Orion-zhen commited on
Commit
8f4c544
·
verified ·
1 Parent(s): 6966233

Upload 3 files

Browse files
models/Jamba/config.json CHANGED
@@ -0,0 +1,41 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "JambaForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "attn_layer_offset": 4,
7
+ "attn_layer_period": 8,
8
+ "bos_token_id": 1,
9
+ "eos_token_id": 2,
10
+ "expert_layer_offset": 1,
11
+ "expert_layer_period": 2,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 8192,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 24576,
16
+ "mamba_conv_bias": true,
17
+ "mamba_d_conv": 4,
18
+ "mamba_d_state": 16,
19
+ "mamba_dt_rank": 512,
20
+ "mamba_expand": 2,
21
+ "mamba_proj_bias": false,
22
+ "max_position_embeddings": 262144,
23
+ "model_type": "jamba",
24
+ "num_attention_heads": 64,
25
+ "num_experts": 16,
26
+ "num_experts_per_tok": 2,
27
+ "num_hidden_layers": 72,
28
+ "num_key_value_heads": 8,
29
+ "num_logits_to_keep": 1,
30
+ "output_router_logits": false,
31
+ "pad_token_id": 0,
32
+ "rms_norm_eps": 1e-06,
33
+ "router_aux_loss_coef": 0.001,
34
+ "sliding_window": null,
35
+ "tie_word_embeddings": false,
36
+ "torch_dtype": "bfloat16",
37
+ "transformers_version": "4.41.2",
38
+ "use_cache": true,
39
+ "use_mamba_kernels": true,
40
+ "vocab_size": 65536
41
+ }
models/Jamba/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
models/Jamba/tokenizer_config.json ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": true,
3
+ "add_eos_token": false,
4
+ "added_tokens_decoder": {
5
+ "0": {
6
+ "content": "<|pad|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "1": {
14
+ "content": "<|startoftext|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "2": {
22
+ "content": "<|endoftext|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "3": {
30
+ "content": "<|unk|>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "518": {
38
+ "content": "<|eom|>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": true
44
+ },
45
+ "519": {
46
+ "content": "<|bom|>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": true
52
+ },
53
+ "520": {
54
+ "content": "<|system|>",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": true
60
+ },
61
+ "521": {
62
+ "content": "<|user|>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": true
68
+ },
69
+ "522": {
70
+ "content": "<|assistant|>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": true
76
+ },
77
+ "523": {
78
+ "content": "<|tool|>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": true
84
+ },
85
+ "524": {
86
+ "content": "<documents>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": true
92
+ },
93
+ "525": {
94
+ "content": "</documents>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": true
100
+ },
101
+ "526": {
102
+ "content": "<tool_definitions>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": true
108
+ },
109
+ "527": {
110
+ "content": "</tool_definitions>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": true
116
+ },
117
+ "528": {
118
+ "content": "<active_output_modes>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": true
124
+ },
125
+ "529": {
126
+ "content": "</active_output_modes>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": true
132
+ },
133
+ "530": {
134
+ "content": "<citations>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": true
140
+ },
141
+ "531": {
142
+ "content": "</citations>",
143
+ "lstrip": false,
144
+ "normalized": false,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": true
148
+ },
149
+ "532": {
150
+ "content": "<tool_calls>",
151
+ "lstrip": false,
152
