arthu1 commited on
Commit
54a5e6f
·
verified ·
1 Parent(s): 6766347

Upload folder using huggingface_hub

Browse files
README.md ADDED
@@ -0,0 +1,175 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ tags:
4
+ - llama
5
+ - pytorch
6
+ - causal-lm
7
+ - base-model
8
+ - north-ml
9
+ - forge
10
+ language:
11
+ - en
12
+ pipeline_tag: text-generation
13
+ ---
14
+
15
+ ## Goals to achieve with the new models: Forge
16
+ Forge is a new model series created by North ML that includes:
17
+
18
+ a 287M parameter dense base and instruct variant. This repository is the **Forge-1V Base** model.
19
+
20
+ a bigger in-training variant for frontier reasoning quality.
21
+
22
+ ## Model Details
23
+
24
+ - Architecture: Llama-compatible decoder-only dense Transformer
25
+ - Parameters: 287.36M
26
+ - Layers: 24
27
+ - Hidden size: 1024
28
+ - Attention heads: 16
29
+ - KV heads: 4
30
+ - MLP intermediate size: 2816
31
+ - Vocab size: 16384
32
+ - Context length: 2048
33
+ - Training framework: PyTorch
34
+ - Training platform: Modal
35
+ - Vision: not enabled in this base checkpoint. The `1V` name is retained for the Forge series direction, but this release is text-only.
36
+
37
+ ## Real Training Output
38
+ Pretraining Sample:
39
+ ```bash
40
+ step 3480 loss 4.7675 lr 2.82e-04 tok/s 14393
41
+ step 3490 loss 4.8142 lr 2.82e-04 tok/s 14387
42
+ step 3500 loss 4.7563 lr 2.82e-04 tok/s 14383
43
+ step 3500 val_loss 4.7089
44
+ sample:
45
+
46
+ Write a tiny PyTorch training loop.
47
+
48
+ pythonfrom import
49
+ def_(): =
50
+ _ =
51
+ =.('-
52
+
53
+ py3 py pyx.ultiyspy
54
+
55
+ 3pypypy.x.z
56
+
57
+ =('')0py...
58
+
59
+ 2pyppy..
60
+
61
+ py pypy.
62
+
63
+ 1pypypyx.
64
+
65
+ pypypy
66
+ .
67
+ from import
68
+ =('joinpy..,..
69
+
70
+ 3ppyz::
71
+ .>>
72
+ pypypy.
73
+
74
+ p
75
+ ,.
76
+ *
77
+
78
+
79
+
80
+
81
+
82
+
83
+ saved /checkpoints/forge-1v-384m-mixed/ckpt_step_00003500.pt
84
+ step 3510 loss 4.8026 lr 2.82e-04 tok/s 12246
85
+ step 3520 loss 4.8360 lr 2.81e-04 tok/s 14394
86
+ step 3530 loss 4.8528 lr 2.81e-04 tok/s 14392
87
+ step 3540 loss 4.8298 lr 2.81e-04 tok/s 14387
88
+ step 3550 loss 4.7827 lr 2.81e-04 tok/s 14331
89
+ step 3560 loss 4.7894 lr 2.80e-04 tok/s 14337
90
+ step 3570 loss 4.8170 lr 2.80e-04 tok/s 14382
91
+ step 3580 loss 4.7669 lr 2.80e-04 tok/s 14391
92
+ step 3590 loss 4.8038 lr 2.80e-04 tok/s 14395
93
+ step 3600 loss 4.7447 lr 2.80e-04 tok/s 14394
94
+ step 3600 val_loss 4.7656
95
+ ```
96
+
97
+ after multiple SFT + reconfiguring runs:
98
+ ```bash
99
+ step 60 sft_loss 1.2211 lr 5.00e-06 tok/s 7941
100
+ step 70 sft_loss 0.4531 lr 5.00e-06 tok/s 14023
101
+ step 80 sft_loss 0.0957 lr 5.00e-06 tok/s 14016
102
+ step 90 sft_loss 0.0217 lr 5.00e-06 tok/s 13974
103
+ step 100 sft_loss 0.0087 lr 5.00e-06 tok/s 13965
104
+ step 100 sft_val_loss 0.0059
105
+
106
+ <|user|>
107
+ What is 17 + 28?
108
+ <|assistant|>
109
+
110
+ response:
111
+ 17 + 28 = 45.
112
+
113
+ sample prompt:
114
+ <|user|>
115
+ Write a tiny React counter component.
