Hugging Face's logo

wmere
/
falcon_moe_1.5Base_test1

Feature Extraction
Transformers
Safetensors
falcon_h1
custom_code
Model card Files
xet
Community
wmere commited on
Commit
24b89ec
·
verified ·
1 Parent(s): c779afe

Upload FalconH1MoEForCausalLM

Browse files
Files changed (12) hide show
  1. README.md +199 -0
  2. config.json +70 -0
  3. configuration_falcon_h1_moe.py +17 -0
  4. generation_config.json +7 -0
  5. model-00001-of-00006.safetensors +3 -0
  6. model-00002-of-00006.safetensors +3 -0
  7. model-00003-of-00006.safetensors +3 -0
  8. model-00004-of-00006.safetensors +3 -0
  9. model-00005-of-00006.safetensors +3 -0
  10. model-00006-of-00006.safetensors +3 -0
  11. model.safetensors.index.json +947 -0
  12. modeling_falcon_h1_moe.py +260 -0
README.md ADDED
@@ -0,0 +1,199 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ tags: []
4
+ ---
5
+
6
+ # Model Card for Model ID
7
+
8
+ <!-- Provide a quick summary of what the model is/does. -->
9
+
10
+
11
+
12
+ ## Model Details
13
+
14
+ ### Model Description
15
+
16
+ <!-- Provide a longer summary of what this model is. -->
17
+
18
+ This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
19
+
20
+ - **Developed by:** [More Information Needed]
21
+ - **Funded by [optional]:** [More Information Needed]
22
+ - **Shared by [optional]:** [More Information Needed]
23
+ - **Model type:** [More Information Needed]
24
+ - **Language(s) (NLP):** [More Information Needed]
25
+ - **License:** [More Information Needed]
26
+ - **Finetuned from model [optional]:** [More Information Needed]
27
+
28
+ ### Model Sources [optional]
29
+
30
+ <!-- Provide the basic links for the model. -->
31
+
32
+ - **Repository:** [More Information Needed]
33
+ - **Paper [optional]:** [More Information Needed]
34
+ - **Demo [optional]:** [More Information Needed]
35
+
36
+ ## Uses
37
+
38
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
39
+
40
+ ### Direct Use
41
+
42
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
43
+
44
+ [More Information Needed]
45
+
46
+ ### Downstream Use [optional]
47
+
48
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
49
+
50
+ [More Information Needed]
51
+
52
+ ### Out-of-Scope Use
53
+
54
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
55
+
56
+ [More Information Needed]
57
+
58
+ ## Bias, Risks, and Limitations
59
+
60
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
61
+
62
+ [More Information Needed]
63
+
64
+ ### Recommendations
65
+
66
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
67
+
68
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
69
+
70
+ ## How to Get Started with the Model
71
+
72
+ Use the code below to get started with the model.
73
+
74
+ [More Information Needed]
75
+
76
+ ## Training Details
77
+
78
+ ### Training Data
79
+
80
+ <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
81
+
82
+ [More Information Needed]
83
+
84
+ ### Training Procedure
85
+
86
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
87
+
88
+ #### Preprocessing [optional]
89
+
90
+ [More Information Needed]
91
+
92
+
93
+ #### Training Hyperparameters
94
+
95
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
96
+
97
+ #### Speeds, Sizes, Times [optional]
98
+
99
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
100
+
101
+ [More Information Needed]
102
+
103
+ ## Evaluation
104
+
105
+ <!-- This section describes the evaluation protocols and provides the results. -->
106
+
107
+ ### Testing Data, Factors & Metrics
108
+
109
+ #### Testing Data
110
+
111
+ <!-- This should link to a Dataset Card if possible. -->
112
+
113
+ [More Information Needed]
114
+
115
+ #### Factors
116
+
117
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
118
+
119
+ [More Information Needed]
120
+
121
+ #### Metrics
122
+
123
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
124
+
125
+ [More Information Needed]
126
+
127
+ ### Results
128
+
129
+ [More Information Needed]
130
+
131
+ #### Summary
132
+
133
+
134
+
135
+ ## Model Examination [optional]
136
+
137
+ <!-- Relevant interpretability work for the model goes here -->
138
+
139
+ [More Information Needed]
140
+
141
+ ## Environmental Impact
142
+
143
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
144
+
145
+ Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
146
+
147
+ - **Hardware Type:** [More Information Needed]
148
+ - **Hours used:** [More Information Needed]
149
+ - **Cloud Provider:** [More Information Needed]
150
+ - **Compute Region:** [More Information Needed]
151
+ - **Carbon Emitted:** [More Information Needed]
152
+
153
+ ## Technical Specifications [optional]
154
+
155
+ ### Model Architecture and Objective
156
+
157
+ [More Information Needed]
158
+
159
+ ### Compute Infrastructure
160
+
161
+ [More Information Needed]
162
+
163
+ #### Hardware
164
+
165
+ [More Information Needed]
166
+
167
+ #### Software
168
+
169
+ [More Information Needed]
170
+
171
+ ## Citation [optional]
172
+
173
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
174
+
175
+ **BibTeX:**
176
+
177
+ [More Information Needed]
178
+
179
+ **APA:**
180
+
181
+ [More Information Needed]
182
+
183
+ ## Glossary [optional]
184
+
185
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
186
+
187
+ [More Information Needed]
188
+
189
+ ## More Information [optional]
190
+
191
+ [More Information Needed]
192
+
193
+ ## Model Card Authors [optional]
194
+
195
+ [More Information Needed]
196
+
197
+ ## Model Card Contact
198
+
199
+ [More Information Needed]
config.json ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "FalconH1MoEForCausalLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "attention_in_multiplier": 1.0,
8
+ "attention_out_multiplier": 0.234375,
9
+ "attn_layer_indices": null,
10
+ "auto_map": {
11
+ "AutoConfig": "configuration_falcon_h1_moe.FalconH1MoEConfig",
12
+ "AutoModel": "modeling_falcon_h1_moe.FalconH1MoEForCausalLM"
13
+ },
14
+ "bos_token_id": 1,
15
+ "embedding_multiplier": 5.656854249492381,
16
+ "eos_token_id": 11,
17
+ "expert_num": 8,
18
+ "head_dim": 128,
19
+ "hidden_act": "silu",
20
+ "hidden_size": 2048,
21
+ "initializer_range": 0.02,
22
+ "intermediate_size": 4608,
23
+ "key_multiplier": 0.06905339660024876,
24
+ "lm_head_multiplier": 0.01953125,
25
+ "mamba_chunk_size": 128,
26
+ "mamba_conv_bias": true,
27
+ "mamba_d_conv": 4,
28
+ "mamba_d_head": 64,
29
+ "mamba_d_ssm": 3072,
30
+ "mamba_d_state": 256,
31
+ "mamba_expand": 2,
32
+ "mamba_n_groups": 1,
33
+ "mamba_n_heads": 48,
34
+ "mamba_norm_before_gate": false,
35
+ "mamba_proj_bias": false,
36
+ "mamba_rms_norm": true,
37
+ "mamba_use_mlp": true,
38
+ "max_position_embeddings": 131072,
39
+ "mlp_bias": false,
40
+ "mlp_expansion_factor": 8,
41
+ "mlp_multipliers": [
42
+ 0.4419417382415922,
43
+ 0.13020833333333331
44
+ ],
45
+ "model_type": "falcon_h1",
46
+ "num_attention_heads": 8,
47
+ "num_hidden_layers": 24,
48
+ "num_key_value_heads": 2,
49
+ "num_logits_to_keep": 1,
50
+ "pad_token_id": 0,
51
+ "projectors_bias": false,
52
+ "rms_norm_eps": 1e-05,
53
+ "rope_scaling": null,
54
+ "rope_theta": 100000000000.0,
55
+ "ssm_in_multiplier": 0.625,
56
+ "ssm_multipliers": [
57
+ 0.3535533905932738,
58
+ 0.25,
59
+ 0.1767766952966369,
60
+ 0.5,
61
+ 0.3535533905932738
62
+ ],
63
+ "ssm_out_multiplier": 0.11785113019775793,
64
+ "tie_word_embeddings": false,
65
+ "topk": 2,
66
+ "torch_dtype": "float32",
67
+ "transformers_version": "4.55.2",
68
+ "use_cache": true,
69
+ "vocab_size": 65536
70
+ }
configuration_falcon_h1_moe.py ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from transformers import FalconH1Config
2
+
3
+
4
+ """FalconH1MoE model configuration"""
5
+ class FalconH1MoEConfig(FalconH1Config):
6
+ def __init__(
7
+ self,
8
+ expert_num=8,
9
+ topk=2,
10
+ **kwargs,
11
+ ):
12
+ super().__init__(**kwargs)
13
+ self.expert_num = expert_num
14
+ self.topk = topk
15
+
16
+
17
+ __all__ = ["FalconH1MoEConfig"]
generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "eos_token_id": 11,
5
+ "pad_token_id": 0,
6
+ "transformers_version": "4.55.2"
7
+ }
model-00001-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e1e73bfd7ed28c0ae035d598c452953f94722ad1ca72653d46807c5d55e798f
3
+ size 4980963920
model-00002-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d96b3602ff76f63532389048d3099467f057cb53ce2f30b7d31dd7c145381c4b
3
+ size 4998238224
model-00003-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e3d4992f90a2cca7a205ebc2660414cf7654e12176cbd14ee64d0a6ed20c34a3
3
+ size 4998238384
model-00004-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:40e0f8b814e7e098648198b86a639901b9ed2ffb8d51e086b0cd0ea974c5560f
3
+ size 4998238416
model-00005-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cad7e85c27a811871353b9d6db628ad47a87c8a561f1a5e6783d596a06755a51
3
+ size 4733938968
model-00006-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db3f453c5bf4fbc3766066825eeb62e54e86663efeedcd2dced55b4356eddce8
3
+ size 536871040
model.safetensors.index.json ADDED
@@ -0,0 +1,947 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_parameters": 6311593344,
4
+ "total_size": 25246373376
5
+ },
6
+ "weight_map": {
7
+ "lm_head.weight": "model-00006-of-00006.safetensors",
8
+ "model.embed_tokens.weight": "model-00001-of-00006.safetensors",
9
+ "model.final_layernorm.weight": "model-00005-of-00006.safetensors",
10
+ "model.layers.0.feed_forward.experts.0.down_proj.weight": "model-00001-of-00006.safetensors",
11
+ "model.layers.0.feed_forward.experts.0.gate_proj.weight": "model-00001-of-00006.safetensors",
12
+ "model.layers.0.feed_forward.experts.0.up_proj.weight": "model-00001-of-00006.safetensors",
13
+ "model.layers.0.feed_forward.experts.1.down_proj.weight": "model-00001-of-00006.safetensors",
14
+ "model.layers.0.feed_forward.experts.1.gate_proj.weight": "model-00001-of-00006.safetensors",
15
+ "model.layers.0.feed_forward.experts.1.up_proj.weight": "model-00001-of-00006.safetensors",
16
+ "model.layers.0.feed_forward.experts.2.down_proj.weight": "model-00001-of-00006.safetensors",
17
+ "model.layers.0.feed_forward.experts.2.gate_proj.weight": "model-00001-of-00006.safetensors",
18
+ "model.layers.0.feed_forward.experts.2.up_proj.weight": "model-00001-of-00006.safetensors",
19
+ "model.layers.0.feed_forward.experts.3.down_proj.weight": "model-00001-of-00006.safetensors",
20
+ "model.layers.0.feed_forward.experts.3.gate_proj.weight": "model-00001-of-00006.safetensors",
21
+ "model.layers.0.feed_forward.experts.3.up_proj.weight": "model-00001-of-00006.safetensors",
22
+ "model.layers.0.feed_forward.experts.4.down_proj.weight": "model-00001-of-00006.safetensors",
23
+ "model.layers.0.feed_forward.experts.4.gate_proj.weight": "model-00001-of-00006.safetensors",
24
+ "model.layers.0.feed_forward.experts.4.up_proj.weight": "model-00001-of-00006.safetensors",
25
+ "model.layers.0.feed_forward.experts.5.down_proj.weight": "model-00001-of-00006.safetensors",
26
+ "model.layers.0.feed_forward.experts.5.gate_proj.weight": "model-00001-of-00006.safetensors",
27
+ "model.layers.0.feed_forward.experts.5.up_proj.weight": "model-00001-of-00006.safetensors",
28
+ "model.layers.0.feed_forward.experts.6.down_proj.weight": "model-00001-of-00006.safetensors",
29
+ "model.layers.0.feed_forward.experts.6.gate_proj.weight": "model-00001-of-00006.safetensors",
30
+ "model.layers.0.feed_forward.experts.6.up_proj.weight": "model-00001-of-00006.safetensors",
31
+ "model.layers.0.feed_forward.experts.7.down_proj.weight": "model-00001-of-00006.safetensors",
32
+ "model.layers.0.feed_forward.experts.7.gate_proj.weight": "model-00001-of-00006.safetensors",
33
+ "model.layers.0.feed_forward.experts.7.up_proj.weight": "model-00001-of-00006.safetensors",
34
+ "model.layers.0.feed_forward.gate.weight": "model-00001-of-00006.safetensors",
35
+ "model.layers.0.input_layernorm.weight": "model-00001-of-00006.safetensors",
36
+ "model.layers.0.mamba.A_log": "model-00001-of-00006.safetensors",
37
+ "model.layers.0.mamba.D": "model-00001-of-00006.safetensors",
38
+ "model.layers.0.mamba.conv1d.bias": "model-00001-of-00006.safetensors",
39
+ "model.layers.0.mamba.conv1d.weight": "model-00001-of-00006.safetensors",
40
+ "model.layers.0.mamba.dt_bias": "model-00001-of-00006.safetensors",
41
+ "model.layers.0.mamba.in_proj.weight": "model-00001-of-00006.safetensors",
42
+ "model.layers.0.mamba.norm.weight": "model-00001-of-00006.safetensors",
43
+ "model.layers.0.mamba.out_proj.weight": "model-00001-of-00006.safetensors",
44
+ "model.layers.0.pre_ff_layernorm.weight": "model-00001-of-00006.safetensors",
45
+ "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
46
+ "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
47
+ "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
48
+ "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
49
+ "model.layers.1.feed_forward.experts.0.down_proj.weight": "model-00001-of-00006.safetensors",
50
+ "model.layers.1.feed_forward.experts.0.gate_proj.weight": "model-00001-of-00006.safetensors",
51
+ "model.layers.1.feed_forward.experts.0.up_proj.weight": "model-00001-of-00006.safetensors",
52
+ "model.layers.1.feed_forward.experts.1.down_proj.weight": "model-00001-of-00006.safetensors",
53
+ "model.layers.1.feed_forward.experts.1.gate_proj.weight": "model-00001-of-00006.safetensors",
54
+ "model.layers.1.feed_forward.experts.1.up_proj.weight": "model-00001-of-00006.safetensors",
55
+ "model.layers.1.feed_forward.experts.2.down_proj.weight": "model-00001-of-00006.safetensors",
56
+ "model.layers.1.feed_forward.experts.2.gate_proj.weight": "model-00001-of-00006.safetensors",
