2139 lines
38 KiB
JSON
2139 lines
38 KiB
JSON
{
|
|
"model": "Kimi-K2.5-1T",
|
|
"quant": "Q4_0",
|
|
"total_shards": 13,
|
|
"shards_completed": 2,
|
|
"total_tensors": 194,
|
|
"tensors": [
|
|
{
|
|
"name": "output.weight",
|
|
"dims": [
|
|
7168,
|
|
163840
|
|
],
|
|
"dtype": "Q6_K",
|
|
"n_elements": 1174405120,
|
|
"theta": 89.97,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "output_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.94,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "token_embd.weight",
|
|
"dims": [
|
|
7168,
|
|
163840
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 1174405120,
|
|
"theta": 89.45,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.0.attn_k_b.weight",
|
|
"dims": [
|
|
128,
|
|
512,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 86.27,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.0.attn_kv_a_mqa.weight",
|
|
"dims": [
|
|
7168,
|
|
576
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4128768,
|
|
"theta": 81.41,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.0.attn_kv_a_norm.weight",
|
|
"dims": [
|
|
512
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 512,
|
|
"theta": 85.41,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.0.attn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 88.49,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.0.attn_output.weight",
|
|
"dims": [
|
|
8192,
|
|
7168
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 58720256,
|
|
"theta": 89.99,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.0.attn_q_a.weight",
|
|
"dims": [
|
|
7168,
|
|
1536
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 11010048,
|
|
"theta": 89.94,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.0.attn_q_a_norm.weight",
|
|
"dims": [
|
|
1536
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 1536,
|
|
"theta": 89.18,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.0.attn_q_b.weight",
|
|
"dims": [
|
|
1536,
|
|
12288
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 18874368,
|
|
"theta": 90.0,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.0.attn_v_b.weight",
|
|
"dims": [
|
|
512,
|
|
128,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 55.6,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.0.ffn_down.weight",
|
|
"dims": [
|
|
18432,
|
|
7168
|
|
],
|
|
"dtype": "Q4_1",
|
|
"n_elements": 132120576,
|
|
"theta": null,
|
|
"status": "unsupported_dtype_3"
|
|
},
|
|
{
|
|
"name": "blk.0.ffn_gate.weight",
|
|
"dims": [
|
|
7168,
|
|
18432
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 132120576,
|
|
"theta": 89.76,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.0.ffn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.35,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.0.ffn_up.weight",
|
|
"dims": [
|
|
7168,
|
|
18432
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 132120576,
|
|
"theta": 89.86,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.attn_k_b.weight",
|
|
"dims": [
|
|
128,
|
|
512,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 81.09,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.attn_kv_a_mqa.weight",
|
|
"dims": [
|
|
7168,
|
|
576
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4128768,
|
|
"theta": 89.5,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.attn_kv_a_norm.weight",
|
|
"dims": [
|
|
512
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 512,
|
|
"theta": 86.39,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.attn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.83,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.attn_output.weight",
|
|
"dims": [
|
|
8192,
|
|
7168
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 58720256,
|
|
"theta": 89.9,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.attn_q_a.weight",
|
|
"dims": [
|
|
7168,
|
|
1536
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 11010048,
|
|
"theta": 89.94,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.attn_q_a_norm.weight",
|
|
"dims": [
|
|
1536
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 1536,
|
|
"theta": 86.97,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.attn_q_b.weight",
|
|
"dims": [
|
|
1536,
|
|
12288
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 18874368,
|
|
"theta": 89.99,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.attn_v_b.weight",
|
|
"dims": [
|
|
512,
|
|
128,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 78.71,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.exp_probs_b.bias",
|
|
"dims": [
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 384,
|
|
"theta": 86.75,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.ffn_down_exps.weight",
|
|
"dims": [
|
|
2048,
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "Q4_1",
|
|
"n_elements": 5637144576,
|
|
"theta": null,
|
|
"status": "unsupported_dtype_3"
|
|
},
|
|
{
|
|
"name": "blk.1.ffn_down_shexp.weight",
|
|
"dims": [
|
|
2048,
|
|
7168
|
|
],
|
|
"dtype": "Q4_1",
|
|
"n_elements": 14680064,
|
|
"theta": null,
|
|
"status": "unsupported_dtype_3"
|
|
},
|
|
{
|
|
"name": "blk.1.ffn_gate_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.98,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.ffn_gate_inp.weight",
|
|
"dims": [
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 2752512,
|
|
"theta": 89.98,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.ffn_gate_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.22,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.ffn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.5,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.ffn_up_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.67,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.1.ffn_up_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.59,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.attn_k_b.weight",
|
|
"dims": [
|
|
128,
|
|
512,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 52.18,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.attn_kv_a_mqa.weight",
|
|
"dims": [
|
|
7168,
|
|
576
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4128768,
|
|
"theta": 82.45,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.attn_kv_a_norm.weight",
|
|
"dims": [
|
|
512
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 512,
|
|