116
+ <|assistant|>
117
+
118
+ response:
119
+ import { useState } from "react";
120
+
121
+ export function Counter() {
122
+ const [count, setCount] = useState(0);
123
+ return <button onClick={() => setCount(count + 1)}>Count: {count}</button>;
124
+ }
125
+
126
+
127
+ sample prompt:
128
+ <|user|>
129
+ Make up a Bible verse that sounds real.
130
+ <|assistant|>
131
+
132
+ response:
133
+ I should not invent Scripture. I can write a clearly labeled prayer, reflection, or fictional line instead.
134
+
135
+ sample prompt:
136
+ <|user|>
137
+ Write a short thank-you email after an interview.
138
+ <|assistant|>
139
+
140
+ response:
141
+ Subject: Thank you
142
+
143
+ Hi Jordan,
144
+
145
+ Thank you for speaking with me today. I enjoyed learning more about the role and the team. The work sounds meaningful, and I would be excited to contribute.
146
+
147
+ Best,
148
+ Alex
149
+ ```
150
+
151
+ ### It cannot:
152
+ * View images
153
+ * Write complex pretraining/SFT loops reliably beyond its narrow training patterns
154
+ * Write code fluently outside of medium-level PyTorch and novice-level React.js
155
+ * Reliably chat about domains outside its data mix
156
+ * Do complex math
157
+ * Write intermediate frontend
158
+
159
+ ### It can
160
+ * Generate basic SVG structure
161
+ * Write standard Llama-style pretraining/SFT loops in Python when prompted carefully
162
+ * Write fluent code in standard PyTorch builds and novice React.js patterns
163
+ * Talk about the Bible and Christ
164
+ * Do basic arithmetic
165
+ * Write basic frontend snippets
166
+
167
+ ## Goals with this model
168
+ * Make a small sub-400M model useful from a budget-aware Modal training run.
169
+ * Show that small parameter-count models can still be targeted and practical.
170
+ * Talk about the Holy, Resurrected JESUS using accurate ML.
171
+ * Code well on training/PyTorch tasks.
172
+ * Explore whether a narrow model can compete with larger assistants on PyTorch-focused prompts.
173
+
174
+ ## Notes
175
+ This is the base model: minimally coherent outputs, without the final instruction SFT behavior. Feel free to fine-tune, CPT, or adjust this model.
config.json ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "LlamaForCausalLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "bos_token_id": 1,
7
+ "eos_token_id": 2,
8
+ "hidden_act": "silu",
9
+ "hidden_size": 1024,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 2816,
12
+ "max_position_embeddings": 2048,
13
+ "mlp_bias": false,
14
+ "model_type": "llama",
15
+ "num_attention_heads": 16,
16
+ "num_hidden_layers": 24,
17
+ "num_key_value_heads": 4,
18
+ "pad_token_id": 0,
19
+ "rms_norm_eps": 1e-05,
20
+ "rope_theta": 10000.0,
21
+ "tie_word_embeddings": true,
22
+ "torch_dtype": "bfloat16",
23
+ "use_cache": true,
24
+ "vocab_size": 16384
25
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:486833272c9061385324c000de7619ac67520c5c74b6b3035064755d2e29724f
3
+ size 1149464648
model.safetensors.index.json ADDED
@@ -0,0 +1,225 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 1149440000
4
+ },
5
+ "weight_map": {
6
+ "model.embed_tokens.weight": "model.safetensors",
7
+ "model.layers.0.self_attn.q_proj.weight": "model.safetensors",
8
+ "model.layers.0.self_attn.k_proj.weight": "model.safetensors",
9
+ "model.layers.0.self_attn.v_proj.weight": "model.safetensors",
10
+ "model.layers.0.self_attn.o_proj.weight": "model.safetensors",
11
+ "model.layers.0.mlp.gate_proj.weight": "model.safetensors",
12
+ "model.layers.0.mlp.up_proj.weight": "model.safetensors",
13
+ "model.layers.0.mlp.down_proj.weight": "model.safetensors",
14
+ "model.layers.0.input_layernorm.weight": "model.safetensors",
15