57
+ "model.layers.1.feed_forward.experts.2.up_proj.weight": "model-00001-of-00006.safetensors",
58
+ "model.layers.1.feed_forward.experts.3.down_proj.weight": "model-00001-of-00006.safetensors",
59
+ "model.layers.1.feed_forward.experts.3.gate_proj.weight": "model-00001-of-00006.safetensors",
60
+ "model.layers.1.feed_forward.experts.3.up_proj.weight": "model-00001-of-00006.safetensors",
61
+ "model.layers.1.feed_forward.experts.4.down_proj.weight": "model-00001-of-00006.safetensors",
62
+ "model.layers.1.feed_forward.experts.4.gate_proj.weight": "model-00001-of-00006.safetensors",
63
+ "model.layers.1.feed_forward.experts.4.up_proj.weight": "model-00001-of-00006.safetensors",
64
+ "model.layers.1.feed_forward.experts.5.down_proj.weight": "model-00001-of-00006.safetensors",
65
+ "model.layers.1.feed_forward.experts.5.gate_proj.weight": "model-00001-of-00006.safetensors",
66
+ "model.layers.1.feed_forward.experts.5.up_proj.weight": "model-00001-of-00006.safetensors",
67
+ "model.layers.1.feed_forward.experts.6.down_proj.weight": "model-00001-of-00006.safetensors",
68
+ "model.layers.1.feed_forward.experts.6.gate_proj.weight": "model-00001-of-00006.safetensors",
69
+ "model.layers.1.feed_forward.experts.6.up_proj.weight": "model-00001-of-00006.safetensors",
70
+ "model.layers.1.feed_forward.experts.7.down_proj.weight": "model-00001-of-00006.safetensors",
71
+ "model.layers.1.feed_forward.experts.7.gate_proj.weight": "model-00001-of-00006.safetensors",
72
+ "model.layers.1.feed_forward.experts.7.up_proj.weight": "model-00001-of-00006.safetensors",
73
+ "model.layers.1.feed_forward.gate.weight": "model-00001-of-00006.safetensors",
74
+ "model.layers.1.input_layernorm.weight": "model-00001-of-00006.safetensors",
75
+ "model.layers.1.mamba.A_log": "model-00001-of-00006.safetensors",
76
+ "model.layers.1.mamba.D": "model-00001-of-00006.safetensors",
77
+ "model.layers.1.mamba.conv1d.bias": "model-00001-of-00006.safetensors",
78
+ "model.layers.1.mamba.conv1d.weight": "model-00001-of-00006.safetensors",
79
+ "model.layers.1.mamba.dt_bias": "model-00001-of-00006.safetensors",
80
+ "model.layers.1.mamba.in_proj.weight": "model-00001-of-00006.safetensors",
81
+ "model.layers.1.mamba.norm.weight": "model-00001-of-00006.safetensors",
82
+ "model.layers.1.mamba.out_proj.weight": "model-00001-of-00006.safetensors",
83
+ "model.layers.1.pre_ff_layernorm.weight": "model-00001-of-00006.safetensors",
84
+ "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
85
+ "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
86
+ "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
87
+ "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
88
+ "model.layers.10.feed_forward.experts.0.down_proj.weight": "model-00003-of-00006.safetensors",
89
+ "model.layers.10.feed_forward.experts.0.gate_proj.weight": "model-00003-of-00006.safetensors",
90
+ "model.layers.10.feed_forward.experts.0.up_proj.weight": "model-00003-of-00006.safetensors",
91
+ "model.layers.10.feed_forward.experts.1.down_proj.weight": "model-00003-of-00006.safetensors",
92
+ "model.layers.10.feed_forward.experts.1.gate_proj.weight": "model-00003-of-00006.safetensors",
93
+ "model.layers.10.feed_forward.experts.1.up_proj.weight": "model-00003-of-00006.safetensors",
94
+ "model.layers.10.feed_forward.experts.2.down_proj.weight": "model-00003-of-00006.safetensors",
95
+ "model.layers.10.feed_forward.experts.2.gate_proj.weight": "model-00003-of-00006.safetensors",
96
+ "model.layers.10.feed_forward.experts.2.up_proj.weight": "model-00003-of-00006.safetensors",
97
+ "model.layers.10.feed_forward.experts.3.down_proj.weight": "model-00003-of-00006.safetensors",
98
+ "model.layers.10.feed_forward.experts.3.gate_proj.weight": "model-00003-of-00006.safetensors",
99
+ "model.layers.10.feed_forward.experts.3.up_proj.weight": "model-00003-of-00006.safetensors",
100
+ "model.layers.10.feed_forward.experts.4.down_proj.weight": "model-00003-of-00006.safetensors",
101
+ "model.layers.10.feed_forward.experts.4.gate_proj.weight": "model-00003-of-00006.safetensors",
102
+ "model.layers.10.feed_forward.experts.4.up_proj.weight": "model-00003-of-00006.safetensors",
103
+ "model.layers.10.feed_forward.experts.5.down_proj.weight": "model-00003-of-00006.safetensors",
104
+ "model.layers.10.feed_forward.experts.5.gate_proj.weight": "model-00003-of-00006.safetensors",
105
+ "model.layers.10.feed_forward.experts.5.up_proj.weight": "model-00003-of-00006.safetensors",
106
+ "model.layers.10.feed_forward.experts.6.down_proj.weight": "model-00003-of-00006.safetensors",
107
+ "model.layers.10.feed_forward.experts.6.gate_proj.weight": "model-00003-of-00006.safetensors",
108
+ "model.layers.10.feed_forward.experts.6.up_proj.weight": "model-00003-of-00006.safetensors",
109
+ "model.layers.10.feed_forward.experts.7.down_proj.weight": "model-00003-of-00006.safetensors",
110
+ "model.layers.10.feed_forward.experts.7.gate_proj.weight": "model-00003-of-00006.safetensors",
111
+ "model.layers.10.feed_forward.experts.7.up_proj.weight": "model-00003-of-00006.safetensors",
112
+ "model.layers.10.feed_forward.gate.weight": "model-00003-of-00006.safetensors",
113
+ "model.layers.10.input_layernorm.weight": "model-00003-of-00006.safetensors",
114
+ "model.layers.10.mamba.A_log": "model-00003-of-00006.safetensors",
115
+ "model.layers.10.mamba.D": "model-00003-of-00006.safetensors",
116
+ "model.layers.10.mamba.conv1d.bias": "model-00003-of-00006.safetensors",
117
+ "model.layers.10.mamba.conv1d.weight": "model-00003-of-00006.safetensors",
118
+ "model.layers.10.mamba.dt_bias": "model-00003-of-00006.safetensors",
119
+ "model.layers.10.mamba.in_proj.weight": "model-00003-of-00006.safetensors",
120
+ "model.layers.10.mamba.norm.weight": "model-00003-of-00006.safetensors",
121
+ "model.layers.10.mamba.out_proj.weight": "model-00003-of-00006.safetensors",
122
+ "model.layers.10.pre_ff_layernorm.weight": "model-00003-of-00006.safetensors",
123
+ "model.layers.10.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
124
+ "model.layers.10.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
125
+ "model.layers.10.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
126
+ "model.layers.10.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
127
+ "model.layers.11.feed_forward.experts.0.down_proj.weight": "model-00003-of-00006.safetensors",
128
+ "model.layers.11.feed_forward.experts.0.gate_proj.weight": "model-00003-of-00006.safetensors",
129
+ "model.layers.11.feed_forward.experts.0.up_proj.weight": "model-00003-of-00006.safetensors",
130
+ "model.layers.11.feed_forward.experts.1.down_proj.weight": "model-00003-of-00006.safetensors",
131
+ "model.layers.11.feed_forward.experts.1.gate_proj.weight": "model-00003-of-00006.safetensors",
132
+ "model.layers.11.feed_forward.experts.1.up_proj.weight": "model-00003-of-00006.safetensors",
133
+ "model.layers.11.feed_forward.experts.2.down_proj.weight": "model-00003-of-00006.safetensors",
134
+ "model.layers.11.feed_forward.experts.2.gate_proj.weight": "model-00003-of-00006.safetensors",
135
+ "model.layers.11.feed_forward.experts.2.up_proj.weight": "model-00003-of-00006.safetensors",
136
+ "model.layers.11.feed_forward.experts.3.down_proj.weight": "model-00003-of-00006.safetensors",
137
+ "model.layers.11.feed_forward.experts.3.gate_proj.weight": "model-00003-of-00006.safetensors",
138
+ "model.layers.11.feed_forward.experts.3.up_proj.weight": "model-00003-of-00006.safetensors",
139
+ "model.layers.11.feed_forward.experts.4.down_proj.weight": "model-00003-of-00006.safetensors",
140
+ "model.layers.11.feed_forward.experts.4.gate_proj.weight": "model-00003-of-00006.safetensors",
141
+ "model.layers.11.feed_forward.experts.4.up_proj.weight": "model-00003-of-00006.safetensors",
142
+ "model.layers.11.feed_forward.experts.5.down_proj.weight": "model-00003-of-00006.safetensors",
143
+ "model.layers.11.feed_forward.experts.5.gate_proj.weight": "model-00003-of-00006.safetensors",
144
+ "model.layers.11.feed_forward.experts.5.up_proj.weight": "model-00003-of-00006.safetensors",
145
+ "model.layers.11.feed_forward.experts.6.down_proj.weight": "model-00003-of-00006.safetensors",
146
+ "model.layers.11.feed_forward.experts.6.gate_proj.weight": "model-00003-of-00006.safetensors",
147
+ "model.layers.11.feed_forward.experts.6.up_proj.weight": "model-00003-of-00006.safetensors",
148
+ "model.layers.11.feed_forward.experts.7.down_proj.weight": "model-00003-of-00006.safetensors",
149
+ "model.layers.11.feed_forward.experts.7.gate_proj.weight": "model-00003-of-00006.safetensors",
150
+ "model.layers.11.feed_forward.experts.7.up_proj.weight": "model-00003-of-00006.safetensors",
151
+ "model.layers.11.feed_forward.gate.weight": "model-00003-of-00006.safetensors",
152
+ "model.layers.11.input_layernorm.weight": "model-00003-of-00006.safetensors",
153
+ "model.layers.11.mamba.A_log": "model-00003-of-00006.safetensors",
154
+ "model.layers.11.mamba.D": "model-00003-of-00006.safetensors",
155
+ "model.layers.11.mamba.conv1d.bias": "model-00003-of-00006.safetensors",
156
+ "model.layers.11.mamba.conv1d.weight": "model-00003-of-00006.safetensors",
157
+ "model.layers.11.mamba.dt_bias": "model-00003-of-00006.safetensors",
158
+ "model.layers.11.mamba.in_proj.weight": "model-00003-of-00006.safetensors",
159
+ "model.layers.11.mamba.norm.weight": "model-00003-of-00006.safetensors",
160
+ "model.layers.11.mamba.out_proj.weight": "model-00003-of-00006.safetensors",
161
+ "model.layers.11.pre_ff_layernorm.weight": "model-00003-of-00006.safetensors",
162
+ "model.layers.11.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
163
+ "model.layers.11.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
164
+ "model.layers.11.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
165
+ "model.layers.11.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
166
+ "model.layers.12.feed_forward.experts.0.down_proj.weight": "model-00003-of-00006.safetensors",
167
+ "model.layers.12.feed_forward.experts.0.gate_proj.weight": "model-00003-of-00006.safetensors",
168
+ "model.layers.12.feed_forward.experts.0.up_proj.weight": "model-00003-of-00006.safetensors",
169
+ "model.layers.12.feed_forward.experts.1.down_proj.weight": "model-00003-of-00006.safetensors",
170
+ "model.layers.12.feed_forward.experts.1.gate_proj.weight": "model-00003-of-00006.safetensors",
171
+ "model.layers.12.feed_forward.experts.1.up_proj.weight": "model-00003-of-00006.safetensors",
172
+ "model.layers.12.feed_forward.experts.2.down_proj.weight": "model-00003-of-00006.safetensors",
173
+ "model.layers.12.feed_forward.experts.2.gate_proj.weight": "model-00003-of-00006.safetensors",
174
+ "model.layers.12.feed_forward.experts.2.up_proj.weight": "model-00003-of-00006.safetensors",
175
+ "model.layers.12.feed_forward.experts.3.down_proj.weight": "model-00003-of-00006.safetensors",
176
+ "model.layers.12.feed_forward.experts.3.gate_proj.weight": "model-00003-of-00006.safetensors",
177
+ "model.layers.12.feed_forward.experts.3.up_proj.weight": "model-00003-of-00006.safetensors",
178
+ "model.layers.12.feed_forward.experts.4.down_proj.weight": "model-00003-of-00006.safetensors",
179
+ "model.layers.12.feed_forward.experts.4.gate_proj.weight": "model-00003-of-00006.safetensors",
180
+ "model.layers.12.feed_forward.experts.4.up_proj.weight": "model-00003-of-00006.safetensors",
181
+ "model.layers.12.feed_forward.experts.5.down_proj.weight": "model-00003-of-00006.safetensors",
182
+ "model.layers.12.feed_forward.experts.5.gate_proj.weight": "model-00003-of-00006.safetensors",
183
+ "model.layers.12.feed_forward.experts.5.up_proj.weight": "model-00003-of-00006.safetensors",
184
+ "model.layers.12.feed_forward.experts.6.down_proj.weight": "model-00003-of-00006.safetensors",
185
+ "model.layers.12.feed_forward.experts.6.gate_proj.weight": "model-00003-of-00006.safetensors",
186
+ "model.layers.12.feed_forward.experts.6.up_proj.weight": "model-00003-of-00006.safetensors",
187
+ "model.layers.12.feed_forward.experts.7.down_proj.weight": "model-00003-of-00006.safetensors",
188
+ "model.layers.12.feed_forward.experts.7.gate_proj.weight": "model-00003-of-00006.safetensors",
189
+ "model.layers.12.feed_forward.experts.7.up_proj.weight": "model-00003-of-00006.safetensors",
190
+ "model.layers.12.feed_forward.gate.weight": "model-00003-of-00006.safetensors",
191
+ "model.layers.12.input_layernorm.weight": "model-00003-of-00006.safetensors",
192
+ "model.layers.12.mamba.A_log": "model-00003-of-00006.safetensors",
193
+ "model.layers.12.mamba.D": "model-00003-of-00006.safetensors",
194
+ "model.layers.12.mamba.conv1d.bias": "model-00003-of-00006.safetensors",
195
+ "model.layers.12.mamba.conv1d.weight": "model-00003-of-00006.safetensors",
196
+ "model.layers.12.mamba.dt_bias": "model-00003-of-00006.safetensors",
197
+ "model.layers.12.mamba.in_proj.weight": "model-00003-of-00006.safetensors",
198
+ "model.layers.12.mamba.norm.weight": "model-00003-of-00006.safetensors",
199
+ "model.layers.12.mamba.out_proj.weight": "model-00003-of-00006.safetensors",
200
+ "model.layers.12.pre_ff_layernorm.weight": "model-00003-of-00006.safetensors",
201
+ "model.layers.12.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
202
+ "model.layers.12.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
203
+ "model.layers.12.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
204
+ "model.layers.12.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
205
+ "model.layers.13.feed_forward.experts.0.down_proj.weight": "model-00003-of-00006.safetensors",
206