"theta": 86.85,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.attn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.71,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.attn_output.weight",
|
|
"dims": [
|
|
8192,
|
|
7168
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 58720256,
|
|
"theta": 89.57,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.attn_q_a.weight",
|
|
"dims": [
|
|
7168,
|
|
1536
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 11010048,
|
|
"theta": 89.97,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.attn_q_a_norm.weight",
|
|
"dims": [
|
|
1536
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 1536,
|
|
"theta": 87.42,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.attn_q_b.weight",
|
|
"dims": [
|
|
1536,
|
|
12288
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 18874368,
|
|
"theta": 89.85,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.attn_v_b.weight",
|
|
"dims": [
|
|
512,
|
|
128,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 53.98,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.exp_probs_b.bias",
|
|
"dims": [
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 384,
|
|
"theta": 88.06,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.ffn_down_exps.weight",
|
|
"dims": [
|
|
2048,
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "Q4_1",
|
|
"n_elements": 5637144576,
|
|
"theta": null,
|
|
"status": "unsupported_dtype_3"
|
|
},
|
|
{
|
|
"name": "blk.2.ffn_down_shexp.weight",
|
|
"dims": [
|
|
2048,
|
|
7168
|
|
],
|
|
"dtype": "Q4_1",
|
|
"n_elements": 14680064,
|
|
"theta": null,
|
|
"status": "unsupported_dtype_3"
|
|
},
|
|
{
|
|
"name": "blk.2.ffn_gate_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.36,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.ffn_gate_inp.weight",
|
|
"dims": [
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 2752512,
|
|
"theta": 89.99,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.ffn_gate_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.65,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.ffn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.72,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.ffn_up_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.93,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.2.ffn_up_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.63,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.attn_k_b.weight",
|
|
"dims": [
|
|
128,
|
|
512,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 74.84,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.attn_kv_a_mqa.weight",
|
|
"dims": [
|
|
7168,
|
|
576
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4128768,
|
|
"theta": 89.06,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.attn_kv_a_norm.weight",
|
|
"dims": [
|
|
512
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 512,
|
|
"theta": 88.61,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.attn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.73,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.attn_output.weight",
|
|
"dims": [
|
|
8192,
|
|
7168
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 58720256,
|
|
"theta": 89.95,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.attn_q_a.weight",
|
|
"dims": [
|
|
7168,
|
|
1536
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 11010048,
|
|
"theta": 89.93,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.attn_q_a_norm.weight",
|
|
"dims": [
|
|
1536
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 1536,
|
|
"theta": 89.5,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.attn_q_b.weight",
|
|
"dims": [
|
|
1536,
|
|
12288
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 18874368,
|
|
"theta": 89.96,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.attn_v_b.weight",
|
|
"dims": [
|
|
512,
|
|
128,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 69.72,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.exp_probs_b.bias",
|
|
"dims": [
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 384,
|
|
"theta": 88.82,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.ffn_down_exps.weight",
|
|
"dims": [
|
|
2048,
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "Q4_1",
|
|
"n_elements": 5637144576,
|
|
"theta": null,
|
|
"status": "unsupported_dtype_3"
|
|
},
|
|
{
|
|
"name": "blk.3.ffn_down_shexp.weight",
|
|
"dims": [
|
|
2048,
|
|
7168
|
|
],
|
|
"dtype": "Q4_1",
|
|
"n_elements": 14680064,
|
|
"theta": null,
|
|
"status": "unsupported_dtype_3"
|
|
},
|
|
{
|
|
"name": "blk.3.ffn_gate_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.88,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.ffn_gate_inp.weight",
|
|
"dims": [
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 2752512,
|
|
"theta": 89.95,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.ffn_gate_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.39,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.ffn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.8,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.ffn_up_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.99,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.3.ffn_up_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.7,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.attn_k_b.weight",
|
|
"dims": [
|
|
128,
|
|
512,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 61.83,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.attn_kv_a_mqa.weight",
|
|
"dims": [
|
|
7168,
|
|
576
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4128768,
|
|
"theta": 89.75,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.attn_kv_a_norm.weight",
|
|
"dims": [
|
|
512
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 512,
|
|
"theta": 87.27,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.attn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.49,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.attn_output.weight",
|
|
"dims": [
|
|
8192,
|
|
7168
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 58720256,
|
|
"theta": 89.99,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.attn_q_a.weight",
|
|
"dims": [
|
|
7168,
|
|
1536
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 11010048,
|
|