+ "normalized": false,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": true
156
+ },
157
+ "533": {
158
+ "content": "</tool_calls>",
159
+ "lstrip": false,
160
+ "normalized": false,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": true
164
+ }
165
+ },
166
+ "additional_special_tokens": [
167
+ "<|eom|>",
168
+ "<|bom|>",
169
+ "<|system|>",
170
+ "<|user|>",
171
+ "<|assistant|>",
172
+ "<|tool|>",
173
+ "<documents>",
174
+ "</documents>",
175
+ "<tool_definitions>",
176
+ "</tool_definitions>",
177
+ "<active_output_modes>",
178
+ "</active_output_modes>",
179
+ "<citations>",
180
+ "</citations>",
181
+ "<tool_calls>",
182
+ "</tool_calls>"
183
+ ],
184
+ "bos_token": "<|startoftext|>",
185
+ "chat_template": "{# Variables #}\n{% set ns = namespace(message_count=0, is_last_checked_defined=False) %}\n{##}\n{% set bom_str = bom_str or \"<|bom|>\" %}\n{% set eom_str = eom_str or \"<|eom|>\" %}\n{% set default_system_message = default_system_message or \"\" %}\n{##}\n{% set documents_prefix = \"<documents>\" %}\n{% set documents_suffix = \"</documents>\" %}\n{% set tool_definitions_prefix = \"<tool_definitions>\" %}\n{% set tool_definitions_suffix = \"</tool_definitions>\" %}\n{% set active_modes_prefix = \"<active_output_modes>\" %}\n{% set active_modes_suffix = \"</active_output_modes>\" %}\n{##}\n{% set tool_calls_prefix = \"<tool_calls>\" %}\n{% set tool_calls_suffix = \"</tool_calls>\" %}\n{% set citations_prefix = \"<citations>\" %}\n{% set citations_suffix = \"</citations>\" %}\n{##}\n{% if add_generation_prompt is not defined %}\n {% set add_generation_prompt = True %}\n{% endif %}\n{% set role_to_predict = role_to_predict or \"assistant\" %}\n{% if messages|length > 0 and messages[0].role == \"system\" %}\n {% set system_message = messages[0].content %}\n {% set loop_messages = messages[1:] %}\n{% else %}\n {% set system_message = default_system_message %}\n {% set loop_messages = messages %}\n{% endif %}\n{##}\n{##}\n{# Macros #}\n{% macro handle_tool_definitions(tools) %}\n {{- tool_definitions_prefix -}}\n {{- \"\\n# Tools\" -}}\n {{- \"\\n\\n## Functions\" -}}\n {% for tool in tools %}\n {% set _ = is_param_set(tool, field=\"type\") %}\n {% set is_tool_type_set = ns.is_last_checked_defined %}\n {% if is_tool_type_set %}\n {% if tool.type == \"function\" %}\n {% set tool = tool.function %}\n {% else %}\n {{ raise_exception(\"Currently, the only supported tool type is `function`\") }}\n {% endif %}\n {% endif %}\n {{- \"\\n\\n\" + (tool|tojson(indent=2)) -}}\n {% endfor %}\n {{- \"\\n\" + tool_definitions_suffix -}}\n{% endmacro %}\n{##}\n{% macro handle_first_system_message(system_message, tools) %}\n {{- bom_str + handle_role(\"system\") -}}\n {% set _ = is_param_set(system_message) %}\n {% set is_system_message_set = ns.is_last_checked_defined %}\n {% if is_system_message_set %}\n {{- system_message -}}\n {% endif %}\n {% set _ = is_param_set(tools, check_length=True) %}\n {% set is_tools_set = ns.is_last_checked_defined %}\n {% if is_tools_set %}\n {% if system_message %}\n {{- \"\\n\\n\" -}}\n {% endif %}\n {{- handle_tool_definitions(tools) -}}\n {% endif %}\n {% set ns.message_count = ns.message_count + 1 %}\n{% endmacro %}\n{##}\n{% macro handle_tool_calls(tool_calls) %}\n {{- tool_calls_prefix + \"[\\n\" -}}\n {% for tool_call in tool_calls %}\n {% set _ = is_param_set(tool_call, field=\"function\") %}\n {% set is_tool_call_function_set = ns.is_last_checked_defined %}\n {% if is_tool_call_function_set %}\n {%- set tool_call = tool_call.function %}\n {%- endif %}\n {% set arguments = tool_call.arguments %}\n {% if arguments is not string %}\n {%- set arguments = arguments|tojson -%}\n {%- endif %}\n {{ \"{\\\"name\\\": \\\"\" + tool_call.name + \"\\\", \\\"arguments\\\": \" + arguments + \"}\" -}}\n {% if not loop.last %}\n {{- \",\" }}\n {% endif %}\n {% endfor %}\n {{- \"\\n]\" + tool_calls_suffix -}}\n{% endmacro %}\n{##}\n{% macro handle_documents(documents) %}\n {{- documents_prefix -}}\n {{- \"\\n# Documents\" -}}\n {{- \"\\n\\nYou can use the following documents for reference:\" -}}\n {% for doc in documents %}\n {{- \"\\n\\n## Document ID: \" + loop.index0|string -}}\n {% set _ = is_param_set(doc, field=\"title\") %}\n {% set is_doc_title_set = ns.is_last_checked_defined %}\n {% if is_doc_title_set %}\n {{- \"\\nTitle: \" + doc.title -}}\n {% endif %}\n {% for key, value in doc.items() %}\n {% if key not in [\"title\", \"text\"] %}\n {{- \"\\n\" + key|title + \": \" + value|string -}}\n {% endif %}\n {% endfor %}\n {{- \"\\nText: \" + doc.text -}}\n {% endfor %}\n {{- \"\\n\" + documents_suffix -}}\n{% endmacro %}\n{##}\n{% macro handle_knobs(knobs) %}\n {{- active_modes_prefix -}}\n {{- \"\\n# Active Modes\" -}}\n {{ \"\\n\\nThe following modes configure the format or style of your responses. You should adhere to all currently\" -}}\n {{ \" active modes simultaneously.\" -}}\n {% if knobs.citation_mode == \"fast\" %}\n {{- \"\\n\\n## Citation Mode\" -}}\n {{- \"\\n\\nProvide a list of references only for the documents you base your response on. Format your response\" -}}\n {{ \" with the original answer followed by a citation section. Use this template:\" -}}\n {{ \" `{answer}\" + citations_prefix + \"DOCUMENT_IDS\" + citations_suffix + \"`, where DOCUMENT_IDS are the relevant document numbers\" -}}\n {{ \" (e.g. [2, 5, 9]), or [] if the answer cannot be supported by the provided documents.\" -}}\n {% endif %}\n {% if knobs.response_format == \"json_object\" %}\n {{- \"\\n\\n## JSON Mode\" -}}\n {{ \"\\n\\nProvide your response in JSON format. Adhere strictly to any schema given by the user.\" -}}\n {{ \" If an appropriate JSON format exists, use it without modification.\" -}}\n {% endif %}\n {{- \"\\n\" + active_modes_suffix -}}\n{% endmacro %}\n{##}\n{% macro get_last_user_index(messages) %}\n {% set ns.last_user_index = 0 %}\n {% for message in messages %}\n {% if message.role == 'user' %}\n {% set ns.last_user_index = loop.index0 %}\n {% endif %}\n {% endfor %}\n {{- ns.last_user_index -}}\n{% endmacro %}\n{##}\n{% macro handle_last_system_message(documents, knobs, use_documents, use_knobs) %}\n {{- bom_str + handle_role(\"system\") -}}\n {% set macros_to_call = [] %}\n {% set params_for_macros = [] %}\n {% if use_documents %}\n {% set macros_to_call = macros_to_call + [handle_documents] %}\n {% set params_for_macros = params_for_macros + [[documents]] %}\n {% endif %}\n {% if use_knobs %}\n {% set macros_to_call = macros_to_call + [handle_knobs] %}\n {% set params_for_macros = params_for_macros + [[knobs]] %}\n {% endif %}\n {% for i in range(macros_to_call|length) %}\n {% if i > 0 %}\n {{- \"\\n\\n\" -}}\n {% endif %}\n {{- macros_to_call[i](*params_for_macros[i]) -}}\n {% endfor %}\n {% set ns.message_count = ns.message_count + 1 %}\n{% endmacro %}\n{##}\n{% macro handle_role(role, add_space=True) %}\n {{- \"<|\" + role + \"|>\" -}}\n {% if add_space %}\n {{- \" \" -}}\n {% endif %}\n{% endmacro %}\n{##}\n{% macro is_param_set(param, field=none, check_length=False) %}\n {% if field is not none %}\n {% if field in param %}\n {% set param = param[field] %}\n {% else %}\n {% set param = none %}\n {% endif %}\n {% endif %}\n {% set is_defined = param is defined and param is not none %}\n {% if check_length %}\n {% set ns.is_last_checked_defined = is_defined and param|length > 0 %}\n {% else %}\n {% set ns.is_last_checked_defined = is_defined %}\n {% endif %}\n{% endmacro %}\n{##}\n{##}\n{# Template #}\n{% if bos_token is defined and bos_token is not none %}\n {{- bos_token -}}\n{% endif %}\n{% set _ = is_param_set(system_message) %}\n{% set is_system_message_set = ns.is_last_checked_defined %}\n{% set _ = is_param_set(tools, check_length=True) %}\n{% set is_tools_set = ns.is_last_checked_defined %}\n{% set has_system_message = (is_system_message_set or is_tools_set) %}\n{% if has_system_message %}\n {{- handle_first_system_message(system_message, tools) -}}\n{% endif %}\n{% set last_user_index = get_last_user_index(loop_messages)|int %}\n{% for message in loop_messages %}\n {% if loop.index0 == last_user_index %}\n {% set _ = is_param_set(documents, check_length=True) %}\n {% set use_documents = ns.is_last_checked_defined %}\n {% set _ = is_param_set(knobs) %}\n {% set use_knobs = ns.is_last_checked_defined and knobs.is_set %}\n {% set add_last_system_message = use_documents or use_knobs %}\n {% if add_last_system_message %}\n {% if ns.message_count > 0 %}\n {{- eom_str -}}\n {% endif %}\n {{- handle_last_system_message(documents, knobs, use_documents, use_knobs) -}}\n {% endif %}\n {% endif %}\n {% set role = message.role %}\n {% set _ = is_param_set(message, field=\"name\") %}\n {% set is_message_name_set = ns.is_last_checked_defined %}\n {% if is_message_name_set %}\n {% set message_prefix = handle_role(role) + \"(\" + message.name + \")\" %}\n {% else %}\n {% set message_prefix = handle_role(role) %}\n {% endif %}\n {% set content = (message.content or \"\") %}\n {% if content is not string %}\n {% set content = content|tojson %}\n {% endif %}\n {% if ns.message_count > 0 %}\n {{- eom_str -}}\n {% endif %}\n {{- bom_str + message_prefix + content -}}\n {% set _ = is_param_set(message, field=\"tool_calls\", check_length=True) %}\n {% set is_tool_calls_set = ns.is_last_checked_defined %}\n {% if role == \"assistant\" and is_tool_calls_set %}\n {{- handle_tool_calls(message.tool_calls) -}}\n {% endif %}\n {% set _ = is_param_set(message, field=\"citations\", check_length=False) %}\n {% set is_citations_set = ns.is_last_checked_defined %}\n {% if role == \"assistant\" and is_citations_set and knobs.is_set and knobs.citation_mode != \"off\" %}\n {{- citations_prefix + message.citations|map(attribute=\"document_id\")|list|string + citations_suffix -}}\n {% endif %}\n {% set ns.message_count = ns.message_count + 1 %}\n{% endfor %}\n{% if add_generation_prompt %}\n {% if ns.message_count > 0 %}\n {{- eom_str -}}\n {% endif %}\n {{- bom_str + handle_role(role_to_predict, add_space=False) -}}\n {% set _ = is_param_set(generation_preamble) %}\n {% set is_generation_preamble_set = ns.is_last_checked_defined %}\n {% if is_generation_preamble_set and generation_preamble.strip() != \"\" %}\n {{- \" \" + generation_preamble -}}\n {% endif %}\n {% set ns.message_count = ns.message_count + 1 %}\n{% else %}\n {% if ns.message_count > 0 %}\n {{- eom_str -}}\n {% endif %}\n{% endif %}\n",
186
+ "clean_up_tokenization_spaces": false,
187
+ "eos_token": "<|endoftext|>",
188
+ "legacy": true,
189
+ "model_max_length": 1000000000000000019884624838656,
190
+ "pad_token": "<|pad|>",
191
+ "spaces_between_special_tokens": false,
192
+ "tokenizer_class": "LlamaTokenizer",
193
+ "unk_token": "<|unk|>",
194
+ "use_default_system_prompt": false
195
+ }