Hugging Face's logo

EasyDeL
/
Kimi-VL-A3B-Instruct

Text Generation
JAX
easydel
kimi_vl
CausalLM
ragged_page_attention_v3
conversational
custom_code
Model card Files
xet
Community
Kimi-VL-A3B-Instruct / tensorstore_index.json
erfanzar's picture
erfanzar
Upload KimiVLForConditionalGeneration
65399b9 verified
raw
history blame contribute delete
131 kB
 {
"format": "tensorstore",
"version": "0.0.85",
"prefixes": {
"model": [
{
"path": "model/language_model/lm_head/kernel",
"shape": [
2048,
163840
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/embed_tokens/embedding",
"shape": [
163840,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/0/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/0/mlp/down_proj/kernel",
"shape": [
11264,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/0/mlp/gate_proj/kernel",
"shape": [
2048,
11264
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/0/mlp/up_proj/kernel",
"shape": [
2048,
11264
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/0/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/0/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/0/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/0/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/0/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/0/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/1/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/2/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/3/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/4/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/5/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/6/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/7/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/8/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/9/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/10/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/11/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/12/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/13/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/14/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/15/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/16/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/17/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/18/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/19/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/20/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/21/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/22/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/23/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/24/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/25/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/input_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/mlp/experts/down_proj/kernel",
"shape": [
64,
1408,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/mlp/experts/gate_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/mlp/experts/up_proj/kernel",
"shape": [
64,
2048,
1408
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/mlp/gate/e_score_correction_bias",
"shape": [
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/mlp/gate/kernel",
"shape": [
2048,
64
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/mlp/shared_experts/down_proj/kernel",
"shape": [
2816,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/mlp/shared_experts/gate_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/mlp/shared_experts/up_proj/kernel",
"shape": [
2048,
2816
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/post_attention_layernorm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/self_attn/kv_a_layernorm/kernel",
"shape": [
512
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/self_attn/kv_a_proj_with_mqa/kernel",
"shape": [
2048,
576
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/self_attn/kv_b_proj/kernel",
"shape": [
512,
4096
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/self_attn/o_proj/kernel",
"shape": [
2048,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/layers/26/self_attn/q_proj/kernel",
"shape": [
2048,
3072
],
"dtype": "bfloat16"
},
{
"path": "model/language_model/model/norm/kernel",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/multi_modal_projector/linear_1/bias",
"shape": [
4608
],
"dtype": "bfloat16"
},
{
"path": "model/multi_modal_projector/linear_1/kernel",
"shape": [
4608,
4608
],
"dtype": "bfloat16"
},
{
"path": "model/multi_modal_projector/linear_2/bias",
"shape": [
2048
],
"dtype": "bfloat16"
},
{
"path": "model/multi_modal_projector/linear_2/kernel",
"shape": [
4608,
2048
],
"dtype": "bfloat16"
},
{
"path": "model/multi_modal_projector/pre_norm/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/multi_modal_projector/pre_norm/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/0/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/0/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/0/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/0/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/0/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/0/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/0/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/0/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/0/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/0/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/0/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/0/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/1/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/1/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/1/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/1/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/1/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/1/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/1/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/1/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/1/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/1/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/1/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/1/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/2/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/2/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/2/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/2/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/2/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/2/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/2/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/2/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/2/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/2/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/2/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/2/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/3/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/3/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/3/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/3/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/3/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/3/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/3/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/3/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/3/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/3/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/3/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/3/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/4/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/4/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/4/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/4/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/4/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/4/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/4/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/4/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/4/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/4/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/4/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/4/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/5/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/5/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/5/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/5/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/5/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/5/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/5/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/5/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/5/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/5/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/5/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/5/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/6/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/6/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/6/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/6/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/6/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/6/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/6/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/6/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/6/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/6/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/6/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/6/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/7/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/7/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/7/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/7/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/7/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/7/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/7/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/7/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/7/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/7/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/7/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/7/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/8/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/8/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/8/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/8/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/8/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/8/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/8/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/8/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/8/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/8/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/8/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/8/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/9/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/9/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/9/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/9/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/9/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/9/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/9/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/9/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/9/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/9/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/9/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/9/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/10/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/10/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/10/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/10/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/10/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/10/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/10/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/10/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/10/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/10/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/10/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/10/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/11/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/11/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/11/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/11/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/11/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/11/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/11/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/11/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/11/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/11/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/11/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/11/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/12/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/12/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/12/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/12/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/12/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/12/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/12/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/12/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/12/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/12/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/12/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/12/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/13/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/13/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/13/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/13/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/13/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/13/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/13/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/13/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/13/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/13/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/13/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/13/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/14/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/14/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/14/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/14/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/14/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/14/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/14/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/14/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/14/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/14/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/14/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/14/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/15/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/15/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/15/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/15/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/15/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/15/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/15/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/15/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/15/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/15/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/15/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/15/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/16/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/16/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/16/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/16/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/16/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/16/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/16/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/16/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/16/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/16/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/16/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/16/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/17/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/17/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/17/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/17/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/17/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/17/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/17/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/17/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/17/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/17/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/17/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/17/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/18/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/18/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/18/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/18/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/18/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/18/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/18/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/18/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/18/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/18/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/18/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/18/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/19/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/19/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/19/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/19/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/19/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/19/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/19/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/19/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/19/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/19/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/19/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/19/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/20/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/20/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/20/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/20/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/20/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/20/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/20/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/20/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/20/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/20/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/20/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/20/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/21/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/21/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/21/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/21/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/21/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/21/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/21/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/21/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/21/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/21/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/21/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/21/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/22/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/22/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/22/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/22/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/22/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/22/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/22/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/22/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/22/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/22/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/22/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/22/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/23/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/23/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/23/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/23/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/23/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/23/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/23/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/23/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/23/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/23/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/23/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/23/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/24/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/24/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/24/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/24/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/24/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/24/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/24/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/24/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/24/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/24/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/24/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/24/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/25/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/25/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/25/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/25/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/25/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/25/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/25/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/25/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/25/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/25/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/25/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/25/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/26/mlp/fc0/bias",
"shape": [
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/26/mlp/fc0/kernel",
"shape": [
1152,
4304
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/26/mlp/fc1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/26/mlp/fc1/kernel",
"shape": [
4304,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/26/norm0/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/26/norm0/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/26/norm1/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/26/norm1/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/26/wo/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/26/wo/kernel",
"shape": [
1152,
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/26/wqkv/bias",
"shape": [
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/blocks/26/wqkv/kernel",
"shape": [
1152,
3456
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/final_layernorm/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/encoder/final_layernorm/scale",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/patch_embed/pos_emb/kernel",
"shape": [
1152,
64,
64
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/patch_embed/proj/bias",
"shape": [
1152
],
"dtype": "bfloat16"
},
{
"path": "model/vision_tower/patch_embed/proj/kernel",
"shape": [
14,
14,
3,
1152
],
"dtype": "bfloat16"
}
]
}
}