"theta": 89.89,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.attn_q_a_norm.weight",
|
|
"dims": [
|
|
1536
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 1536,
|
|
"theta": 87.83,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.attn_q_b.weight",
|
|
"dims": [
|
|
1536,
|
|
12288
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 18874368,
|
|
"theta": 89.83,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.attn_v_b.weight",
|
|
"dims": [
|
|
512,
|
|
128,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 68.28,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.exp_probs_b.bias",
|
|
"dims": [
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 384,
|
|
"theta": 87.44,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.ffn_down_exps.weight",
|
|
"dims": [
|
|
2048,
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "Q4_1",
|
|
"n_elements": 5637144576,
|
|
"theta": null,
|
|
"status": "unsupported_dtype_3"
|
|
},
|
|
{
|
|
"name": "blk.4.ffn_down_shexp.weight",
|
|
"dims": [
|
|
2048,
|
|
7168
|
|
],
|
|
"dtype": "Q4_1",
|
|
"n_elements": 14680064,
|
|
"theta": null,
|
|
"status": "unsupported_dtype_3"
|
|
},
|
|
{
|
|
"name": "blk.4.ffn_gate_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.75,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.ffn_gate_inp.weight",
|
|
"dims": [
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 2752512,
|
|
"theta": 89.98,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.ffn_gate_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 88.39,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.ffn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.75,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.ffn_up_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.53,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.4.ffn_up_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.25,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.attn_k_b.weight",
|
|
"dims": [
|
|
128,
|
|
512,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 49.88,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.attn_kv_a_mqa.weight",
|
|
"dims": [
|
|
7168,
|
|
576
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4128768,
|
|
"theta": 89.98,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.attn_kv_a_norm.weight",
|
|
"dims": [
|
|
512
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 512,
|
|
"theta": 89.04,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.attn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 88.57,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.attn_output.weight",
|
|
"dims": [
|
|
8192,
|
|
7168
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 58720256,
|
|
"theta": 89.79,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.attn_q_a.weight",
|
|
"dims": [
|
|
7168,
|
|
1536
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 11010048,
|
|
"theta": 89.67,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.attn_q_a_norm.weight",
|
|
"dims": [
|
|
1536
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 1536,
|
|
"theta": 89.53,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.attn_q_b.weight",
|
|
"dims": [
|
|
1536,
|
|
12288
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 18874368,
|
|
"theta": 89.87,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.attn_v_b.weight",
|
|
"dims": [
|
|
512,
|
|
128,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 89.4,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.exp_probs_b.bias",
|
|
"dims": [
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 384,
|
|
"theta": 88.88,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.ffn_down_exps.weight",
|
|
"dims": [
|
|
2048,
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "Q4_1",
|
|
"n_elements": 5637144576,
|
|
"theta": null,
|
|
"status": "unsupported_dtype_3"
|
|
},
|
|
{
|
|
"name": "blk.5.ffn_down_shexp.weight",
|
|
"dims": [
|
|
2048,
|
|
7168
|
|
],
|
|
"dtype": "Q4_1",
|
|
"n_elements": 14680064,
|
|
"theta": null,
|
|
"status": "unsupported_dtype_3"
|
|
},
|
|
{
|
|
"name": "blk.5.ffn_gate_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.66,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.ffn_gate_inp.weight",
|
|
"dims": [
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 2752512,
|
|
"theta": 89.98,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.ffn_gate_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 87.99,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.ffn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.94,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.ffn_up_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.56,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.5.ffn_up_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 88.27,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.attn_k_b.weight",
|
|
"dims": [
|
|
128,
|
|
512,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 45.21,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.attn_kv_a_mqa.weight",
|
|
"dims": [
|
|
7168,
|
|
576
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4128768,
|
|
"theta": 86.53,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.attn_kv_a_norm.weight",
|
|
"dims": [
|
|
512
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 512,
|
|
"theta": 89.21,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.attn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.28,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.attn_output.weight",
|
|
"dims": [
|
|
8192,
|
|
7168
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 58720256,
|
|
"theta": 89.83,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.attn_q_a.weight",
|
|
"dims": [
|
|
7168,
|
|
1536
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 11010048,
|
|
"theta": 89.99,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.attn_q_a_norm.weight",
|
|
"dims": [
|
|
1536
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 1536,
|
|
"theta": 89.19,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.attn_q_b.weight",
|
|
"dims": [
|
|
1536,
|
|
12288
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 18874368,
|
|
"theta": 89.59,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.attn_v_b.weight",
|
|
"dims": [
|
|
512,
|
|
128,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 59.69,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.exp_probs_b.bias",