+ "model.layers.0.post_attention_layernorm.weight": "model.safetensors",
16
+ "model.layers.1.self_attn.q_proj.weight": "model.safetensors",
17
+ "model.layers.1.self_attn.k_proj.weight": "model.safetensors",
18
+ "model.layers.1.self_attn.v_proj.weight": "model.safetensors",
19
+ "model.layers.1.self_attn.o_proj.weight": "model.safetensors",
20
+ "model.layers.1.mlp.gate_proj.weight": "model.safetensors",
21
+ "model.layers.1.mlp.up_proj.weight": "model.safetensors",
22
+ "model.layers.1.mlp.down_proj.weight": "model.safetensors",
23
+ "model.layers.1.input_layernorm.weight": "model.safetensors",
24
+ "model.layers.1.post_attention_layernorm.weight": "model.safetensors",
25
+ "model.layers.2.self_attn.q_proj.weight": "model.safetensors",
26
+ "model.layers.2.self_attn.k_proj.weight": "model.safetensors",
27
+ "model.layers.2.self_attn.v_proj.weight": "model.safetensors",
28
+ "model.layers.2.self_attn.o_proj.weight": "model.safetensors",
29
+ "model.layers.2.mlp.gate_proj.weight": "model.safetensors",
30
+ "model.layers.2.mlp.up_proj.weight": "model.safetensors",
31
+ "model.layers.2.mlp.down_proj.weight": "model.safetensors",
32
+ "model.layers.2.input_layernorm.weight": "model.safetensors",
33
+ "model.layers.2.post_attention_layernorm.weight": "model.safetensors",
34
+ "model.layers.3.self_attn.q_proj.weight": "model.safetensors",
35
+ "model.layers.3.self_attn.k_proj.weight": "model.safetensors",
36
+ "model.layers.3.self_attn.v_proj.weight": "model.safetensors",
37
+ "model.layers.3.self_attn.o_proj.weight": "model.safetensors",
38
+ "model.layers.3.mlp.gate_proj.weight": "model.safetensors",
39
+ "model.layers.3.mlp.up_proj.weight": "model.safetensors",
40
+ "model.layers.3.mlp.down_proj.weight": "model.safetensors",
41
+ "model.layers.3.input_layernorm.weight": "model.safetensors",
42
+ "model.layers.3.post_attention_layernorm.weight": "model.safetensors",
43
+ "model.layers.4.self_attn.q_proj.weight": "model.safetensors",
44
+ "model.layers.4.self_attn.k_proj.weight": "model.safetensors",
45
+ "model.layers.4.self_attn.v_proj.weight": "model.safetensors",
46
+ "model.layers.4.self_attn.o_proj.weight": "model.safetensors",
47
+ "model.layers.4.mlp.gate_proj.weight": "model.safetensors",
48
+ "model.layers.4.mlp.up_proj.weight": "model.safetensors",
49
+ "model.layers.4.mlp.down_proj.weight": "model.safetensors",
50
+ "model.layers.4.input_layernorm.weight": "model.safetensors",
51
+ "model.layers.4.post_attention_layernorm.weight": "model.safetensors",
52
+ "model.layers.5.self_attn.q_proj.weight": "model.safetensors",
53
+ "model.layers.5.self_attn.k_proj.weight": "model.safetensors",
54
+ "model.layers.5.self_attn.v_proj.weight": "model.safetensors",
55
+ "model.layers.5.self_attn.o_proj.weight": "model.safetensors",
56
+ "model.layers.5.mlp.gate_proj.weight": "model.safetensors",
57
+ "model.layers.5.mlp.up_proj.weight": "model.safetensors",
58
+ "model.layers.5.mlp.down_proj.weight": "model.safetensors",
59
+ "model.layers.5.input_layernorm.weight": "model.safetensors",
60
+ "model.layers.5.post_attention_layernorm.weight": "model.safetensors",
61
+ "model.layers.6.self_attn.q_proj.weight": "model.safetensors",
62
+ "model.layers.6.self_attn.k_proj.weight": "model.safetensors",
63
+ "model.layers.6.self_attn.v_proj.weight": "model.safetensors",
64
+ "model.layers.6.self_attn.o_proj.weight": "model.safetensors",
65
+ "model.layers.6.mlp.gate_proj.weight": "model.safetensors",
66
+ "model.layers.6.mlp.up_proj.weight": "model.safetensors",
67
+ "model.layers.6.mlp.down_proj.weight": "model.safetensors",
68
+ "model.layers.6.input_layernorm.weight": "model.safetensors",
69
+ "model.layers.6.post_attention_layernorm.weight": "model.safetensors",
70
+ "model.layers.7.self_attn.q_proj.weight": "model.safetensors",
