model.layers.68.mlp.down_proj.weight": "model-30-of-35.safetensors",
693
+ "model.layers.68.input_layernorm.weight": "model-30-of-35.safetensors",
694
+ "model.layers.68.post_attention_layernorm.weight": "model-30-of-35.safetensors",
695
+ "model.layers.69.self_attn.q_proj.weight": "model-30-of-35.safetensors",
696
+ "model.layers.69.self_attn.k_proj.weight": "model-30-of-35.safetensors",
697
+ "model.layers.69.self_attn.v_proj.weight": "model-30-of-35.safetensors",
698
+ "model.layers.69.self_attn.o_proj.weight": "model-30-of-35.safetensors",
699
+ "model.layers.69.self_attn.rotary_emb.inv_freq": "model-30-of-35.safetensors",
700
+ "model.layers.69.mlp.gate_proj.weight": "model-30-of-35.safetensors",
701
+ "model.layers.69.mlp.up_proj.weight": "model-31-of-35.safetensors",
702
+ "model.layers.69.mlp.down_proj.weight": "model-31-of-35.safetensors",
703
+ "model.layers.69.input_layernorm.weight": "model-31-of-35.safetensors",
704
+ "model.layers.69.post_attention_layernorm.weight": "model-31-of-35.safetensors",
705
+ "model.layers.70.self_attn.q_proj.weight": "model-31-of-35.safetensors",
706
+ "model.layers.70.self_attn.k_proj.weight": "model-31-of-35.safetensors",
707
+ "model.layers.70.self_attn.v_proj.weight": "model-31-of-35.safetensors",
708
+ "model.layers.70.self_attn.o_proj.weight": "model-31-of-35.safetensors",
709
+ "model.layers.70.self_attn.rotary_emb.inv_freq": "model-31-of-35.safetensors",
710
+ "model.layers.70.mlp.gate_proj.weight": "model-31-of-35.safetensors",
711
+ "model.layers.70.mlp.up_proj.weight": "model-31-of-35.safetensors",
712
+ "model.layers.70.mlp.down_proj.weight": "model-31-of-35.safetensors",
713
+ "model.layers.70.input_layernorm.weight": "model-31-of-35.safetensors",
714
+ "model.layers.70.post_attention_layernorm.weight": "model-31-of-35.safetensors",
715
+ "model.layers.71.self_attn.q_proj.weight": "model-31-of-35.safetensors",
716
+ "model.layers.71.self_attn.k_proj.weight": "model-31-of-35.safetensors",
717
+ "model.layers.71.self_attn.v_proj.weight": "model-31-of-35.safetensors",
718
+ "model.layers.71.self_attn.o_proj.weight": "model-31-of-35.safetensors",
719
+ "model.layers.71.self_attn.rotary_emb.inv_freq": "model-31-of-35.safetensors",
720
+ "model.layers.71.mlp.gate_proj.weight": "model-31-of-35.safetensors",
721
+ "model.layers.71.mlp.up_proj.weight": "model-31-of-35.safetensors",
722
+ "model.layers.71.mlp.down_proj.weight": "model-32-of-35.safetensors",
723
+ "model.layers.71.input_layernorm.weight": "model-32-of-35.safetensors",
724
+ "model.layers.71.post_attention_layernorm.weight": "model-32-of-35.safetensors",
725
+ "model.layers.72.self_attn.q_proj.weight": "model-32-of-35.safetensors",
726
+ "model.layers.72.self_attn.k_proj.weight": "model-32-of-35.safetensors",
727
+ "model.layers.72.self_attn.v_proj.weight": "model-32-of-35.safetensors",
728
+ "model.layers.72.self_attn.o_proj.weight": "model-32-of-35.safetensors",
729
+ "model.layers.72.self_attn.rotary_emb.inv_freq": "model-32-of-35.safetensors",
730
+ "model.layers.72.mlp.gate_proj.weight": "model-32-of-35.safetensors",
731
+ "model.layers.72.mlp.up_proj.weight": "model-32-of-35.safetensors",
732
+ "model.layers.72.mlp.down_proj.weight": "model-32-of-35.safetensors",
733
+ "model.layers.72.input_layernorm.weight": "model-32-of-35.safetensors",
734
+ "model.layers.72.post_attention_layernorm.weight": "model-32-of-35.safetensors",
735
+ "model.layers.73.self_attn.q_proj.weight": "model-32-of-35.safetensors",
736
+ "model.layers.73.self_attn.k_proj.weight": "model-32-of-35.safetensors",
737
+ "model.layers.73.self_attn.v_proj.weight": "model-32-of-35.safetensors",
738
+ "model.layers.73.self_attn.o_proj.weight": "model-32-of-35.safetensors",
739
+ "model.layers.73.self_attn.rotary_emb.inv_freq": "model-32-of-35.safetensors",
740
+ "model.layers.73.mlp.gate_proj.weight": "model-32-of-35.safetensors",
741
+ "model.layers.73.mlp.up_proj.weight": "model-32-of-35.safetensors",
742
+ "model.layers.73.mlp.down_proj.weight": "model-32-of-35.safetensors",
743
+ "model.layers.73.input_layernorm.weight": "model-32-of-35.safetensors",
744
+ "model.layers.73.post_attention_layernorm.weight": "model-32-of-35.safetensors",
745
+ "model.layers.74.self_attn.q_proj.weight": "model-32-of-35.safetensors",
746
+ "model.layers.74.self_attn.k_proj.weight": "model-32-of-35.safetensors",
747
+ "model.layers.74.self_attn.v_proj.weight": "model-32-of-35.safetensors",
748
+ "model.layers.74.self_attn.o_proj.weight": "model-32-of-35.safetensors",
749