+ "model.layers.13.feed_forward.experts.0.gate_proj.weight": "model-00003-of-00006.safetensors",
207
+ "model.layers.13.feed_forward.experts.0.up_proj.weight": "model-00003-of-00006.safetensors",
208
+ "model.layers.13.feed_forward.experts.1.down_proj.weight": "model-00003-of-00006.safetensors",
209
+ "model.layers.13.feed_forward.experts.1.gate_proj.weight": "model-00003-of-00006.safetensors",
210
+ "model.layers.13.feed_forward.experts.1.up_proj.weight": "model-00003-of-00006.safetensors",
211
+ "model.layers.13.feed_forward.experts.2.down_proj.weight": "model-00003-of-00006.safetensors",
212
+ "model.layers.13.feed_forward.experts.2.gate_proj.weight": "model-00003-of-00006.safetensors",
213
+ "model.layers.13.feed_forward.experts.2.up_proj.weight": "model-00003-of-00006.safetensors",
214
+ "model.layers.13.feed_forward.experts.3.down_proj.weight": "model-00003-of-00006.safetensors",
215
+ "model.layers.13.feed_forward.experts.3.gate_proj.weight": "model-00003-of-00006.safetensors",
216
+ "model.layers.13.feed_forward.experts.3.up_proj.weight": "model-00003-of-00006.safetensors",
217
+ "model.layers.13.feed_forward.experts.4.down_proj.weight": "model-00003-of-00006.safetensors",
218
+ "model.layers.13.feed_forward.experts.4.gate_proj.weight": "model-00003-of-00006.safetensors",
219
+ "model.layers.13.feed_forward.experts.4.up_proj.weight": "model-00003-of-00006.safetensors",
220
+ "model.layers.13.feed_forward.experts.5.down_proj.weight": "model-00003-of-00006.safetensors",
221
+ "model.layers.13.feed_forward.experts.5.gate_proj.weight": "model-00003-of-00006.safetensors",
222
+ "model.layers.13.feed_forward.experts.5.up_proj.weight": "model-00003-of-00006.safetensors",
223
+ "model.layers.13.feed_forward.experts.6.down_proj.weight": "model-00003-of-00006.safetensors",
224
+ "model.layers.13.feed_forward.experts.6.gate_proj.weight": "model-00003-of-00006.safetensors",
225
+ "model.layers.13.feed_forward.experts.6.up_proj.weight": "model-00003-of-00006.safetensors",
226
+ "model.layers.13.feed_forward.experts.7.down_proj.weight": "model-00003-of-00006.safetensors",
227
+ "model.layers.13.feed_forward.experts.7.gate_proj.weight": "model-00003-of-00006.safetensors",
228
+ "model.layers.13.feed_forward.experts.7.up_proj.weight": "model-00003-of-00006.safetensors",
229
+ "model.layers.13.feed_forward.gate.weight": "model-00003-of-00006.safetensors",
230
+ "model.layers.13.input_layernorm.weight": "model-00003-of-00006.safetensors",
231
+ "model.layers.13.mamba.A_log": "model-00003-of-00006.safetensors",
232
+ "model.layers.13.mamba.D": "model-00003-of-00006.safetensors",
233
+ "model.layers.13.mamba.conv1d.bias": "model-00003-of-00006.safetensors",
234
+ "model.layers.13.mamba.conv1d.weight": "model-00003-of-00006.safetensors",
235
+ "model.layers.13.mamba.dt_bias": "model-00003-of-00006.safetensors",
236
+ "model.layers.13.mamba.in_proj.weight": "model-00003-of-00006.safetensors",
237
+ "model.layers.13.mamba.norm.weight": "model-00003-of-00006.safetensors",
238
+ "model.layers.13.mamba.out_proj.weight": "model-00003-of-00006.safetensors",
239
+ "model.layers.13.pre_ff_layernorm.weight": "model-00003-of-00006.safetensors",
240
+ "model.layers.13.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
241
+ "model.layers.13.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
242
+ "model.layers.13.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
243
+ "model.layers.13.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
244
+ "model.layers.14.feed_forward.experts.0.down_proj.weight": "model-00003-of-00006.safetensors",
245
+ "model.layers.14.feed_forward.experts.0.gate_proj.weight": "model-00003-of-00006.safetensors",
246
+ "model.layers.14.feed_forward.experts.0.up_proj.weight": "model-00003-of-00006.safetensors",
247
+ "model.layers.14.feed_forward.experts.1.down_proj.weight": "model-00003-of-00006.safetensors",
248
+ "model.layers.14.feed_forward.experts.1.gate_proj.weight": "model-00003-of-00006.safetensors",
249
+ "model.layers.14.feed_forward.experts.1.up_proj.weight": "model-00003-of-00006.safetensors",
250
+ "model.layers.14.feed_forward.experts.2.down_proj.weight": "model-00003-of-00006.safetensors",
251
+ "model.layers.14.feed_forward.experts.2.gate_proj.weight": "model-00003-of-00006.safetensors",
252
+ "model.layers.14.feed_forward.experts.2.up_proj.weight": "model-00003-of-00006.safetensors",
253
+ "model.layers.14.feed_forward.experts.3.down_proj.weight": "model-00004-of-00006.safetensors",
254
+ "model.layers.14.feed_forward.experts.3.gate_proj.weight": "model-00004-of-00006.safetensors",
255
+ "model.layers.14.feed_forward.experts.3.up_proj.weight": "model-00004-of-00006.safetensors",
256
+ "model.layers.14.feed_forward.experts.4.down_proj.weight": "model-00004-of-00006.safetensors",
257
+ "model.layers.14.feed_forward.experts.4.gate_proj.weight": "model-00004-of-00006.safetensors",
258
+ "model.layers.14.feed_forward.experts.4.up_proj.weight": "model-00004-of-00006.safetensors",
259
+ "model.layers.14.feed_forward.experts.5.down_proj.weight": "model-00004-of-00006.safetensors",
260
+ "model.layers.14.feed_forward.experts.5.gate_proj.weight": "model-00004-of-00006.safetensors",
261
+ "model.layers.14.feed_forward.experts.5.up_proj.weight": "model-00004-of-00006.safetensors",
262
+ "model.layers.14.feed_forward.experts.6.down_proj.weight": "model-00004-of-00006.safetensors",
263
+ "model.layers.14.feed_forward.experts.6.gate_proj.weight": "model-00004-of-00006.safetensors",
264
+ "model.layers.14.feed_forward.experts.6.up_proj.weight": "model-00004-of-00006.safetensors",
265
+ "model.layers.14.feed_forward.experts.7.down_proj.weight": "model-00004-of-00006.safetensors",
266
+ "model.layers.14.feed_forward.experts.7.gate_proj.weight": "model-00004-of-00006.safetensors",
267
+ "model.layers.14.feed_forward.experts.7.up_proj.weight": "model-00004-of-00006.safetensors",
268
+ "model.layers.14.feed_forward.gate.weight": "model-00003-of-00006.safetensors",
269
+ "model.layers.14.input_layernorm.weight": "model-00004-of-00006.safetensors",
270
+ "model.layers.14.mamba.A_log": "model-00004-of-00006.safetensors",
271
+ "model.layers.14.mamba.D": "model-00004-of-00006.safetensors",
272
+ "model.layers.14.mamba.conv1d.bias": "model-00004-of-00006.safetensors",
273
+ "model.layers.14.mamba.conv1d.weight": "model-00004-of-00006.safetensors",
274
+ "model.layers.14.mamba.dt_bias": "model-00004-of-00006.safetensors",
275
+ "model.layers.14.mamba.in_proj.weight": "model-00004-of-00006.safetensors",
276
+ "model.layers.14.mamba.norm.weight": "model-00004-of-00006.safetensors",
277
+ "model.layers.14.mamba.out_proj.weight": "model-00004-of-00006.safetensors",
278
+ "model.layers.14.pre_ff_layernorm.weight": "model-00004-of-00006.safetensors",
279
+ "model.layers.14.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
280
+ "model.layers.14.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
281
+ "model.layers.14.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
282
+ "model.layers.14.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
283
+ "model.layers.15.feed_forward.experts.0.down_proj.weight": "model-00004-of-00006.safetensors",
284
+ "model.layers.15.feed_forward.experts.0.gate_proj.weight": "model-00004-of-00006.safetensors",
285
+ "model.layers.15.feed_forward.experts.0.up_proj.weight": "model-00004-of-00006.safetensors",
286
+ "model.layers.15.feed_forward.experts.1.down_proj.weight": "model-00004-of-00006.safetensors",
287
+ "model.layers.15.feed_forward.experts.1.gate_proj.weight": "model-00004-of-00006.safetensors",
288
+ "model.layers.15.feed_forward.experts.1.up_proj.weight": "model-00004-of-00006.safetensors",
289
+ "model.layers.15.feed_forward.experts.2.down_proj.weight": "model-00004-of-00006.safetensors",
290
+ "model.layers.15.feed_forward.experts.2.gate_proj.weight": "model-00004-of-00006.safetensors",
291
+ "model.layers.15.feed_forward.experts.2.up_proj.weight": "model-00004-of-00006.safetensors",
292
+ "model.layers.15.feed_forward.experts.3.down_proj.weight": "model-00004-of-00006.safetensors",
293
+ "model.layers.15.feed_forward.experts.3.gate_proj.weight": "model-00004-of-00006.safetensors",
294
+ "model.layers.15.feed_forward.experts.3.up_proj.weight": "model-00004-of-00006.safetensors",
295
+ "model.layers.15.feed_forward.experts.4.down_proj.weight": "model-00004-of-00006.safetensors",
296
+ "model.layers.15.feed_forward.experts.4.gate_proj.weight": "model-00004-of-00006.safetensors",
297
+ "model.layers.15.feed_forward.experts.4.up_proj.weight": "model-00004-of-00006.safetensors",
298
+ "model.layers.15.feed_forward.experts.5.down_proj.weight": "model-00004-of-00006.safetensors",
299
+ "model.layers.15.feed_forward.experts.5.gate_proj.weight": "model-00004-of-00006.safetensors",
300
+ "model.layers.15.feed_forward.experts.5.up_proj.weight": "model-00004-of-00006.safetensors",
301
+ "model.layers.15.feed_forward.experts.6.down_proj.weight": "model-00004-of-00006.safetensors",
302
+ "model.layers.15.feed_forward.experts.6.gate_proj.weight": "model-00004-of-00006.safetensors",
303
+ "model.layers.15.feed_forward.experts.6.up_proj.weight": "model-00004-of-00006.safetensors",
304
+ "model.layers.15.feed_forward.experts.7.down_proj.weight": "model-00004-of-00006.safetensors",
305
+ "model.layers.15.feed_forward.experts.7.gate_proj.weight": "model-00004-of-00006.safetensors",
306
+ "model.layers.15.feed_forward.experts.7.up_proj.weight": "model-00004-of-00006.safetensors",
307
+ "model.layers.15.feed_forward.gate.weight": "model-00004-of-00006.safetensors",
308
+ "model.layers.15.input_layernorm.weight": "model-00004-of-00006.safetensors",
309
+ "model.layers.15.mamba.A_log": "model-00004-of-00006.safetensors",
310
+ "model.layers.15.mamba.D": "model-00004-of-00006.safetensors",
311
+ "model.layers.15.mamba.conv1d.bias": "model-00004-of-00006.safetensors",
312
+ "model.layers.15.mamba.conv1d.weight": "model-00004-of-00006.safetensors",
313
+ "model.layers.15.mamba.dt_bias": "model-00004-of-00006.safetensors",
314
+ "model.layers.15.mamba.in_proj.weight": "model-00004-of-00006.safetensors",
315
+ "model.layers.15.mamba.norm.weight": "model-00004-of-00006.safetensors",
316
+ "model.layers.15.mamba.out_proj.weight": "model-00004-of-00006.safetensors",
317
+ "model.layers.15.pre_ff_layernorm.weight": "model-00004-of-00006.safetensors",
318
+ "model.layers.15.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
319
+ "model.layers.15.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
320
+ "model.layers.15.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
321
+ "model.layers.15.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
322
+ "model.layers.16.feed_forward.experts.0.down_proj.weight": "model-00004-of-00006.safetensors",
323
+ "model.layers.16.feed_forward.experts.0.gate_proj.weight": "model-00004-of-00006.safetensors",
324
+ "model.layers.16.feed_forward.experts.0.up_proj.weight": "model-00004-of-00006.safetensors",
325
+ "model.layers.16.feed_forward.experts.1.down_proj.weight": "model-00004-of-00006.safetensors",
326
+ "model.layers.16.feed_forward.experts.1.gate_proj.weight": "model-00004-of-00006.safetensors",
327
+ "model.layers.16.feed_forward.experts.1.up_proj.weight": "model-00004-of-00006.safetensors",
328
+ "model.layers.16.feed_forward.experts.2.down_proj.weight": "model-00004-of-00006.safetensors",
329
+ "model.layers.16.feed_forward.experts.2.gate_proj.weight": "model-00004-of-00006.safetensors",
330
+ "model.layers.16.feed_forward.experts.2.up_proj.weight": "model-00004-of-00006.safetensors",
331
+ "model.layers.16.feed_forward.experts.3.down_proj.weight": "model-00004-of-00006.safetensors",
332
+ "model.layers.16.feed_forward.experts.3.gate_proj.weight": "model-00004-of-00006.safetensors",
333
+ "model.layers.16.feed_forward.experts.3.up_proj.weight": "model-00004-of-00006.safetensors",
334
+ "model.layers.16.feed_forward.experts.4.down_proj.weight": "model-00004-of-00006.safetensors",
335
+ "model.layers.16.feed_forward.experts.4.gate_proj.weight": "model-00004-of-00006.safetensors",
336
+ "model.layers.16.feed_forward.experts.4.up_proj.weight": "model-00004-of-00006.safetensors",
337
+ "model.layers.16.feed_forward.experts.5.down_proj.weight": "model-00004-of-00006.safetensors",
338
+ "model.layers.16.feed_forward.experts.5.gate_proj.weight": "model-00004-of-00006.safetensors",
339
+ "model.layers.16.feed_forward.experts.5.up_proj.weight": "model-00004-of-00006.safetensors",
340
+ "model.layers.16.feed_forward.experts.6.down_proj.weight": "model-00004-of-00006.safetensors",
341
+ "model.layers.16.feed_forward.experts.6.gate_proj.weight": "model-00004-of-00006.safetensors",
342
+ "model.layers.16.feed_forward.experts.6.up_proj.weight": "model-00004-of-00006.safetensors",
343
+ "model.layers.16.feed_forward.experts.7.down_proj.weight": "model-00004-of-00006.safetensors",
344
+ "model.layers.16.feed_forward.experts.7.gate_proj.weight": "model-00004-of-00006.safetensors",
345
+ "model.layers.16.feed_forward.experts.7.up_proj.weight": "model-00004-of-00006.safetensors",
346
+ "model.layers.16.feed_forward.gate.weight": "model-00004-of-00006.safetensors",
347
+ "model.layers.16.input_layernorm.weight": "model-00004-of-00006.safetensors",
348
+ "model.layers.16.mamba.A_log": "model-00004-of-00006.safetensors",
349
+ "model.layers.16.mamba.D": "model-00004-of-00006.safetensors",
350
+ "model.layers.16.mamba.conv1d.bias": "model-00004-of-00006.safetensors",
351
+ "model.layers.16.mamba.conv1d.weight": "model-00004-of-00006.safetensors",
352