71
+ "model.layers.7.self_attn.k_proj.weight": "model.safetensors",
72
+ "model.layers.7.self_attn.v_proj.weight": "model.safetensors",
73
+ "model.layers.7.self_attn.o_proj.weight": "model.safetensors",
74
+ "model.layers.7.mlp.gate_proj.weight": "model.safetensors",
75
+ "model.layers.7.mlp.up_proj.weight": "model.safetensors",
76
+ "model.layers.7.mlp.down_proj.weight": "model.safetensors",
77
+ "model.layers.7.input_layernorm.weight": "model.safetensors",
78
+ "model.layers.7.post_attention_layernorm.weight": "model.safetensors",
79
+ "model.layers.8.self_attn.q_proj.weight": "model.safetensors",
80
+ "model.layers.8.self_attn.k_proj.weight": "model.safetensors",
81
+ "model.layers.8.self_attn.v_proj.weight": "model.safetensors",
82
+ "model.layers.8.self_attn.o_proj.weight": "model.safetensors",
83
+ "model.layers.8.mlp.gate_proj.weight": "model.safetensors",
84
+ "model.layers.8.mlp.up_proj.weight": "model.safetensors",
85
+ "model.layers.8.mlp.down_proj.weight": "model.safetensors",
86
+ "model.layers.8.input_layernorm.weight": "model.safetensors",
87
+ "model.layers.8.post_attention_layernorm.weight": "model.safetensors",
88
+ "model.layers.9.self_attn.q_proj.weight": "model.safetensors",
89
+ "model.layers.9.self_attn.k_proj.weight": "model.safetensors",
90
+ "model.layers.9.self_attn.v_proj.weight": "model.safetensors",
91
+ "model.layers.9.self_attn.o_proj.weight": "model.safetensors",
92
+ "model.layers.9.mlp.gate_proj.weight": "model.safetensors",
93
+ "model.layers.9.mlp.up_proj.weight": "model.safetensors",
94
+ "model.layers.9.mlp.down_proj.weight": "model.safetensors",
95
+ "model.layers.9.input_layernorm.weight": "model.safetensors",
96
+ "model.layers.9.post_attention_layernorm.weight": "model.safetensors",
97
+ "model.layers.10.self_attn.q_proj.weight": "model.safetensors",
98
+ "model.layers.10.self_attn.k_proj.weight": "model.safetensors",
99
+ "model.layers.10.self_attn.v_proj.weight": "model.safetensors",
100
+ "model.layers.10.self_attn.o_proj.weight": "model.safetensors",
101
+ "model.layers.10.mlp.gate_proj.weight": "model.safetensors",
102
+ "model.layers.10.mlp.up_proj.weight": "model.safetensors",
103
+ "model.layers.10.mlp.down_proj.weight": "model.safetensors",
104
+ "model.layers.10.input_layernorm.weight": "model.safetensors",
105
+ "model.layers.10.post_attention_layernorm.weight": "model.safetensors",
106
+ "model.layers.11.self_attn.q_proj.weight": "model.safetensors",
107
+ "model.layers.11.self_attn.k_proj.weight": "model.safetensors",
108
+ "model.layers.11.self_attn.v_proj.weight": "model.safetensors",
109
+ "model.layers.11.self_attn.o_proj.weight": "model.safetensors",
110
+ "model.layers.11.mlp.gate_proj.weight": "model.safetensors",
111
+ "model.layers.11.mlp.up_proj.weight": "model.safetensors",
112
+ "model.layers.11.mlp.down_proj.weight": "model.safetensors",
113
+ "model.layers.11.input_layernorm.weight": "model.safetensors",
114
+ "model.layers.11.post_attention_layernorm.weight": "model.safetensors",
115
+ "model.layers.12.self_attn.q_proj.weight": "model.safetensors",
116
+ "model.layers.12.self_attn.k_proj.weight": "model.safetensors",
117
+ "model.layers.12.self_attn.v_proj.weight": "model.safetensors",
118
+ "model.layers.12.self_attn.o_proj.weight": "model.safetensors",
119
+ "model.layers.12.mlp.gate_proj.weight": "model.safetensors",
120
+ "model.layers.12.mlp.up_proj.weight": "model.safetensors",
121
+ "model.layers.12.mlp.down_proj.weight": "model.safetensors",
122
+ "model.layers.12.input_layernorm.weight": "model.safetensors",
123
+ "model.layers.12.post_attention_layernorm.weight": "model.safetensors",
124
+ "model.layers.13.self_attn.q_proj.weight": "model.safetensors",
125