|
|
"dims": [
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 384,
|
|
"theta": 88.2,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.ffn_down_exps.weight",
|
|
"dims": [
|
|
2048,
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "Q4_1",
|
|
"n_elements": 5637144576,
|
|
"theta": null,
|
|
"status": "unsupported_dtype_3"
|
|
},
|
|
{
|
|
"name": "blk.6.ffn_down_shexp.weight",
|
|
"dims": [
|
|
2048,
|
|
7168
|
|
],
|
|
"dtype": "Q4_1",
|
|
"n_elements": 14680064,
|
|
"theta": null,
|
|
"status": "unsupported_dtype_3"
|
|
},
|
|
{
|
|
"name": "blk.6.ffn_gate_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.89,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.ffn_gate_inp.weight",
|
|
"dims": [
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 2752512,
|
|
"theta": 89.98,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.ffn_gate_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.51,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.ffn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.64,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.ffn_up_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.83,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.6.ffn_up_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.99,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.attn_k_b.weight",
|
|
"dims": [
|
|
128,
|
|
512,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 40.66,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.attn_kv_a_mqa.weight",
|
|
"dims": [
|
|
7168,
|
|
576
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4128768,
|
|
"theta": 88.75,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.attn_kv_a_norm.weight",
|
|
"dims": [
|
|
512
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 512,
|
|
"theta": 87.74,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.attn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.95,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.attn_output.weight",
|
|
"dims": [
|
|
8192,
|
|
7168
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 58720256,
|
|
"theta": 89.72,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.attn_q_a.weight",
|
|
"dims": [
|
|
7168,
|
|
1536
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 11010048,
|
|
"theta": 89.88,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.attn_q_a_norm.weight",
|
|
"dims": [
|
|
1536
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 1536,
|
|
"theta": 88.69,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.attn_q_b.weight",
|
|
"dims": [
|
|
1536,
|
|
12288
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 18874368,
|
|
"theta": 89.98,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.attn_v_b.weight",
|
|
"dims": [
|
|
512,
|
|
128,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 64.82,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.exp_probs_b.bias",
|
|
"dims": [
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 384,
|
|
"theta": 86.12,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.ffn_down_exps.weight",
|
|
"dims": [
|
|
2048,
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.85,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.ffn_down_shexp.weight",
|
|
"dims": [
|
|
2048,
|
|
7168
|
|
],
|
|
"dtype": "Q6_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.73,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.ffn_gate_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.98,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.ffn_gate_inp.weight",
|
|
"dims": [
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 2752512,
|
|
"theta": 89.99,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.ffn_gate_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.75,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.ffn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.53,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.ffn_up_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.67,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.7.ffn_up_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.83,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.attn_k_b.weight",
|
|
"dims": [
|
|
128,
|
|
512,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 77.55,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.attn_kv_a_mqa.weight",
|
|
"dims": [
|
|
7168,
|
|
576
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4128768,
|
|
"theta": 87.5,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.attn_kv_a_norm.weight",
|
|
"dims": [
|
|
512
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 512,
|
|
"theta": 84.23,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.attn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.76,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.attn_output.weight",
|
|
"dims": [
|
|
8192,
|
|
7168
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 58720256,
|
|
"theta": 89.81,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.attn_q_a.weight",
|
|
"dims": [
|
|
7168,
|
|
1536
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 11010048,
|
|
"theta": 89.84,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.attn_q_a_norm.weight",
|
|
"dims": [
|
|
1536
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 1536,
|
|
"theta": 88.72,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.attn_q_b.weight",
|
|
"dims": [
|
|
1536,
|
|
12288
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 18874368,
|
|
"theta": 89.76,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.attn_v_b.weight",
|
|
"dims": [
|
|
512,
|
|
128,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 86.0,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.exp_probs_b.bias",
|
|
"dims": [
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 384,
|
|
"theta": 88.83,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.ffn_down_exps.weight",
|
|
"dims": [
|
|
2048,
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.64,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.ffn_down_shexp.weight",
|
|
"dims": [
|
|
2048,
|
|
7168
|
|
],
|
|
"dtype": "Q6_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.77,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.ffn_gate_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.64,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.ffn_gate_inp.weight",