+ "model.layers.16.mamba.dt_bias": "model-00004-of-00006.safetensors",
353
+ "model.layers.16.mamba.in_proj.weight": "model-00004-of-00006.safetensors",
354
+ "model.layers.16.mamba.norm.weight": "model-00004-of-00006.safetensors",
355
+ "model.layers.16.mamba.out_proj.weight": "model-00004-of-00006.safetensors",
356
+ "model.layers.16.pre_ff_layernorm.weight": "model-00004-of-00006.safetensors",
357
+ "model.layers.16.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
358
+ "model.layers.16.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
359
+ "model.layers.16.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
360
+ "model.layers.16.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
361
+ "model.layers.17.feed_forward.experts.0.down_proj.weight": "model-00004-of-00006.safetensors",
362
+ "model.layers.17.feed_forward.experts.0.gate_proj.weight": "model-00004-of-00006.safetensors",
363
+ "model.layers.17.feed_forward.experts.0.up_proj.weight": "model-00004-of-00006.safetensors",
364
+ "model.layers.17.feed_forward.experts.1.down_proj.weight": "model-00004-of-00006.safetensors",
365
+ "model.layers.17.feed_forward.experts.1.gate_proj.weight": "model-00004-of-00006.safetensors",
366
+ "model.layers.17.feed_forward.experts.1.up_proj.weight": "model-00004-of-00006.safetensors",
367
+ "model.layers.17.feed_forward.experts.2.down_proj.weight": "model-00004-of-00006.safetensors",
368
+ "model.layers.17.feed_forward.experts.2.gate_proj.weight": "model-00004-of-00006.safetensors",
369
+ "model.layers.17.feed_forward.experts.2.up_proj.weight": "model-00004-of-00006.safetensors",
370
+ "model.layers.17.feed_forward.experts.3.down_proj.weight": "model-00004-of-00006.safetensors",
371
+ "model.layers.17.feed_forward.experts.3.gate_proj.weight": "model-00004-of-00006.safetensors",
372
+ "model.layers.17.feed_forward.experts.3.up_proj.weight": "model-00004-of-00006.safetensors",
373
+ "model.layers.17.feed_forward.experts.4.down_proj.weight": "model-00004-of-00006.safetensors",
374
+ "model.layers.17.feed_forward.experts.4.gate_proj.weight": "model-00004-of-00006.safetensors",
375
+ "model.layers.17.feed_forward.experts.4.up_proj.weight": "model-00004-of-00006.safetensors",
376
+ "model.layers.17.feed_forward.experts.5.down_proj.weight": "model-00004-of-00006.safetensors",
377
+ "model.layers.17.feed_forward.experts.5.gate_proj.weight": "model-00004-of-00006.safetensors",
378
+ "model.layers.17.feed_forward.experts.5.up_proj.weight": "model-00004-of-00006.safetensors",
379
+ "model.layers.17.feed_forward.experts.6.down_proj.weight": "model-00004-of-00006.safetensors",
380
+ "model.layers.17.feed_forward.experts.6.gate_proj.weight": "model-00004-of-00006.safetensors",
381
+ "model.layers.17.feed_forward.experts.6.up_proj.weight": "model-00004-of-00006.safetensors",
382
+ "model.layers.17.feed_forward.experts.7.down_proj.weight": "model-00004-of-00006.safetensors",
383
+ "model.layers.17.feed_forward.experts.7.gate_proj.weight": "model-00004-of-00006.safetensors",
384
+ "model.layers.17.feed_forward.experts.7.up_proj.weight": "model-00004-of-00006.safetensors",
385
+ "model.layers.17.feed_forward.gate.weight": "model-00004-of-00006.safetensors",
386
+ "model.layers.17.input_layernorm.weight": "model-00004-of-00006.safetensors",
387
+ "model.layers.17.mamba.A_log": "model-00004-of-00006.safetensors",
388
+ "model.layers.17.mamba.D": "model-00004-of-00006.safetensors",
389
+ "model.layers.17.mamba.conv1d.bias": "model-00004-of-00006.safetensors",
390
+ "model.layers.17.mamba.conv1d.weight": "model-00004-of-00006.safetensors",
391
+ "model.layers.17.mamba.dt_bias": "model-00004-of-00006.safetensors",
392
+ "model.layers.17.mamba.in_proj.weight": "model-00004-of-00006.safetensors",
393
+ "model.layers.17.mamba.norm.weight": "model-00004-of-00006.safetensors",
394
+ "model.layers.17.mamba.out_proj.weight": "model-00004-of-00006.safetensors",
395
+ "model.layers.17.pre_ff_layernorm.weight": "model-00004-of-00006.safetensors",
396
+ "model.layers.17.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
397
+ "model.layers.17.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
398
+ "model.layers.17.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
399
+ "model.layers.17.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
400
+ "model.layers.18.feed_forward.experts.0.down_proj.weight": "model-00004-of-00006.safetensors",
401
+ "model.layers.18.feed_forward.experts.0.gate_proj.weight": "model-00004-of-00006.safetensors",
402
+ "model.layers.18.feed_forward.experts.0.up_proj.weight": "model-00004-of-00006.safetensors",
403
+ "model.layers.18.feed_forward.experts.1.down_proj.weight": "model-00004-of-00006.safetensors",
404
+ "model.layers.18.feed_forward.experts.1.gate_proj.weight": "model-00004-of-00006.safetensors",
405
+ "model.layers.18.feed_forward.experts.1.up_proj.weight": "model-00004-of-00006.safetensors",
406
+ "model.layers.18.feed_forward.experts.2.down_proj.weight": "model-00004-of-00006.safetensors",
407
+ "model.layers.18.feed_forward.experts.2.gate_proj.weight": "model-00004-of-00006.safetensors",
408
+ "model.layers.18.feed_forward.experts.2.up_proj.weight": "model-00004-of-00006.safetensors",
409
+ "model.layers.18.feed_forward.experts.3.down_proj.weight": "model-00004-of-00006.safetensors",
410
+ "model.layers.18.feed_forward.experts.3.gate_proj.weight": "model-00004-of-00006.safetensors",
411
+ "model.layers.18.feed_forward.experts.3.up_proj.weight": "model-00004-of-00006.safetensors",
412
+ "model.layers.18.feed_forward.experts.4.down_proj.weight": "model-00004-of-00006.safetensors",
413
+ "model.layers.18.feed_forward.experts.4.gate_proj.weight": "model-00004-of-00006.safetensors",
414
+ "model.layers.18.feed_forward.experts.4.up_proj.weight": "model-00004-of-00006.safetensors",
415
+ "model.layers.18.feed_forward.experts.5.down_proj.weight": "model-00004-of-00006.safetensors",
416
+ "model.layers.18.feed_forward.experts.5.gate_proj.weight": "model-00004-of-00006.safetensors",
417
+ "model.layers.18.feed_forward.experts.5.up_proj.weight": "model-00004-of-00006.safetensors",
418
+ "model.layers.18.feed_forward.experts.6.down_proj.weight": "model-00004-of-00006.safetensors",
419
+ "model.layers.18.feed_forward.experts.6.gate_proj.weight": "model-00004-of-00006.safetensors",
420
+ "model.layers.18.feed_forward.experts.6.up_proj.weight": "model-00004-of-00006.safetensors",
421
+ "model.layers.18.feed_forward.experts.7.down_proj.weight": "model-00004-of-00006.safetensors",
422
+ "model.layers.18.feed_forward.experts.7.gate_proj.weight": "model-00004-of-00006.safetensors",
423
+ "model.layers.18.feed_forward.experts.7.up_proj.weight": "model-00004-of-00006.safetensors",
424
+ "model.layers.18.feed_forward.gate.weight": "model-00004-of-00006.safetensors",
425
+ "model.layers.18.input_layernorm.weight": "model-00004-of-00006.safetensors",
426
+ "model.layers.18.mamba.A_log": "model-00004-of-00006.safetensors",
427
+ "model.layers.18.mamba.D": "model-00004-of-00006.safetensors",
428
+ "model.layers.18.mamba.conv1d.bias": "model-00004-of-00006.safetensors",
429
+ "model.layers.18.mamba.conv1d.weight": "model-00004-of-00006.safetensors",
430
+ "model.layers.18.mamba.dt_bias": "model-00004-of-00006.safetensors",
431
+ "model.layers.18.mamba.in_proj.weight": "model-00004-of-00006.safetensors",
432
+ "model.layers.18.mamba.norm.weight": "model-00004-of-00006.safetensors",
433
+ "model.layers.18.mamba.out_proj.weight": "model-00004-of-00006.safetensors",
434
+ "model.layers.18.pre_ff_layernorm.weight": "model-00004-of-00006.safetensors",
435
+ "model.layers.18.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
436
+ "model.layers.18.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
437
+ "model.layers.18.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
438
+ "model.layers.18.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
439
+ "model.layers.19.feed_forward.experts.0.down_proj.weight": "model-00004-of-00006.safetensors",
440
+ "model.layers.19.feed_forward.experts.0.gate_proj.weight": "model-00004-of-00006.safetensors",
441
+ "model.layers.19.feed_forward.experts.0.up_proj.weight": "model-00004-of-00006.safetensors",
442
+ "model.layers.19.feed_forward.experts.1.down_proj.weight": "model-00004-of-00006.safetensors",
443
+ "model.layers.19.feed_forward.experts.1.gate_proj.weight": "model-00004-of-00006.safetensors",
444
+ "model.layers.19.feed_forward.experts.1.up_proj.weight": "model-00004-of-00006.safetensors",
445
+ "model.layers.19.feed_forward.experts.2.down_proj.weight": "model-00005-of-00006.safetensors",
446
+ "model.layers.19.feed_forward.experts.2.gate_proj.weight": "model-00004-of-00006.safetensors",
447
+ "model.layers.19.feed_forward.experts.2.up_proj.weight": "model-00004-of-00006.safetensors",
448
+ "model.layers.19.feed_forward.experts.3.down_proj.weight": "model-00005-of-00006.safetensors",
449
+ "model.layers.19.feed_forward.experts.3.gate_proj.weight": "model-00005-of-00006.safetensors",
450
+ "model.layers.19.feed_forward.experts.3.up_proj.weight": "model-00005-of-00006.safetensors",
451
+ "model.layers.19.feed_forward.experts.4.down_proj.weight": "model-00005-of-00006.safetensors",
452
+ "model.layers.19.feed_forward.experts.4.gate_proj.weight": "model-00005-of-00006.safetensors",
453
+ "model.layers.19.feed_forward.experts.4.up_proj.weight": "model-00005-of-00006.safetensors",
454
+ "model.layers.19.feed_forward.experts.5.down_proj.weight": "model-00005-of-00006.safetensors",
455
+ "model.layers.19.feed_forward.experts.5.gate_proj.weight": "model-00005-of-00006.safetensors",
456
+ "model.layers.19.feed_forward.experts.5.up_proj.weight": "model-00005-of-00006.safetensors",
457
+ "model.layers.19.feed_forward.experts.6.down_proj.weight": "model-00005-of-00006.safetensors",
458
+ "model.layers.19.feed_forward.experts.6.gate_proj.weight": "model-00005-of-00006.safetensors",
459
+ "model.layers.19.feed_forward.experts.6.up_proj.weight": "model-00005-of-00006.safetensors",
460
+ "model.layers.19.feed_forward.experts.7.down_proj.weight": "model-00005-of-00006.safetensors",
461
+ "model.layers.19.feed_forward.experts.7.gate_proj.weight": "model-00005-of-00006.safetensors",
462
+ "model.layers.19.feed_forward.experts.7.up_proj.weight": "model-00005-of-00006.safetensors",
463
+ "model.layers.19.feed_forward.gate.weight": "model-00004-of-00006.safetensors",
464
+ "model.layers.19.input_layernorm.weight": "model-00005-of-00006.safetensors",
465
+ "model.layers.19.mamba.A_log": "model-00005-of-00006.safetensors",
466
+ "model.layers.19.mamba.D": "model-00005-of-00006.safetensors",
467
+ "model.layers.19.mamba.conv1d.bias": "model-00005-of-00006.safetensors",
468
+ "model.layers.19.mamba.conv1d.weight": "model-00005-of-00006.safetensors",
469
+ "model.layers.19.mamba.dt_bias": "model-00005-of-00006.safetensors",
470
+ "model.layers.19.mamba.in_proj.weight": "model-00005-of-00006.safetensors",
471
+ "model.layers.19.mamba.norm.weight": "model-00005-of-00006.safetensors",
472
+ "model.layers.19.mamba.out_proj.weight": "model-00005-of-00006.safetensors",
473
+ "model.layers.19.pre_ff_layernorm.weight": "model-00005-of-00006.safetensors",
474
+ "model.layers.19.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
475
+ "model.layers.19.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
476
+ "model.layers.19.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
477
+ "model.layers.19.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
478
+ "model.layers.2.feed_forward.experts.0.down_proj.weight": "model-00001-of-00006.safetensors",
479
+ "model.layers.2.feed_forward.experts.0.gate_proj.weight": "model-00001-of-00006.safetensors",
480
+ "model.layers.2.feed_forward.experts.0.up_proj.weight": "model-00001-of-00006.safetensors",
481
+ "model.layers.2.feed_forward.experts.1.down_proj.weight": "model-00001-of-00006.safetensors",
482
+ "model.layers.2.feed_forward.experts.1.gate_proj.weight": "model-00001-of-00006.safetensors",
483
+ "model.layers.2.feed_forward.experts.1.up_proj.weight": "model-00001-of-00006.safetensors",
484
+ "model.layers.2.feed_forward.experts.2.down_proj.weight": "model-00001-of-00006.safetensors",
485
+ "model.layers.2.feed_forward.experts.2.gate_proj.weight": "model-00001-of-00006.safetensors",
486
+ "model.layers.2.feed_forward.experts.2.up_proj.weight": "model-00001-of-00006.safetensors",
487
+ "model.layers.2.feed_forward.experts.3.down_proj.weight": "model-00001-of-00006.safetensors",
488
+ "model.layers.2.feed_forward.experts.3.gate_proj.weight": "model-00001-of-00006.safetensors",
489
+ "model.layers.2.feed_forward.experts.3.up_proj.weight": "model-00001-of-00006.safetensors",
490
+ "model.layers.2.feed_forward.experts.4.down_proj.weight": "model-00001-of-00006.safetensors",
491
+ "model.layers.2.feed_forward.experts.4.gate_proj.weight": "model-00001-of-00006.safetensors",
492
+ "model.layers.2.feed_forward.experts.4.up_proj.weight": "model-00001-of-00006.safetensors",
493
+ "model.layers.2.feed_forward.experts.5.down_proj.weight": "model-00001-of-00006.safetensors",
494
+ "model.layers.2.feed_forward.experts.5.gate_proj.weight": "model-00001-of-00006.safetensors",
495
+ "model.layers.2.feed_forward.experts.5.up_proj.weight": "model-00001-of-00006.safetensors",
496
+ "model.layers.2.feed_forward.experts.6.down_proj.weight": "model-00001-of-00006.safetensors",
497
+ "model.layers.2.feed_forward.experts.6.gate_proj.weight": "model-00001-of-00006.safetensors",
498
+ "model.layers.2.feed_forward.experts.6.up_proj.weight": "model-00001-of-00006.safetensors",
499