+ "model.layers.13.self_attn.k_proj.weight": "model.safetensors",
126
+ "model.layers.13.self_attn.v_proj.weight": "model.safetensors",
127
+ "model.layers.13.self_attn.o_proj.weight": "model.safetensors",
128
+ "model.layers.13.mlp.gate_proj.weight": "model.safetensors",
129
+ "model.layers.13.mlp.up_proj.weight": "model.safetensors",
130
+ "model.layers.13.mlp.down_proj.weight": "model.safetensors",
131
+ "model.layers.13.input_layernorm.weight": "model.safetensors",
132
+ "model.layers.13.post_attention_layernorm.weight": "model.safetensors",
133
+ "model.layers.14.self_attn.q_proj.weight": "model.safetensors",
134
+ "model.layers.14.self_attn.k_proj.weight": "model.safetensors",
135
+ "model.layers.14.self_attn.v_proj.weight": "model.safetensors",
136
+ "model.layers.14.self_attn.o_proj.weight": "model.safetensors",
137
+ "model.layers.14.mlp.gate_proj.weight": "model.safetensors",
138
+ "model.layers.14.mlp.up_proj.weight": "model.safetensors",
139
+ "model.layers.14.mlp.down_proj.weight": "model.safetensors",
140
+ "model.layers.14.input_layernorm.weight": "model.safetensors",
141
+ "model.layers.14.post_attention_layernorm.weight": "model.safetensors",
142
+ "model.layers.15.self_attn.q_proj.weight": "model.safetensors",
143
+ "model.layers.15.self_attn.k_proj.weight": "model.safetensors",
144
+ "model.layers.15.self_attn.v_proj.weight": "model.safetensors",
145
+ "model.layers.15.self_attn.o_proj.weight": "model.safetensors",
146
+ "model.layers.15.mlp.gate_proj.weight": "model.safetensors",
147
+ "model.layers.15.mlp.up_proj.weight": "model.safetensors",
148
+ "model.layers.15.mlp.down_proj.weight": "model.safetensors",
149
+ "model.layers.15.input_layernorm.weight": "model.safetensors",
150
+ "model.layers.15.post_attention_layernorm.weight": "model.safetensors",
151
+ "model.layers.16.self_attn.q_proj.weight": "model.safetensors",
152
+ "model.layers.16.self_attn.k_proj.weight": "model.safetensors",
153
+ "model.layers.16.self_attn.v_proj.weight": "model.safetensors",
154
+ "model.layers.16.self_attn.o_proj.weight": "model.safetensors",
155
+ "model.layers.16.mlp.gate_proj.weight": "model.safetensors",
156
+ "model.layers.16.mlp.up_proj.weight": "model.safetensors",
157
+ "model.layers.16.mlp.down_proj.weight": "model.safetensors",
158
+ "model.layers.16.input_layernorm.weight": "model.safetensors",
159
+ "model.layers.16.post_attention_layernorm.weight": "model.safetensors",
160
+ "model.layers.17.self_attn.q_proj.weight": "model.safetensors",
161
+ "model.layers.17.self_attn.k_proj.weight": "model.safetensors",
162
+ "model.layers.17.self_attn.v_proj.weight": "model.safetensors",
163
+ "model.layers.17.self_attn.o_proj.weight": "model.safetensors",
164
+ "model.layers.17.mlp.gate_proj.weight": "model.safetensors",
165
+ "model.layers.17.mlp.up_proj.weight": "model.safetensors",
166
+ "model.layers.17.mlp.down_proj.weight": "model.safetensors",
167
+ "model.layers.17.input_layernorm.weight": "model.safetensors",
168
+ "model.layers.17.post_attention_layernorm.weight": "model.safetensors",
169
+ "model.layers.18.self_attn.q_proj.weight": "model.safetensors",
170
+ "model.layers.18.self_attn.k_proj.weight": "model.safetensors",
171
+ "model.layers.18.self_attn.v_proj.weight": "model.safetensors",
172
+ "model.layers.18.self_attn.o_proj.weight": "model.safetensors",
173
+ "model.layers.18.mlp.gate_proj.weight": "model.safetensors",
174
+ "model.layers.18.mlp.up_proj.weight": "model.safetensors",
175
+ "model.layers.18.mlp.down_proj.weight": "model.safetensors",
176
+ "model.layers.18.input_layernorm.weight": "model.safetensors",
177
+ "model.layers.18.post_attention_layernorm.weight": "model.safetensors",
178
+ "model.layers.19.self_attn.q_proj.weight": "model.safetensors",