|
|
"dims": [
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 2752512,
|
|
"theta": 89.98,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.ffn_gate_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.88,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.ffn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.89,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.ffn_up_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.99,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.8.ffn_up_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.55,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.attn_k_b.weight",
|
|
"dims": [
|
|
128,
|
|
512,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 65.89,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.attn_kv_a_mqa.weight",
|
|
"dims": [
|
|
7168,
|
|
576
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4128768,
|
|
"theta": 89.61,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.attn_kv_a_norm.weight",
|
|
"dims": [
|
|
512
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 512,
|
|
"theta": 88.51,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.attn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.41,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.attn_output.weight",
|
|
"dims": [
|
|
8192,
|
|
7168
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 58720256,
|
|
"theta": 89.95,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.attn_q_a.weight",
|
|
"dims": [
|
|
7168,
|
|
1536
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 11010048,
|
|
"theta": 89.76,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.attn_q_a_norm.weight",
|
|
"dims": [
|
|
1536
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 1536,
|
|
"theta": 89.7,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.attn_q_b.weight",
|
|
"dims": [
|
|
1536,
|
|
12288
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 18874368,
|
|
"theta": 89.42,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.attn_v_b.weight",
|
|
"dims": [
|
|
512,
|
|
128,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 89.4,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.exp_probs_b.bias",
|
|
"dims": [
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 384,
|
|
"theta": 88.37,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.ffn_down_exps.weight",
|
|
"dims": [
|
|
2048,
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.95,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.ffn_down_shexp.weight",
|
|
"dims": [
|
|
2048,
|
|
7168
|
|
],
|
|
"dtype": "Q6_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.86,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.ffn_gate_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.53,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.ffn_gate_inp.weight",
|
|
"dims": [
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 2752512,
|
|
"theta": 89.99,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.ffn_gate_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.19,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.ffn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.81,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.ffn_up_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.6,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.9.ffn_up_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.5,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.attn_k_b.weight",
|
|
"dims": [
|
|
128,
|
|
512,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 59.9,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.attn_kv_a_mqa.weight",
|
|
"dims": [
|
|
7168,
|
|
576
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4128768,
|
|
"theta": 89.54,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.attn_kv_a_norm.weight",
|
|
"dims": [
|
|
512
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 512,
|
|
"theta": 86.35,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.attn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.44,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.attn_output.weight",
|
|
"dims": [
|
|
8192,
|
|
7168
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 58720256,
|
|
"theta": 89.8,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.attn_q_a.weight",
|
|
"dims": [
|
|
7168,
|
|
1536
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 11010048,
|
|
"theta": 89.62,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.attn_q_a_norm.weight",
|
|
"dims": [
|
|
1536
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 1536,
|
|
"theta": 87.33,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.attn_q_b.weight",
|
|
"dims": [
|
|
1536,
|
|
12288
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 18874368,
|
|
"theta": 89.85,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.attn_v_b.weight",
|
|
"dims": [
|
|
512,
|
|
128,
|
|
64
|
|
],
|
|
"dtype": "Q8_0",
|
|
"n_elements": 4194304,
|
|
"theta": 65.84,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.exp_probs_b.bias",
|
|
"dims": [
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 384,
|
|
"theta": 89.42,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.ffn_down_exps.weight",
|
|
"dims": [
|
|
2048,
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.96,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.ffn_down_shexp.weight",
|
|
"dims": [
|
|
2048,
|
|
7168
|
|
],
|
|
"dtype": "Q6_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.45,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.ffn_gate_exps.weight",
|
|
"dims": [
|
|
7168,
|
|
2048,
|
|
384
|
|
],
|
|
"dtype": "Q4_0",
|
|
"n_elements": 5637144576,
|
|
"theta": 89.87,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.ffn_gate_inp.weight",
|
|
"dims": [
|
|
7168,
|
|
384
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 2752512,
|
|
"theta": 89.96,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.ffn_gate_shexp.weight",
|
|
"dims": [
|
|
7168,
|
|
2048
|
|
],
|
|
"dtype": "Q5_K",
|
|
"n_elements": 14680064,
|
|
"theta": 89.89,
|
|
"status": "ok"
|
|
},
|
|
{
|
|
"name": "blk.10.ffn_norm.weight",
|
|
"dims": [
|
|
7168
|
|
],
|
|
"dtype": "F32",
|
|
"n_elements": 7168,
|
|
"theta": 89.55,
|
|
"status": "ok"
|
|
}
|
|
]
|
|
} |