+ "model.layers.2.feed_forward.experts.7.down_proj.weight": "model-00001-of-00006.safetensors",
500
+ "model.layers.2.feed_forward.experts.7.gate_proj.weight": "model-00001-of-00006.safetensors",
501
+ "model.layers.2.feed_forward.experts.7.up_proj.weight": "model-00001-of-00006.safetensors",
502
+ "model.layers.2.feed_forward.gate.weight": "model-00001-of-00006.safetensors",
503
+ "model.layers.2.input_layernorm.weight": "model-00001-of-00006.safetensors",
504
+ "model.layers.2.mamba.A_log": "model-00001-of-00006.safetensors",
505
+ "model.layers.2.mamba.D": "model-00001-of-00006.safetensors",
506
+ "model.layers.2.mamba.conv1d.bias": "model-00001-of-00006.safetensors",
507
+ "model.layers.2.mamba.conv1d.weight": "model-00001-of-00006.safetensors",
508
+ "model.layers.2.mamba.dt_bias": "model-00001-of-00006.safetensors",
509
+ "model.layers.2.mamba.in_proj.weight": "model-00001-of-00006.safetensors",
510
+ "model.layers.2.mamba.norm.weight": "model-00001-of-00006.safetensors",
511
+ "model.layers.2.mamba.out_proj.weight": "model-00001-of-00006.safetensors",
512
+ "model.layers.2.pre_ff_layernorm.weight": "model-00001-of-00006.safetensors",
513
+ "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
514
+ "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
515
+ "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
516
+ "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
517
+ "model.layers.20.feed_forward.experts.0.down_proj.weight": "model-00005-of-00006.safetensors",
518
+ "model.layers.20.feed_forward.experts.0.gate_proj.weight": "model-00005-of-00006.safetensors",
519
+ "model.layers.20.feed_forward.experts.0.up_proj.weight": "model-00005-of-00006.safetensors",
520
+ "model.layers.20.feed_forward.experts.1.down_proj.weight": "model-00005-of-00006.safetensors",
521
+ "model.layers.20.feed_forward.experts.1.gate_proj.weight": "model-00005-of-00006.safetensors",
522
+ "model.layers.20.feed_forward.experts.1.up_proj.weight": "model-00005-of-00006.safetensors",
523
+ "model.layers.20.feed_forward.experts.2.down_proj.weight": "model-00005-of-00006.safetensors",
524
+ "model.layers.20.feed_forward.experts.2.gate_proj.weight": "model-00005-of-00006.safetensors",
525
+ "model.layers.20.feed_forward.experts.2.up_proj.weight": "model-00005-of-00006.safetensors",
526
+ "model.layers.20.feed_forward.experts.3.down_proj.weight": "model-00005-of-00006.safetensors",
527
+ "model.layers.20.feed_forward.experts.3.gate_proj.weight": "model-00005-of-00006.safetensors",
528
+ "model.layers.20.feed_forward.experts.3.up_proj.weight": "model-00005-of-00006.safetensors",
529
+ "model.layers.20.feed_forward.experts.4.down_proj.weight": "model-00005-of-00006.safetensors",
530
+ "model.layers.20.feed_forward.experts.4.gate_proj.weight": "model-00005-of-00006.safetensors",
531
+ "model.layers.20.feed_forward.experts.4.up_proj.weight": "model-00005-of-00006.safetensors",
532
+ "model.layers.20.feed_forward.experts.5.down_proj.weight": "model-00005-of-00006.safetensors",
533
+ "model.layers.20.feed_forward.experts.5.gate_proj.weight": "model-00005-of-00006.safetensors",
534
+ "model.layers.20.feed_forward.experts.5.up_proj.weight": "model-00005-of-00006.safetensors",
535
+ "model.layers.20.feed_forward.experts.6.down_proj.weight": "model-00005-of-00006.safetensors",
536
+ "model.layers.20.feed_forward.experts.6.gate_proj.weight": "model-00005-of-00006.safetensors",
537
+ "model.layers.20.feed_forward.experts.6.up_proj.weight": "model-00005-of-00006.safetensors",
538
+ "model.layers.20.feed_forward.experts.7.down_proj.weight": "model-00005-of-00006.safetensors",
539
+ "model.layers.20.feed_forward.experts.7.gate_proj.weight": "model-00005-of-00006.safetensors",
540
+ "model.layers.20.feed_forward.experts.7.up_proj.weight": "model-00005-of-00006.safetensors",
541
+ "model.layers.20.feed_forward.gate.weight": "model-00005-of-00006.safetensors",
542
+ "model.layers.20.input_layernorm.weight": "model-00005-of-00006.safetensors",
543
+ "model.layers.20.mamba.A_log": "model-00005-of-00006.safetensors",
544
+ "model.layers.20.mamba.D": "model-00005-of-00006.safetensors",
545
+ "model.layers.20.mamba.conv1d.bias": "model-00005-of-00006.safetensors",
546
+ "model.layers.20.mamba.conv1d.weight": "model-00005-of-00006.safetensors",
547
+ "model.layers.20.mamba.dt_bias": "model-00005-of-00006.safetensors",
548
+ "model.layers.20.mamba.in_proj.weight": "model-00005-of-00006.safetensors",
549
+ "model.layers.20.mamba.norm.weight": "model-00005-of-00006.safetensors",
550
+ "model.layers.20.mamba.out_proj.weight": "model-00005-of-00006.safetensors",
551
+ "model.layers.20.pre_ff_layernorm.weight": "model-00005-of-00006.safetensors",
552
+ "model.layers.20.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
553
+ "model.layers.20.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
554
+ "model.layers.20.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
555
+ "model.layers.20.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
556
+ "model.layers.21.feed_forward.experts.0.down_proj.weight": "model-00005-of-00006.safetensors",
557
+ "model.layers.21.feed_forward.experts.0.gate_proj.weight": "model-00005-of-00006.safetensors",
558
+ "model.layers.21.feed_forward.experts.0.up_proj.weight": "model-00005-of-00006.safetensors",
559
+ "model.layers.21.feed_forward.experts.1.down_proj.weight": "model-00005-of-00006.safetensors",
560
+ "model.layers.21.feed_forward.experts.1.gate_proj.weight": "model-00005-of-00006.safetensors",
561
+ "model.layers.21.feed_forward.experts.1.up_proj.weight": "model-00005-of-00006.safetensors",
562
+ "model.layers.21.feed_forward.experts.2.down_proj.weight": "model-00005-of-00006.safetensors",
563
+ "model.layers.21.feed_forward.experts.2.gate_proj.weight": "model-00005-of-00006.safetensors",
564
+ "model.layers.21.feed_forward.experts.2.up_proj.weight": "model-00005-of-00006.safetensors",
565
+ "model.layers.21.feed_forward.experts.3.down_proj.weight": "model-00005-of-00006.safetensors",
566
+ "model.layers.21.feed_forward.experts.3.gate_proj.weight": "model-00005-of-00006.safetensors",
567
+ "model.layers.21.feed_forward.experts.3.up_proj.weight": "model-00005-of-00006.safetensors",
568
+ "model.layers.21.feed_forward.experts.4.down_proj.weight": "model-00005-of-00006.safetensors",
569
+ "model.layers.21.feed_forward.experts.4.gate_proj.weight": "model-00005-of-00006.safetensors",
570
+ "model.layers.21.feed_forward.experts.4.up_proj.weight": "model-00005-of-00006.safetensors",
571
+ "model.layers.21.feed_forward.experts.5.down_proj.weight": "model-00005-of-00006.safetensors",
572
+ "model.layers.21.feed_forward.experts.5.gate_proj.weight": "model-00005-of-00006.safetensors",
573
+ "model.layers.21.feed_forward.experts.5.up_proj.weight": "model-00005-of-00006.safetensors",
574
+ "model.layers.21.feed_forward.experts.6.down_proj.weight": "model-00005-of-00006.safetensors",
575
+ "model.layers.21.feed_forward.experts.6.gate_proj.weight": "model-00005-of-00006.safetensors",
576
+ "model.layers.21.feed_forward.experts.6.up_proj.weight": "model-00005-of-00006.safetensors",
577
+ "model.layers.21.feed_forward.experts.7.down_proj.weight": "model-00005-of-00006.safetensors",
578
+ "model.layers.21.feed_forward.experts.7.gate_proj.weight": "model-00005-of-00006.safetensors",
579
+ "model.layers.21.feed_forward.experts.7.up_proj.weight": "model-00005-of-00006.safetensors",
580
+ "model.layers.21.feed_forward.gate.weight": "model-00005-of-00006.safetensors",
581
+ "model.layers.21.input_layernorm.weight": "model-00005-of-00006.safetensors",
582
+ "model.layers.21.mamba.A_log": "model-00005-of-00006.safetensors",
583
+ "model.layers.21.mamba.D": "model-00005-of-00006.safetensors",
584
+ "model.layers.21.mamba.conv1d.bias": "model-00005-of-00006.safetensors",
585
+ "model.layers.21.mamba.conv1d.weight": "model-00005-of-00006.safetensors",
586
+ "model.layers.21.mamba.dt_bias": "model-00005-of-00006.safetensors",
587
+ "model.layers.21.mamba.in_proj.weight": "model-00005-of-00006.safetensors",
588
+ "model.layers.21.mamba.norm.weight": "model-00005-of-00006.safetensors",
589
+ "model.layers.21.mamba.out_proj.weight": "model-00005-of-00006.safetensors",
590
+ "model.layers.21.pre_ff_layernorm.weight": "model-00005-of-00006.safetensors",
591
+ "model.layers.21.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
592
+ "model.layers.21.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
593
+ "model.layers.21.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
594
+ "model.layers.21.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
595
+ "model.layers.22.feed_forward.experts.0.down_proj.weight": "model-00005-of-00006.safetensors",
596
+ "model.layers.22.feed_forward.experts.0.gate_proj.weight": "model-00005-of-00006.safetensors",
597
+ "model.layers.22.feed_forward.experts.0.up_proj.weight": "model-00005-of-00006.safetensors",
598
+ "model.layers.22.feed_forward.experts.1.down_proj.weight": "model-00005-of-00006.safetensors",
599
+ "model.layers.22.feed_forward.experts.1.gate_proj.weight": "model-00005-of-00006.safetensors",
600
+ "model.layers.22.feed_forward.experts.1.up_proj.weight": "model-00005-of-00006.safetensors",
601
+ "model.layers.22.feed_forward.experts.2.down_proj.weight": "model-00005-of-00006.safetensors",
602
+ "model.layers.22.feed_forward.experts.2.gate_proj.weight": "model-00005-of-00006.safetensors",
603
+ "model.layers.22.feed_forward.experts.2.up_proj.weight": "model-00005-of-00006.safetensors",
604
+ "model.layers.22.feed_forward.experts.3.down_proj.weight": "model-00005-of-00006.safetensors",
605
+ "model.layers.22.feed_forward.experts.3.gate_proj.weight": "model-00005-of-00006.safetensors",
606
+ "model.layers.22.feed_forward.experts.3.up_proj.weight": "model-00005-of-00006.safetensors",
607
+ "model.layers.22.feed_forward.experts.4.down_proj.weight": "model-00005-of-00006.safetensors",
608
+ "model.layers.22.feed_forward.experts.4.gate_proj.weight": "model-00005-of-00006.safetensors",
609
+ "model.layers.22.feed_forward.experts.4.up_proj.weight": "model-00005-of-00006.safetensors",
610
+ "model.layers.22.feed_forward.experts.5.down_proj.weight": "model-00005-of-00006.safetensors",
611
+ "model.layers.22.feed_forward.experts.5.gate_proj.weight": "model-00005-of-00006.safetensors",
612
+ "model.layers.22.feed_forward.experts.5.up_proj.weight": "model-00005-of-00006.safetensors",
613
+ "model.layers.22.feed_forward.experts.6.down_proj.weight": "model-00005-of-00006.safetensors",
614
+ "model.layers.22.feed_forward.experts.6.gate_proj.weight": "model-00005-of-00006.safetensors",
615
+ "model.layers.22.feed_forward.experts.6.up_proj.weight": "model-00005-of-00006.safetensors",
616
+ "model.layers.22.feed_forward.experts.7.down_proj.weight": "model-00005-of-00006.safetensors",
617
+ "model.layers.22.feed_forward.experts.7.gate_proj.weight": "model-00005-of-00006.safetensors",
618
+ "model.layers.22.feed_forward.experts.7.up_proj.weight": "model-00005-of-00006.safetensors",
619
+ "model.layers.22.feed_forward.gate.weight": "model-00005-of-00006.safetensors",
620
+ "model.layers.22.input_layernorm.weight": "model-00005-of-00006.safetensors",
621
+ "model.layers.22.mamba.A_log": "model-00005-of-00006.safetensors",
622
+ "model.layers.22.mamba.D": "model-00005-of-00006.safetensors",
623
+ "model.layers.22.mamba.conv1d.bias": "model-00005-of-00006.safetensors",
624
+ "model.layers.22.mamba.conv1d.weight": "model-00005-of-00006.safetensors",
625
+ "model.layers.22.mamba.dt_bias": "model-00005-of-00006.safetensors",
626
+ "model.layers.22.mamba.in_proj.weight": "model-00005-of-00006.safetensors",
627
+ "model.layers.22.mamba.norm.weight": "model-00005-of-00006.safetensors",
628
+ "model.layers.22.mamba.out_proj.weight": "model-00005-of-00006.safetensors",
629
+ "model.layers.22.pre_ff_layernorm.weight": "model-00005-of-00006.safetensors",
630
+ "model.layers.22.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
631
+ "model.layers.22.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
632
+ "model.layers.22.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
633
+ "model.layers.22.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
634
+ "model.layers.23.feed_forward.experts.0.down_proj.weight": "model-00005-of-00006.safetensors",
635
+ "model.layers.23.feed_forward.experts.0.gate_proj.weight": "model-00005-of-00006.safetensors",
636
+ "model.layers.23.feed_forward.experts.0.up_proj.weight": "model-00005-of-00006.safetensors",
637
+ "model.layers.23.feed_forward.experts.1.down_proj.weight": "model-00005-of-00006.safetensors",
638
+ "model.layers.23.feed_forward.experts.1.gate_proj.weight": "model-00005-of-00006.safetensors",
639
+ "model.layers.23.feed_forward.experts.1.up_proj.weight": "model-00005-of-00006.safetensors",
640
+ "model.layers.23.feed_forward.experts.2.down_proj.weight": "model-00005-of-00006.safetensors",
641
+ "model.layers.23.feed_forward.experts.2.gate_proj.weight": "model-00005-of-00006.safetensors",
642
+ "model.layers.23.feed_forward.experts.2.up_proj.weight": "model-00005-of-00006.safetensors",
643
+ "model.layers.23.feed_forward.experts.3.down_proj.weight": "model-00005-of-00006.safetensors",
644
+ "model.layers.23.feed_forward.experts.3.gate_proj.weight": "model-00005-of-00006.safetensors",
645
+ "model.layers.23.feed_forward.experts.3.up_proj.weight": "model-00005-of-00006.safetensors",
646
+ "model.layers.23.feed_forward.experts.4.down_proj.weight": "model-00005-of-00006.safetensors",
647
+ "model.layers.23.feed_forward.experts.4.gate_proj.weight": "model-00005-of-00006.safetensors",
648