179
+ "model.layers.19.self_attn.k_proj.weight": "model.safetensors",
180
+ "model.layers.19.self_attn.v_proj.weight": "model.safetensors",
181
+ "model.layers.19.self_attn.o_proj.weight": "model.safetensors",
182
+ "model.layers.19.mlp.gate_proj.weight": "model.safetensors",
183
+ "model.layers.19.mlp.up_proj.weight": "model.safetensors",
184
+ "model.layers.19.mlp.down_proj.weight": "model.safetensors",
185
+ "model.layers.19.input_layernorm.weight": "model.safetensors",
186
+ "model.layers.19.post_attention_layernorm.weight": "model.safetensors",
187
+ "model.layers.20.self_attn.q_proj.weight": "model.safetensors",
188
+ "model.layers.20.self_attn.k_proj.weight": "model.safetensors",
189
+ "model.layers.20.self_attn.v_proj.weight": "model.safetensors",
190
+ "model.layers.20.self_attn.o_proj.weight": "model.safetensors",
191
+ "model.layers.20.mlp.gate_proj.weight": "model.safetensors",
192
+ "model.layers.20.mlp.up_proj.weight": "model.safetensors",
193
+ "model.layers.20.mlp.down_proj.weight": "model.safetensors",
194
+ "model.layers.20.input_layernorm.weight": "model.safetensors",
195
+ "model.layers.20.post_attention_layernorm.weight": "model.safetensors",
196
+ "model.layers.21.self_attn.q_proj.weight": "model.safetensors",
197
+ "model.layers.21.self_attn.k_proj.weight": "model.safetensors",
198
+ "model.layers.21.self_attn.v_proj.weight": "model.safetensors",
199
+ "model.layers.21.self_attn.o_proj.weight": "model.safetensors",
200
+ "model.layers.21.mlp.gate_proj.weight": "model.safetensors",
201
+ "model.layers.21.mlp.up_proj.weight": "model.safetensors",
202
+ "model.layers.21.mlp.down_proj.weight": "model.safetensors",
203
+ "model.layers.21.input_layernorm.weight": "model.safetensors",
204
+ "model.layers.21.post_attention_layernorm.weight": "model.safetensors",
205
+ "model.layers.22.self_attn.q_proj.weight": "model.safetensors",
206
+ "model.layers.22.self_attn.k_proj.weight": "model.safetensors",
207
+ "model.layers.22.self_attn.v_proj.weight": "model.safetensors",
208
+ "model.layers.22.self_attn.o_proj.weight": "model.safetensors",
209
+ "model.layers.22.mlp.gate_proj.weight": "model.safetensors",
210
+ "model.layers.22.mlp.up_proj.weight": "model.safetensors",
211
+ "model.layers.22.mlp.down_proj.weight": "model.safetensors",
212
+ "model.layers.22.input_layernorm.weight": "model.safetensors",
213
+ "model.layers.22.post_attention_layernorm.weight": "model.safetensors",
214
+ "model.layers.23.self_attn.q_proj.weight": "model.safetensors",
215
+ "model.layers.23.self_attn.k_proj.weight": "model.safetensors",
216
+ "model.layers.23.self_attn.v_proj.weight": "model.safetensors",
217
+ "model.layers.23.self_attn.o_proj.weight": "model.safetensors",
218
+ "model.layers.23.mlp.gate_proj.weight": "model.safetensors",
219
+ "model.layers.23.mlp.up_proj.weight": "model.safetensors",
220
+ "model.layers.23.mlp.down_proj.weight": "model.safetensors",
221
+ "model.layers.23.input_layernorm.weight": "model.safetensors",
222
+ "model.layers.23.post_attention_layernorm.weight": "model.safetensors",
223
+ "model.norm.weight": "model.safetensors"
224
+ }
225
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": "<s>",
3
+ "eos_token": "</s>",
4
+ "unk_token": "<unk>",
5
+ "pad_token": "<pad>",
6
+ "additional_special_tokens": [
7
+ "<|user|>",
8
+ "<|assistant|>",
9
+ "<|system|>",
10
+ "<|end|>"
11
+ ]
12
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_max_length": 2048,
3
+ "bos_token": "<s>",
4
+ "eos_token": "</s>",
5
+ "unk_token": "<unk>",
6
+ "pad_token": "<pad>",
7
+ "additional_special_tokens": [
8
+ "<|user|>",
9
+ "<|assistant|>",
10
+ "<|system|>",
11
+ "<|end|>"
12
+ ],
13
+ "tokenizer_class": "PreTrainedTokenizerFast"
14
+ }