+ "model.layers.74.self_attn.rotary_emb.inv_freq": "model-32-of-35.safetensors",
750
+ "model.layers.74.mlp.gate_proj.weight": "model-33-of-35.safetensors",
751
+ "model.layers.74.mlp.up_proj.weight": "model-33-of-35.safetensors",
752
+ "model.layers.74.mlp.down_proj.weight": "model-33-of-35.safetensors",
753
+ "model.layers.74.input_layernorm.weight": "model-33-of-35.safetensors",
754
+ "model.layers.74.post_attention_layernorm.weight": "model-33-of-35.safetensors",
755
+ "model.layers.75.self_attn.q_proj.weight": "model-33-of-35.safetensors",
756
+ "model.layers.75.self_attn.k_proj.weight": "model-33-of-35.safetensors",
757
+ "model.layers.75.self_attn.v_proj.weight": "model-33-of-35.safetensors",
758
+ "model.layers.75.self_attn.o_proj.weight": "model-33-of-35.safetensors",
759
+ "model.layers.75.self_attn.rotary_emb.inv_freq": "model-33-of-35.safetensors",
760
+ "model.layers.75.mlp.gate_proj.weight": "model-33-of-35.safetensors",
761
+ "model.layers.75.mlp.up_proj.weight": "model-33-of-35.safetensors",
762
+ "model.layers.75.mlp.down_proj.weight": "model-33-of-35.safetensors",
763
+ "model.layers.75.input_layernorm.weight": "model-33-of-35.safetensors",
764
+ "model.layers.75.post_attention_layernorm.weight": "model-33-of-35.safetensors",
765
+ "model.layers.76.self_attn.q_proj.weight": "model-33-of-35.safetensors",
766
+ "model.layers.76.self_attn.k_proj.weight": "model-33-of-35.safetensors",
767
+ "model.layers.76.self_attn.v_proj.weight": "model-33-of-35.safetensors",
768
+ "model.layers.76.self_attn.o_proj.weight": "model-33-of-35.safetensors",
769
+ "model.layers.76.self_attn.rotary_emb.inv_freq": "model-33-of-35.safetensors",
770
+ "model.layers.76.mlp.gate_proj.weight": "model-33-of-35.safetensors",
771
+ "model.layers.76.mlp.up_proj.weight": "model-34-of-35.safetensors",
772
+ "model.layers.76.mlp.down_proj.weight": "model-34-of-35.safetensors",
773
+ "model.layers.76.input_layernorm.weight": "model-34-of-35.safetensors",
774
+ "model.layers.76.post_attention_layernorm.weight": "model-34-of-35.safetensors",
775
+ "model.layers.77.self_attn.q_proj.weight": "model-34-of-35.safetensors",
776
+ "model.layers.77.self_attn.k_proj.weight": "model-34-of-35.safetensors",
777
+ "model.layers.77.self_attn.v_proj.weight": "model-34-of-35.safetensors",
778
+ "model.layers.77.self_attn.o_proj.weight": "model-34-of-35.safetensors",
779
+ "model.layers.77.self_attn.rotary_emb.inv_freq": "model-34-of-35.safetensors",
780
+ "model.layers.77.mlp.gate_proj.weight": "model-34-of-35.safetensors",
781
+ "model.layers.77.mlp.up_proj.weight": "model-34-of-35.safetensors",
782
+ "model.layers.77.mlp.down_proj.weight": "model-34-of-35.safetensors",
783
+ "model.layers.77.input_layernorm.weight": "model-34-of-35.safetensors",
784
+ "model.layers.77.post_attention_layernorm.weight": "model-34-of-35.safetensors",
785
+ "model.layers.78.self_attn.q_proj.weight": "model-34-of-35.safetensors",
786
+ "model.layers.78.self_attn.k_proj.weight": "model-34-of-35.safetensors",
787
+ "model.layers.78.self_attn.v_proj.weight": "model-34-of-35.safetensors",
788
+ "model.layers.78.self_attn.o_proj.weight": "model-34-of-35.safetensors",
789
+ "model.layers.78.self_attn.rotary_emb.inv_freq": "model-34-of-35.safetensors",
790
+ "model.layers.78.mlp.gate_proj.weight": "model-34-of-35.safetensors",
791
+ "model.layers.78.mlp.up_proj.weight": "model-34-of-35.safetensors",
792
+ "model.layers.78.mlp.down_proj.weight": "model-35-of-35.safetensors",
793
+ "model.layers.78.input_layernorm.weight": "model-35-of-35.safetensors",
794
+ "model.layers.78.post_attention_layernorm.weight": "model-35-of-35.safetensors",
795
+ "model.layers.79.self_attn.q_proj.weight": "model-35-of-35.safetensors",
796
+ "model.layers.79.self_attn.k_proj.weight": "model-35-of-35.safetensors",
797
+ "model.layers.79.self_attn.v_proj.weight": "model-35-of-35.safetensors",
798
+ "model.layers.79.self_attn.o_proj.weight": "model-35-of-35.safetensors",
799
+ "model.layers.79.self_attn.rotary_emb.inv_freq": "model-35-of-35.safetensors",
800
+ "model.layers.79.mlp.gate_proj.weight": "model-35-of-35.safetensors",
801
+ "model.layers.79.mlp.up_proj.weight": "model-35-of-35.safetensors",
802
+ "model.layers.79.mlp.down_proj.weight": "model-35-of-35.safetensors",
803
+ "model.layers.79.input_layernorm.weight": "model-35-of-35.safetensors",
804
+ "model.layers.79.post_attention_layernorm.weight": "model-35-of-35.safetensors",
805
+ "model.norm.weight": "model-35-of-35.safetensors",
806
+ "lm_head.weight": "model-35-of-35.safetensors"
807
+ }
808
+ }