+ "model.layers.23.feed_forward.experts.4.up_proj.weight": "model-00005-of-00006.safetensors",
649
+ "model.layers.23.feed_forward.experts.5.down_proj.weight": "model-00005-of-00006.safetensors",
650
+ "model.layers.23.feed_forward.experts.5.gate_proj.weight": "model-00005-of-00006.safetensors",
651
+ "model.layers.23.feed_forward.experts.5.up_proj.weight": "model-00005-of-00006.safetensors",
652
+ "model.layers.23.feed_forward.experts.6.down_proj.weight": "model-00005-of-00006.safetensors",
653
+ "model.layers.23.feed_forward.experts.6.gate_proj.weight": "model-00005-of-00006.safetensors",
654
+ "model.layers.23.feed_forward.experts.6.up_proj.weight": "model-00005-of-00006.safetensors",
655
+ "model.layers.23.feed_forward.experts.7.down_proj.weight": "model-00005-of-00006.safetensors",
656
+ "model.layers.23.feed_forward.experts.7.gate_proj.weight": "model-00005-of-00006.safetensors",
657
+ "model.layers.23.feed_forward.experts.7.up_proj.weight": "model-00005-of-00006.safetensors",
658
+ "model.layers.23.feed_forward.gate.weight": "model-00005-of-00006.safetensors",
659
+ "model.layers.23.input_layernorm.weight": "model-00005-of-00006.safetensors",
660
+ "model.layers.23.mamba.A_log": "model-00005-of-00006.safetensors",
661
+ "model.layers.23.mamba.D": "model-00005-of-00006.safetensors",
662
+ "model.layers.23.mamba.conv1d.bias": "model-00005-of-00006.safetensors",
663
+ "model.layers.23.mamba.conv1d.weight": "model-00005-of-00006.safetensors",
664
+ "model.layers.23.mamba.dt_bias": "model-00005-of-00006.safetensors",
665
+ "model.layers.23.mamba.in_proj.weight": "model-00005-of-00006.safetensors",
666
+ "model.layers.23.mamba.norm.weight": "model-00005-of-00006.safetensors",
667
+ "model.layers.23.mamba.out_proj.weight": "model-00005-of-00006.safetensors",
668
+ "model.layers.23.pre_ff_layernorm.weight": "model-00005-of-00006.safetensors",
669
+ "model.layers.23.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
670
+ "model.layers.23.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
671
+ "model.layers.23.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
672
+ "model.layers.23.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
673
+ "model.layers.3.feed_forward.experts.0.down_proj.weight": "model-00001-of-00006.safetensors",
674
+ "model.layers.3.feed_forward.experts.0.gate_proj.weight": "model-00001-of-00006.safetensors",
675
+ "model.layers.3.feed_forward.experts.0.up_proj.weight": "model-00001-of-00006.safetensors",
676
+ "model.layers.3.feed_forward.experts.1.down_proj.weight": "model-00001-of-00006.safetensors",
677
+ "model.layers.3.feed_forward.experts.1.gate_proj.weight": "model-00001-of-00006.safetensors",
678
+ "model.layers.3.feed_forward.experts.1.up_proj.weight": "model-00001-of-00006.safetensors",
679
+ "model.layers.3.feed_forward.experts.2.down_proj.weight": "model-00001-of-00006.safetensors",
680
+ "model.layers.3.feed_forward.experts.2.gate_proj.weight": "model-00001-of-00006.safetensors",
681
+ "model.layers.3.feed_forward.experts.2.up_proj.weight": "model-00001-of-00006.safetensors",
682
+ "model.layers.3.feed_forward.experts.3.down_proj.weight": "model-00001-of-00006.safetensors",
683
+ "model.layers.3.feed_forward.experts.3.gate_proj.weight": "model-00001-of-00006.safetensors",
684
+ "model.layers.3.feed_forward.experts.3.up_proj.weight": "model-00001-of-00006.safetensors",
685
+ "model.layers.3.feed_forward.experts.4.down_proj.weight": "model-00001-of-00006.safetensors",
686
+ "model.layers.3.feed_forward.experts.4.gate_proj.weight": "model-00001-of-00006.safetensors",
687
+ "model.layers.3.feed_forward.experts.4.up_proj.weight": "model-00001-of-00006.safetensors",
688
+ "model.layers.3.feed_forward.experts.5.down_proj.weight": "model-00001-of-00006.safetensors",
689
+ "model.layers.3.feed_forward.experts.5.gate_proj.weight": "model-00001-of-00006.safetensors",
690
+ "model.layers.3.feed_forward.experts.5.up_proj.weight": "model-00001-of-00006.safetensors",
691
+ "model.layers.3.feed_forward.experts.6.down_proj.weight": "model-00001-of-00006.safetensors",
692
+ "model.layers.3.feed_forward.experts.6.gate_proj.weight": "model-00001-of-00006.safetensors",
693
+ "model.layers.3.feed_forward.experts.6.up_proj.weight": "model-00001-of-00006.safetensors",
694
+ "model.layers.3.feed_forward.experts.7.down_proj.weight": "model-00001-of-00006.safetensors",
695
+ "model.layers.3.feed_forward.experts.7.gate_proj.weight": "model-00001-of-00006.safetensors",
696
+ "model.layers.3.feed_forward.experts.7.up_proj.weight": "model-00001-of-00006.safetensors",
697
+ "model.layers.3.feed_forward.gate.weight": "model-00001-of-00006.safetensors",
698
+ "model.layers.3.input_layernorm.weight": "model-00001-of-00006.safetensors",
699
+ "model.layers.3.mamba.A_log": "model-00001-of-00006.safetensors",
700
+ "model.layers.3.mamba.D": "model-00001-of-00006.safetensors",
701
+ "model.layers.3.mamba.conv1d.bias": "model-00001-of-00006.safetensors",
702
+ "model.layers.3.mamba.conv1d.weight": "model-00001-of-00006.safetensors",
703
+ "model.layers.3.mamba.dt_bias": "model-00001-of-00006.safetensors",
704
+ "model.layers.3.mamba.in_proj.weight": "model-00001-of-00006.safetensors",
705
+ "model.layers.3.mamba.norm.weight": "model-00001-of-00006.safetensors",
706
+ "model.layers.3.mamba.out_proj.weight": "model-00001-of-00006.safetensors",
707
+ "model.layers.3.pre_ff_layernorm.weight": "model-00001-of-00006.safetensors",
708
+ "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
709
+ "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
710
+ "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
711
+ "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
712
+ "model.layers.4.feed_forward.experts.0.down_proj.weight": "model-00001-of-00006.safetensors",
713
+ "model.layers.4.feed_forward.experts.0.gate_proj.weight": "model-00001-of-00006.safetensors",
714
+ "model.layers.4.feed_forward.experts.0.up_proj.weight": "model-00001-of-00006.safetensors",
715
+ "model.layers.4.feed_forward.experts.1.down_proj.weight": "model-00001-of-00006.safetensors",
716
+ "model.layers.4.feed_forward.experts.1.gate_proj.weight": "model-00001-of-00006.safetensors",
717
+ "model.layers.4.feed_forward.experts.1.up_proj.weight": "model-00001-of-00006.safetensors",
718
+ "model.layers.4.feed_forward.experts.2.down_proj.weight": "model-00001-of-00006.safetensors",
719
+ "model.layers.4.feed_forward.experts.2.gate_proj.weight": "model-00001-of-00006.safetensors",
720
+ "model.layers.4.feed_forward.experts.2.up_proj.weight": "model-00001-of-00006.safetensors",
721
+ "model.layers.4.feed_forward.experts.3.down_proj.weight": "model-00002-of-00006.safetensors",
722
+ "model.layers.4.feed_forward.experts.3.gate_proj.weight": "model-00001-of-00006.safetensors",
723
+ "model.layers.4.feed_forward.experts.3.up_proj.weight": "model-00001-of-00006.safetensors",
724
+ "model.layers.4.feed_forward.experts.4.down_proj.weight": "model-00002-of-00006.safetensors",
725
+ "model.layers.4.feed_forward.experts.4.gate_proj.weight": "model-00002-of-00006.safetensors",
726
+ "model.layers.4.feed_forward.experts.4.up_proj.weight": "model-00002-of-00006.safetensors",
727
+ "model.layers.4.feed_forward.experts.5.down_proj.weight": "model-00002-of-00006.safetensors",
728
+ "model.layers.4.feed_forward.experts.5.gate_proj.weight": "model-00002-of-00006.safetensors",
729
+ "model.layers.4.feed_forward.experts.5.up_proj.weight": "model-00002-of-00006.safetensors",
730
+ "model.layers.4.feed_forward.experts.6.down_proj.weight": "model-00002-of-00006.safetensors",
731
+ "model.layers.4.feed_forward.experts.6.gate_proj.weight": "model-00002-of-00006.safetensors",
732
+ "model.layers.4.feed_forward.experts.6.up_proj.weight": "model-00002-of-00006.safetensors",
733
+ "model.layers.4.feed_forward.experts.7.down_proj.weight": "model-00002-of-00006.safetensors",
734
+ "model.layers.4.feed_forward.experts.7.gate_proj.weight": "model-00002-of-00006.safetensors",
735
+ "model.layers.4.feed_forward.experts.7.up_proj.weight": "model-00002-of-00006.safetensors",
736
+ "model.layers.4.feed_forward.gate.weight": "model-00001-of-00006.safetensors",
737
+ "model.layers.4.input_layernorm.weight": "model-00002-of-00006.safetensors",
738
+ "model.layers.4.mamba.A_log": "model-00002-of-00006.safetensors",
739
+ "model.layers.4.mamba.D": "model-00002-of-00006.safetensors",
740
+ "model.layers.4.mamba.conv1d.bias": "model-00002-of-00006.safetensors",
741
+ "model.layers.4.mamba.conv1d.weight": "model-00002-of-00006.safetensors",
742
+ "model.layers.4.mamba.dt_bias": "model-00002-of-00006.safetensors",
743
+ "model.layers.4.mamba.in_proj.weight": "model-00002-of-00006.safetensors",
744
+ "model.layers.4.mamba.norm.weight": "model-00002-of-00006.safetensors",
745
+ "model.layers.4.mamba.out_proj.weight": "model-00002-of-00006.safetensors",
746
+ "model.layers.4.pre_ff_layernorm.weight": "model-00002-of-00006.safetensors",
747
+ "model.layers.4.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
748
+ "model.layers.4.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
749
+ "model.layers.4.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
750
+ "model.layers.4.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
751
+ "model.layers.5.feed_forward.experts.0.down_proj.weight": "model-00002-of-00006.safetensors",
752
+ "model.layers.5.feed_forward.experts.0.gate_proj.weight": "model-00002-of-00006.safetensors",
753
+ "model.layers.5.feed_forward.experts.0.up_proj.weight": "model-00002-of-00006.safetensors",
754
+ "model.layers.5.feed_forward.experts.1.down_proj.weight": "model-00002-of-00006.safetensors",
755
+ "model.layers.5.feed_forward.experts.1.gate_proj.weight": "model-00002-of-00006.safetensors",
756
+ "model.layers.5.feed_forward.experts.1.up_proj.weight": "model-00002-of-00006.safetensors",
757
+ "model.layers.5.feed_forward.experts.2.down_proj.weight": "model-00002-of-00006.safetensors",
758
+ "model.layers.5.feed_forward.experts.2.gate_proj.weight": "model-00002-of-00006.safetensors",
759
+ "model.layers.5.feed_forward.experts.2.up_proj.weight": "model-00002-of-00006.safetensors",
760
+ "model.layers.5.feed_forward.experts.3.down_proj.weight": "model-00002-of-00006.safetensors",
761
+ "model.layers.5.feed_forward.experts.3.gate_proj.weight": "model-00002-of-00006.safetensors",
762
+ "model.layers.5.feed_forward.experts.3.up_proj.weight": "model-00002-of-00006.safetensors",
763
+ "model.layers.5.feed_forward.experts.4.down_proj.weight": "model-00002-of-00006.safetensors",
764
+ "model.layers.5.feed_forward.experts.4.gate_proj.weight": "model-00002-of-00006.safetensors",
765
+ "model.layers.5.feed_forward.experts.4.up_proj.weight": "model-00002-of-00006.safetensors",
766
+ "model.layers.5.feed_forward.experts.5.down_proj.weight": "model-00002-of-00006.safetensors",
767
+ "model.layers.5.feed_forward.experts.5.gate_proj.weight": "model-00002-of-00006.safetensors",
768
+ "model.layers.5.feed_forward.experts.5.up_proj.weight": "model-00002-of-00006.safetensors",
769
+ "model.layers.5.feed_forward.experts.6.down_proj.weight": "model-00002-of-00006.safetensors",
770
+ "model.layers.5.feed_forward.experts.6.gate_proj.weight": "model-00002-of-00006.safetensors",
771
+ "model.layers.5.feed_forward.experts.6.up_proj.weight": "model-00002-of-00006.safetensors",
772
+ "model.layers.5.feed_forward.experts.7.down_proj.weight": "model-00002-of-00006.safetensors",
773
+ "model.layers.5.feed_forward.experts.7.gate_proj.weight": "model-00002-of-00006.safetensors",
774
+ "model.layers.5.feed_forward.experts.7.up_proj.weight": "model-00002-of-00006.safetensors",
775
+ "model.layers.5.feed_forward.gate.weight": "model-00002-of-00006.safetensors",
776
+ "model.layers.5.input_layernorm.weight": "model-00002-of-00006.safetensors",
777
+ "model.layers.5.mamba.A_log": "model-00002-of-00006.safetensors",
778
+ "model.layers.5.mamba.D": "model-00002-of-00006.safetensors",
779
+ "model.layers.5.mamba.conv1d.bias": "model-00002-of-00006.safetensors",
780
+ "model.layers.5.mamba.conv1d.weight": "model-00002-of-00006.safetensors",
781
+ "model.layers.5.mamba.dt_bias": "model-00002-of-00006.safetensors",
782
+ "model.layers.5.mamba.in_proj.weight": "model-00002-of-00006.safetensors",
783
+ "model.layers.5.mamba.norm.weight": "model-00002-of-00006.safetensors",
784
+ "model.layers.5.mamba.out_proj.weight": "model-00002-of-00006.safetensors",
785
+ "model.layers.5.pre_ff_layernorm.weight": "model-00002-of-00006.safetensors",
786
+ "model.layers.5.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
787
+ "model.layers.5.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
788
+ "model.layers.5.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
789
+ "model.layers.5.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
790
+ "model.layers.6.feed_forward.experts.0.down_proj.weight": "model-00002-of-00006.safetensors",
791
+ "model.layers.6.feed_forward.experts.0.gate_proj.weight": "model-00002-of-00006.safetensors",
792
+ "model.layers.6.feed_forward.experts.0.up_proj.weight": "model-00002-of-00006.safetensors",
793
+ "model.layers.6.feed_forward.experts.1.down_proj.weight": "model-00002-of-00006.safetensors",
794
+ "model.layers.6.feed_forward.experts.1.gate_proj.weight": "model-00002-of-00006.safetensors",
795
+ "model.layers.6.feed_forward.experts.1.up_proj.weight": "model-00002-of-00006.safetensors",
796
+ "model.layers.6.feed_forward.experts.2.down_proj.weight": "model-00002-of-00006.safetensors",
797
+ "model.layers.6.feed_forward.experts.2.gate_proj.weight": "model-00002-of-00006.safetensors",
798
+ "model.layers.6.feed_forward.experts.2.up_proj.weight": "model-00002-of-00006.safetensors",
799
+ "model.layers.6.feed_forward.experts.3.down_proj.weight": "model-00002-of-00006.safetensors",
800
+ "model.layers.6.feed_forward.experts.3.gate_proj.weight": "model-00002-of-00006.safetensors",
801
+ "model.layers.6.feed_forward.experts.3.up_proj.weight": "model-00002-of-00006.safetensors",
802
+ "model.layers.6.feed_forward.experts.4.down_proj.weight": "model-00002-of-00006.safetensors",
803
+ "model.layers.6.feed_forward.experts.4.gate_proj.weight": "model-00002-of-00006.safetensors",
804
+ "model.layers.6.feed_forward.experts.4.up_proj.weight": "model-00002-of-00006.safetensors",
805
+ "model.layers.6.feed_forward.experts.5.down_proj.weight": "model-00002-of-00006.safetensors",
806
+ "model.layers.6.feed_forward.experts.5.gate_proj.weight": "model-00002-of-00006.safetensors",
807
+ "model.layers.6.feed_forward.experts.5.up_proj.weight": "model-00002-of-00006.safetensors",
808
+ "model.layers.6.feed_forward.experts.6.down_proj.weight": "model-00002-of-00006.safetensors",
809
+ "model.layers.6.feed_forward.experts.6.gate_proj.weight": "model-00002-of-00006.safetensors",
810
+ "model.layers.6.feed_forward.experts.6.up_proj.weight": "model-00002-of-00006.safetensors",
811
+ "model.layers.6.feed_forward.experts.7.down_proj.weight": "model-00002-of-00006.safetensors",
812
+ "model.layers.6.feed_forward.experts.7.gate_proj.weight": "model-00002-of-00006.safetensors",
813
+ "model.layers.6.feed_forward.experts.7.up_proj.weight": "model-00002-of-00006.safetensors",
814
+ "model.layers.6.feed_forward.gate.weight": "model-00002-of-00006.safetensors",
815
+ "model.layers.6.input_layernorm.weight": "model-00002-of-00006.safetensors",
816
+ "model.layers.6.mamba.A_log": "model-00002-of-00006.safetensors",
817
+ "model.layers.6.mamba.D": "model-00002-of-00006.safetensors",
818
+ "model.layers.6.mamba.conv1d.bias": "model-00002-of-00006.safetensors",
819
+ "model.layers.6.mamba.conv1d.weight": "model-00002-of-00006.safetensors",
820
+ "model.layers.6.mamba.dt_bias": "model-00002-of-00006.safetensors",
821
+ "model.layers.6.mamba.in_proj.weight": "model-00002-of-00006.safetensors",
822
+ "model.layers.6.mamba.norm.weight": "model-00002-of-00006.safetensors",
823
+ "model.layers.6.mamba.out_proj.weight": "model-00002-of-00006.safetensors",
824
+ "model.layers.6.pre_ff_layernorm.weight": "model-00002-of-00006.safetensors",
825
+ "model.layers.6.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
826
+ "model.layers.6.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
827
+ "model.layers.6.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
828
+ "model.layers.6.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
829
+ "model.layers.7.feed_forward.experts.0.down_proj.weight": "model-00002-of-00006.safetensors",
830
+ "model.layers.7.feed_forward.experts.0.gate_proj.weight": "model-00002-of-00006.safetensors",
831
+ "model.layers.7.feed_forward.experts.0.up_proj.weight": "model-00002-of-00006.safetensors",
832
+ "model.layers.7.feed_forward.experts.1.down_proj.weight": "model-00002-of-00006.safetensors",
833
+ "model.layers.7.feed_forward.experts.1.gate_proj.weight": "model-00002-of-00006.safetensors",
834
+ "model.layers.7.feed_forward.experts.1.up_proj.weight": "model-00002-of-00006.safetensors",
835
+ "model.layers.7.feed_forward.experts.2.down_proj.weight": "model-00002-of-00006.safetensors",
836
+ "model.layers.7.feed_forward.experts.2.gate_proj.weight": "model-00002-of-00006.safetensors",
837
+ "model.layers.7.feed_forward.experts.2.up_proj.weight": "model-00002-of-00006.safetensors",
838
+ "model.layers.7.feed_forward.experts.3.down_proj.weight": "model-00002-of-00006.safetensors",
839
+ "model.layers.7.feed_forward.experts.3.gate_proj.weight": "model-00002-of-00006.safetensors",
840
+ "model.layers.7.feed_forward.experts.3.up_proj.weight": "model-00002-of-00006.safetensors",
841
+ "model.layers.7.feed_forward.experts.4.down_proj.weight": "model-00002-of-00006.safetensors",
842
+ "model.layers.7.feed_forward.experts.4.gate_proj.weight": "model-00002-of-00006.safetensors",
843
+ "model.layers.7.feed_forward.experts.4.up_proj.weight": "model-00002-of-00006.safetensors",
844
+ "model.layers.7.feed_forward.experts.5.down_proj.weight": "model-00002-of-00006.safetensors",
845
+ "model.layers.7.feed_forward.experts.5.gate_proj.weight": "model-00002-of-00006.safetensors",
846
+ "model.layers.7.feed_forward.experts.5.up_proj.weight": "model-00002-of-00006.safetensors",
847
+ "model.layers.7.feed_forward.experts.6.down_proj.weight": "model-00002-of-00006.safetensors",
848
+ "model.layers.7.feed_forward.experts.6.gate_proj.weight": "model-00002-of-00006.safetensors",
849
+ "model.layers.7.feed_forward.experts.6.up_proj.weight": "model-00002-of-00006.safetensors",
850
+ "model.layers.7.feed_forward.experts.7.down_proj.weight": "model-00002-of-00006.safetensors",
851
+ "model.layers.7.feed_forward.experts.7.gate_proj.weight": "model-00002-of-00006.safetensors",
852
+ "model.layers.7.feed_forward.experts.7.up_proj.weight": "model-00002-of-00006.safetensors",
853
+ "model.layers.7.feed_forward.gate.weight": "model-00002-of-00006.safetensors",
854
+ "model.layers.7.input_layernorm.weight": "model-00002-of-00006.safetensors",
855
+ "model.layers.7.mamba.A_log": "model-00002-of-00006.safetensors",
856
+ "model.layers.7.mamba.D": "model-00002-of-00006.safetensors",
857
+ "model.layers.7.mamba.conv1d.bias": "model-00002-of-00006.safetensors",
858
+ "model.layers.7.mamba.conv1d.weight": "model-00002-of-00006.safetensors",
859
+ "model.layers.7.mamba.dt_bias": "model-00002-of-00006.safetensors",
860
+ "model.layers.7.mamba.in_proj.weight": "model-00002-of-00006.safetensors",
861
+ "model.layers.7.mamba.norm.weight": "model-00002-of-00006.safetensors",
862
+ "model.layers.7.mamba.out_proj.weight": "model-00002-of-00006.safetensors",
863
+ "model.layers.7.pre_ff_layernorm.weight": "model-00002-of-00006.safetensors",
864
+ "model.layers.7.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
865
+ "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
866
+ "model.layers.7.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
867
+ "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
868
+ "model.layers.8.feed_forward.experts.0.down_proj.weight": "model-00002-of-00006.safetensors",
869
+ "model.layers.8.feed_forward.experts.0.gate_proj.weight": "model-00002-of-00006.safetensors",
870
+ "model.layers.8.feed_forward.experts.0.up_proj.weight": "model-00002-of-00006.safetensors",
871
+ "model.layers.8.feed_forward.experts.1.down_proj.weight": "model-00002-of-00006.safetensors",
872
+ "model.layers.8.feed_forward.experts.1.gate_proj.weight": "model-00002-of-00006.safetensors",
873
+ "model.layers.8.feed_forward.experts.1.up_proj.weight": "model-00002-of-00006.safetensors",
874
+ "model.layers.8.feed_forward.experts.2.down_proj.weight": "model-00002-of-00006.safetensors",
875
+ "model.layers.8.feed_forward.experts.2.gate_proj.weight": "model-00002-of-00006.safetensors",
876
+ "model.layers.8.feed_forward.experts.2.up_proj.weight": "model-00002-of-00006.safetensors",
877
+ "model.layers.8.feed_forward.experts.3.down_proj.weight": "model-00002-of-00006.safetensors",
878
+ "model.layers.8.feed_forward.experts.3.gate_proj.weight": "model-00002-of-00006.safetensors",
879
+ "model.layers.8.feed_forward.experts.3.up_proj.weight": "model-00002-of-00006.safetensors",
880
+ "model.layers.8.feed_forward.experts.4.down_proj.weight": "model-00002-of-00006.safetensors",
881
+ "model.layers.8.feed_forward.experts.4.gate_proj.weight": "model-00002-of-00006.safetensors",
882
+ "model.layers.8.feed_forward.experts.4.up_proj.weight": "model-00002-of-00006.safetensors",
883
+ "model.layers.8.feed_forward.experts.5.down_proj.weight": "model-00002-of-00006.safetensors",
884
+ "model.layers.8.feed_forward.experts.5.gate_proj.weight": "model-00002-of-00006.safetensors",
885
+ "model.layers.8.feed_forward.experts.5.up_proj.weight": "model-00002-of-00006.safetensors",
886
+ "model.layers.8.feed_forward.experts.6.down_proj.weight": "model-00002-of-00006.safetensors",
887
+ "model.layers.8.feed_forward.experts.6.gate_proj.weight": "model-00002-of-00006.safetensors",
888
+ "model.layers.8.feed_forward.experts.6.up_proj.weight": "model-00002-of-00006.safetensors",
889
+ "model.layers.8.feed_forward.experts.7.down_proj.weight": "model-00002-of-00006.safetensors",
890
+ "model.layers.8.feed_forward.experts.7.gate_proj.weight": "model-00002-of-00006.safetensors",
891
+ "model.layers.8.feed_forward.experts.7.up_proj.weight": "model-00002-of-00006.safetensors",
892
+ "model.layers.8.feed_forward.gate.weight": "model-00002-of-00006.safetensors",
893
+ "model.layers.8.input_layernorm.weight": "model-00002-of-00006.safetensors",
894
+ "model.layers.8.mamba.A_log": "model-00002-of-00006.safetensors",
895
+ "model.layers.8.mamba.D": "model-00002-of-00006.safetensors",
896
+ "model.layers.8.mamba.conv1d.bias": "model-00002-of-00006.safetensors",
897
+ "model.layers.8.mamba.conv1d.weight": "model-00002-of-00006.safetensors",
898
+ "model.layers.8.mamba.dt_bias": "model-00002-of-00006.safetensors",
899
+ "model.layers.8.mamba.in_proj.weight": "model-00002-of-00006.safetensors",
900
+ "model.layers.8.mamba.norm.weight": "model-00002-of-00006.safetensors",
901
+ "model.layers.8.mamba.out_proj.weight": "model-00002-of-00006.safetensors",
902
+ "model.layers.8.pre_ff_layernorm.weight": "model-00002-of-00006.safetensors",
903
+ "model.layers.8.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
904
+ "model.layers.8.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
905
+ "model.layers.8.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
906
+ "model.layers.8.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
907
+ "model.layers.9.feed_forward.experts.0.down_proj.weight": "model-00002-of-00006.safetensors",
908
+ "model.layers.9.feed_forward.experts.0.gate_proj.weight": "model-00002-of-00006.safetensors",
909
+ "model.layers.9.feed_forward.experts.0.up_proj.weight": "model-00002-of-00006.safetensors",
910
+ "model.layers.9.feed_forward.experts.1.down_proj.weight": "model-00002-of-00006.safetensors",
911
+ "model.layers.9.feed_forward.experts.1.gate_proj.weight": "model-00002-of-00006.safetensors",
912
+ "model.layers.9.feed_forward.experts.1.up_proj.weight": "model-00002-of-00006.safetensors",
913
+ "model.layers.9.feed_forward.experts.2.down_proj.weight": "model-00002-of-00006.safetensors",
914
+ "model.layers.9.feed_forward.experts.2.gate_proj.weight": "model-00002-of-00006.safetensors",
915
+ "model.layers.9.feed_forward.experts.2.up_proj.weight": "model-00002-of-00006.safetensors",
916
+ "model.layers.9.feed_forward.experts.3.down_proj.weight": "model-00003-of-00006.safetensors",
917
+ "model.layers.9.feed_forward.experts.3.gate_proj.weight": "model-00002-of-00006.safetensors",
918
+ "model.layers.9.feed_forward.experts.3.up_proj.weight": "model-00003-of-00006.safetensors",
919
+ "model.layers.9.feed_forward.experts.4.down_proj.weight": "model-00003-of-00006.safetensors",
920
+ "model.layers.9.feed_forward.experts.4.gate_proj.weight": "model-00003-of-00006.safetensors",
921
+ "model.layers.9.feed_forward.experts.4.up_proj.weight": "model-00003-of-00006.safetensors",
922
+ "model.layers.9.feed_forward.experts.5.down_proj.weight": "model-00003-of-00006.safetensors",
923
+ "model.layers.9.feed_forward.experts.5.gate_proj.weight": "model-00003-of-00006.safetensors",
924
+ "model.layers.9.feed_forward.experts.5.up_proj.weight": "model-00003-of-00006.safetensors",
925
+ "model.layers.9.feed_forward.experts.6.down_proj.weight": "model-00003-of-00006.safetensors",
926
+ "model.layers.9.feed_forward.experts.6.gate_proj.weight": "model-00003-of-00006.safetensors",
927
+ "model.layers.9.feed_forward.experts.6.up_proj.weight": "model-00003-of-00006.safetensors",
928
+ "model.layers.9.feed_forward.experts.7.down_proj.weight": "model-00003-of-00006.safetensors",
929
+ "model.layers.9.feed_forward.experts.7.gate_proj.weight": "model-00003-of-00006.safetensors",
930
+ "model.layers.9.feed_forward.experts.7.up_proj.weight": "model-00003-of-00006.safetensors",
931
+ "model.layers.9.feed_forward.gate.weight": "model-00002-of-00006.safetensors",
932
+ "model.layers.9.input_layernorm.weight": "model-00003-of-00006.safetensors",
933
+ "model.layers.9.mamba.A_log": "model-00003-of-00006.safetensors",
934
+ "model.layers.9.mamba.D": "model-00003-of-00006.safetensors",
935
+ "model.layers.9.mamba.conv1d.bias": "model-00003-of-00006.safetensors",
936
+ "model.layers.9.mamba.conv1d.weight": "model-00003-of-00006.safetensors",
937
+ "model.layers.9.mamba.dt_bias": "model-00003-of-00006.safetensors",
938
+ "model.layers.9.mamba.in_proj.weight": "model-00003-of-00006.safetensors",
939
+ "model.layers.9.mamba.norm.weight": "model-00003-of-00006.safetensors",
940
+ "model.layers.9.mamba.out_proj.weight": "model-00003-of-00006.safetensors",
941
+ "model.layers.9.pre_ff_layernorm.weight": "model-00003-of-00006.safetensors",
942
+ "model.layers.9.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
943
+ "model.layers.9.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
944
+ "model.layers.9.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
945
+ "model.layers.9.self_attn.v_proj.weight": "model-00003-of-00006.safetensors"
946
+ }
947
+ }
modeling_falcon_h1_moe.py ADDED
@@ -0,0 +1,260 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ # This file is adapted from NVIDIA Megatron-LM.
3
+ #
4
+ # Original source:
5
+ # https://github.com/NVIDIA/Megatron-LM/blob/991138eedc580ba89ce8e44e771783f083cd9dee/megatron/core/transformer/moe/moe_utils.py
6
+ #
7
+ # Copyright (c) NVIDIA Corporation.
8
+ # Licensed under the Apache License, Version 2.0.
9
+ #
10
+ # Modifications in this file:
11
+ # - Removed distributed and communication-related logic
12
+ # - Retained only single-process / local computation components
13
+ # - Minor refactoring for integration with this project
14
+
15
+
16
+
17
+ import torch
18
+ import torch.nn.functional as F
19
+ from transformers import FalconH1ForCausalLM, FalconH1Model
20
+ from openrlhf.moe_utils import FalconH1MoEConfig
21
+ from transformers.models.falcon_h1.modeling_falcon_h1 import FalconH1DecoderLayer, FalconH1MLP, compute_mup_vector
22
+ from torch import nn
23
+
24
+ class FalconH1MoEModel(FalconH1Model):
25
+ def __init__(self, config: FalconH1MoEConfig):
26
+ super().__init__(config)
27
+ decoder_layers = []
28
+ for i in range(config.num_hidden_layers):
29
+ decoder_layers.append(FalconH1MoEDecoderLayer(config, layer_idx=i))
30
+ self.layers = nn.ModuleList(decoder_layers)
31
+ mup_vector = compute_mup_vector(config)
32
+ for layer in self.layers:
33
+ layer.mamba.register_buffer("mup_vector", mup_vector, persistent=False)
34
+
35
+
36
+
37
+
38
+
39
+
40
+
41
+
42
+
43
+
44
+
45
+
46
+ class FalconH1MoEMLP(nn.Module):
47
+ def __init__(self, config):
48
+ super().__init__()
49
+ self.config = config
50
+ self.num_experts = config.expert_num
51
+ self.top_k = config.topk
52
+ self.norm_topk_prob = False
53
+
54
+ # gating
55
+ self.gate = nn.Linear(config.hidden_size, config.expert_num, bias=False)
56
+ self.experts = nn.ModuleList(
57
+ [FalconH1MLP(config) for _ in range(self.num_experts)]
58
+ )
59
+
60
+ # self.shared_expert = FalconH1MLP(config)
61
+ # self.shared_expert_gate = torch.nn.Linear(config.hidden_size, 1, bias=False)
62
+
63
+
64
+ def fuse_moe(self, batch_size, sequence_length, hidden_dim, selected_experts, hidden_states, routing_weights):
65
+ final_hidden_states = torch.zeros(
66
+ (batch_size * sequence_length, hidden_dim), dtype=hidden_states.dtype, device=hidden_states.device
67
+ )
68
+
69
+ # One hot encode the selected experts to create an expert mask
70
+ # this will be used to easily index which expert is going to be sollicitated
71
+ expert_mask = torch.nn.functional.one_hot(selected_experts, num_classes=self.num_experts).permute(2, 1, 0)
72
+
73
+ # Loop over all available experts in the model and perform the computation on each expert
74
+ expert_hitted = torch.greater(expert_mask.sum(dim=(-1, -2)), 0).nonzero()
75
+ for expert_idx in expert_hitted:
76
+ expert_layer = self.experts[expert_idx]
77
+ idx, top_x = torch.where(expert_mask[expert_idx].squeeze(0))
78
+
79
+ # Index the correct hidden states and compute the expert hidden state for
80
+ # the current expert. We need to make sure to multiply the output hidden
81
+ # states by `routing_weights` on the corresponding tokens (top-1 and top-2)
82
+ current_state = hidden_states[None, top_x].reshape(-1, hidden_dim)
83
+ current_hidden_states = expert_layer(current_state) * routing_weights[top_x, idx, None]
84
+
85
+ # However `index_add_` only support torch tensors for indexing so we'll use
86
+ # the `top_x` tensor here.
87
+ final_hidden_states.index_add_(0, top_x, current_hidden_states.to(hidden_states.dtype))
88
+ return final_hidden_states
89
+
90
+ def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
91
+ """ """
92
+ batch_size, sequence_length, hidden_dim = hidden_states.shape
93
+ hidden_states = hidden_states.view(-1, hidden_dim)
94
+ # router_logits: (batch * sequence_length, n_experts)
95
+ router_logits = self.gate(hidden_states)
96
+ routing_weights = F.softmax(router_logits, dim=1, dtype=torch.float)
97
+ routing_weights, selected_experts = torch.topk(routing_weights, self.top_k, dim=-1)
98
+
99
+ if self.norm_topk_prob:
100
+ routing_weights /= routing_weights.sum(dim=-1, keepdim=True)
101
+ # we cast back to the input dtype
102
+ routing_weights = routing_weights.to(hidden_states.dtype)
103
+
104
+
105
+ final_hidden_states = self.fuse_moe(batch_size, sequence_length, hidden_dim, selected_experts, hidden_states, routing_weights)
106
+
107
+ # shared_expert_output = self.shared_expert(hidden_states)
108
+ # shared_expert_output = F.sigmoid(self.shared_expert_gate(hidden_states)) * shared_expert_output
109
+
110
+ # final_hidden_states = final_hidden_states + shared_expert_output
111
+
112
+ final_hidden_states = final_hidden_states.reshape(batch_size, sequence_length, hidden_dim)
113
+ return final_hidden_states, router_logits
114
+
115
+
116
+
117
+
118
+
119
+
120
+
121
+
122
+
123
+
124
+
125
+
126
+
127
+
128
+
129
+ # class FalconH1MoEMLP(nn.Module):
130
+ # def __init__(self, config: FalconH1MoEConfig):
131
+ # super().__init__()
132
+ # self.config = config
133
+ # self.num_local_experts = config.expert_num
134
+ # self.topk=config.topk
135
+ # '''build experts'''
136
+ # self.experts = torch.nn.ModuleList()
137
+ # for _ in range(self.num_local_experts):
138
+ # expert = FalconH1MLP(config)
139
+ # self.experts.append(expert)
140
+
141
+
142
+ # '''build router'''
143
+ # self.weight = torch.nn.Parameter(
144
+ # torch.empty((self.num_local_experts, self.config.hidden_size), dtype=torch.float32)
145
+ # )
146
+ # torch.nn.init.xavier_uniform_(self.weight)
147
+
148
+
149
+
150
+ # def forward(self, x):
151
+ # x = x.transpose(0, 1).contiguous() #x: [seq_len, bs, hidden_size]
152
+ # '''fixed parameters'''
153
+ # inp_shape = x.shape
154
+ # num_tokens = inp_shape[0] * inp_shape[1]
155
+ # hidden = inp_shape[-1]
156
+ # num_experts = self.num_local_experts
157
+ # x = x.view(-1, inp_shape[-1]) #x: [token_num, hidden_size]
158
+ # restore_shape = x.shape
159
+
160
+
161
+
162
+ # """Routing , compute the experts' weight for each token, all following step is on token level.
163
+ # Args:
164
+ # Returns:
165
+ # routing_probs, token -> expert_prob
166
+ # [[0.0000, 0.0000, 0.4006, 0.5994],
167
+ # ...,
168
+ # [0.0373, 0.0000, 0.9627, 0.0000]]
169
+ # ------------
170
+ # routing_map, token -> expert_idx
171
+ # [[False, False, True, True],
172
+ # ...,
173
+ # [ True, False, True, False]])
174
+ # """
175
+ # y = torch.mm(x, self.weight.to(x.dtype).t()) #y: [token_num, expert_num]
176
+ # scores, top_indices = torch.topk(y, k=self.topk, dim=1)
177
+ # probs = torch.softmax(scores, dim=-1, dtype=torch.float32).type_as(y)
178
+ # routing_probs = torch.zeros_like(y).scatter(1, top_indices, probs)
179
+ # routing_map = torch.zeros_like(y).int().scatter(1, top_indices, 1).bool()
180
+
181
+
182
+
183
+ # """Dispatch: experts-to-tokens
184
+
185
+ # Args:
186
+ # Returns:
187
+ # probs: [expert0{token4_prob, token2_prob,token8_prob}.....expertn]
188
+ # x: [expert0{token4_idx, token2_idx, token8_idx}.....]
189
+ # """
190
+ # num_local_tokens_per_expert = routing_map.sum(dim=0).long() # [token_num_e_1, ...., token_num_e_n]
191
+ # routing_map = routing_map.bool().T.contiguous() # expert-to-token, [expert_num, token_num]
192
+ # '''
193
+ # [False, False, False, ..., False, True, True],
194
+ # [False, False, False, ..., True, False, False],
195
+ # [ True, True, True, ..., True, True, True],
196
+ # [ True, True, True, ..., False, False, False]]
197
+ # '''
198
+ # token_indices = (
199
+ # torch.arange(num_tokens, device=routing_map.device).unsqueeze(0).expand(num_experts, -1)
200
+ # ) # [expert_num, token_num]
201
+ # '''
202
+ # [[ 0, 1, 2, ..., 1021, 1022, 1023],
203
+ # [ 0, 1, 2, ..., 1021, 1022, 1023],
204
+ # [ 0, 1, 2, ..., 1021, 1022, 1023],
205
+ # [ 0, 1, 2, ..., 1021, 1022, 1023]]
206
+ # '''
207
+
208
+ # sorted_indices = token_indices.masked_select(routing_map) # [topk * token_num]
209
+ # '''
210
+ # [ 8, 9, 12, ..., 1015, 1016, 1017],
211
+ # sorted_indices[:idx_1]->expert0
212
+ # sorted_indices[idx_1:idx_2]->expert1
213
+ # sorted_indices[idx_2:idx_3]->expert2
214
+ # sorted_indices[idx_3:idx_4]->expert3
215
+ # '''
216
+ # probs = routing_probs.T.contiguous().masked_select(routing_map) # [topk * token_num]
217
+ # '''
218
+ # [0.6458, 0.6458, 0.5577, ..., 0.4983, 0.0520, 0.0520]
219
+ # '''
220
+ # x = x.index_select(0, sorted_indices) # [token_num * topk, hidden]
221
+
222
+
223
+
224
+ # tokens_list = torch.split(x, num_local_tokens_per_expert.tolist())
225
+ # probs_list = torch.split(probs, num_local_tokens_per_expert.tolist())
226
+
227
+ # output_local_list = []
228
+
229
+ # for expert, tokens, prob in zip(self.experts, tokens_list, probs_list):
230
+ # output = expert(tokens) * prob.unsqueeze(-1)
231
+
232
+ # output_local_list.append(output)
233
+ # permuted_tokens = torch.cat(output_local_list, dim=0)
234
+
235
+ # output_tokens = torch.zeros(
236
+ # restore_shape, dtype=permuted_tokens.dtype, device=permuted_tokens.device
237
+ # )
238
+ # # Scatter add the permuted_input back to the original positions
239
+ # output_tokens.scatter_add_(0, sorted_indices.unsqueeze(1).expand(-1, hidden), permuted_tokens)
240
+ # output = output_tokens.view(inp_shape).transpose(0, 1)
241
+
242
+ # return output
243
+
244
+
245
+
246
+ class FalconH1MoEDecoderLayer(FalconH1DecoderLayer):
247
+ def __init__(self, config: FalconH1MoEConfig, layer_idx: int):
248
+ super().__init__(config, layer_idx)
249
+ self.feed_forward = FalconH1MoEMLP(config)
250
+
251
+
252
+
253
+
254
+
255
+ class FalconH1MoEForCausalLM(FalconH1ForCausalLM):
256
+ def __init__(self, config: FalconH1MoEConfig):
257
+ super().__init__(config)
258
+ self.model = FalconH1MoEModel(config)
259
+
260
+ __all__ = ["FalconH1MoEForCausalLM"]