Skip to content

gguf-dump.py: add --markdown dump output - #7853

Merged
mofosyne merged 11 commits into
ggml-org:masterfrom
mofosyne:gguf-dump-markdown
Jun 17, 2024
Merged

gguf-dump.py: add --markdown dump output#7853
mofosyne merged 11 commits into
ggml-org:masterfrom
mofosyne:gguf-dump-markdown

Conversation

@mofosyne

@mofosyne mofosyne commented Jun 10, 2024 •
edited
Loading

Copy link
Copy Markdown
Collaborator

This will allow $gguf-dump.py Tinyllama-5M-v0.2-Q8_0.gguf --markdown to output a markdown formatted dump that is designed to be as easy as possible to read as a markdown file.

It's sent to stdout so you can do stuff like $gguf-dump.py Tinyllama-5M-v0.2-Q8_0.gguf --markdown | mdless to render the markdown dump directly. Alternatively it might be part of your workflow to render this file when creating a new gguf.

Why do this when you can still manually dump it whenever? Well in a github / huggingface repo, it still might be good courtesy to have an easy to read technical dump of the layers. Note that I also added a function that reads the tensor name and converts it into a human friendly name as devs coming across it may not necessarily know what ffn etc.. means.

Below is how it currently renders. Feel free to suggest changes to it to make it as useful as possible for you.

Example Output 
# phi-2.Q6_K.gguf - GGUF Internal File Dump
* Endian: LITTLE endian

## Key Value Metadata Store

There is 23 key/value pair(s) in this file

| POS | TYPE     | Count | Key                               | Value                                                                             |
|----:|:---------|------:|:----------------------------------|:----------------------------------------------------------------------------------|
|   1 | UINT32   |     1 | GGUF.version                      | 3                                                                                 |
|   2 | UINT64   |     1 | GGUF.tensor_count                 | 325                                                                               |
|   3 | UINT64   |     1 | GGUF.kv_count                     | 20                                                                                |
|   4 | STRING   |     1 | general.architecture              | 'phi2'                                                                            |
|   5 | STRING   |     1 | general.name                      | 'Phi2'                                                                            |
|   6 | UINT32   |     1 | phi2.context_length               | 2048                                                                              |
|   7 | UINT32   |     1 | phi2.embedding_length             | 2560                                                                              |
|   8 | UINT32   |     1 | phi2.feed_forward_length          | 10240                                                                             |
|   9 | UINT32   |     1 | phi2.block_count                  | 32                                                                                |
|  10 | UINT32   |     1 | phi2.attention.head_count         | 32                                                                                |
|  11 | UINT32   |     1 | phi2.attention.head_count_kv      | 32                                                                                |
|  12 | FLOAT32  |     1 | phi2.attention.layer_norm_epsilon | 1e-05                                                                             |
|  13 | UINT32   |     1 | phi2.rope.dimension_count         | 32                                                                                |
|  14 | UINT32   |     1 | general.file_type                 | 18                                                                                |
|  15 | BOOL     |     1 | tokenizer.ggml.add_bos_token      | False                                                                             |
|  16 | STRING   |     1 | tokenizer.ggml.model              | 'gpt2'                                                                            |
|  17 | [STRING] | 51200 | tokenizer.ggml.tokens             | [ '[PAD5', '\n\x00\x00\x00\x00', '[PAD5', '\n\x00\x00\x00\x00', '[PAD5',  ... ]   |
|  18 | [INT32]  | 51200 | tokenizer.ggml.token_type         | [ 4, 4, 4, 4, 4, 4, 4,  ... ]                                                     |
|  19 | [STRING] | 50000 | tokenizer.ggml.merges             | [ 'Ġg az', '\x08\x00\x00\x00\x00', 'Ġinfo', '\r\x00\x00\x00\x00', 'ĠColl',  ... ] |
|  20 | UINT32   |     1 | tokenizer.ggml.bos_token_id       | 50256                                                                             |
|  21 | UINT32   |     1 | tokenizer.ggml.eos_token_id       | 50256                                                                             |
|  22 | UINT32   |     1 | tokenizer.ggml.unknown_token_id   | 50256                                                                             |
|  23 | UINT32   |     1 | general.quantization_version      | 2                                                                                 |

## Tensors Overview ~3B Elements

Total number of elements in all tensors: 2779683840 Elements

- [Base Tensor Group - ~262M Elements](#base)
- [Block 0 Tensor Group - ~79M Elements](#blk_0)
- [Block 1 Tensor Group - ~79M Elements](#blk_1)
- [Block 10 Tensor Group - ~79M Elements](#blk_10)
- [Block 11 Tensor Group - ~79M Elements](#blk_11)
- [Block 12 Tensor Group - ~79M Elements](#blk_12)
- [Block 13 Tensor Group - ~79M Elements](#blk_13)
- [Block 14 Tensor Group - ~79M Elements](#blk_14)
- [Block 15 Tensor Group - ~79M Elements](#blk_15)
- [Block 16 Tensor Group - ~79M Elements](#blk_16)
- [Block 17 Tensor Group - ~79M Elements](#blk_17)
- [Block 18 Tensor Group - ~79M Elements](#blk_18)
- [Block 19 Tensor Group - ~79M Elements](#blk_19)
- [Block 2 Tensor Group - ~79M Elements](#blk_2)
- [Block 20 Tensor Group - ~79M Elements](#blk_20)
- [Block 21 Tensor Group - ~79M Elements](#blk_21)
- [Block 22 Tensor Group - ~79M Elements](#blk_22)
- [Block 23 Tensor Group - ~79M Elements](#blk_23)
- [Block 24 Tensor Group - ~79M Elements](#blk_24)
- [Block 25 Tensor Group - ~79M Elements](#blk_25)
- [Block 26 Tensor Group - ~79M Elements](#blk_26)
- [Block 27 Tensor Group - ~79M Elements](#blk_27)
- [Block 28 Tensor Group - ~79M Elements](#blk_28)
- [Block 29 Tensor Group - ~79M Elements](#blk_29)
- [Block 3 Tensor Group - ~79M Elements](#blk_3)
- [Block 30 Tensor Group - ~79M Elements](#blk_30)
- [Block 4 Tensor Group - ~79M Elements](#blk_4)
- [Block 5 Tensor Group - ~79M Elements](#blk_5)
- [Block 6 Tensor Group - ~79M Elements](#blk_6)
- [Block 7 Tensor Group - ~79M Elements](#blk_7)
- [Block 8 Tensor Group - ~79M Elements](#blk_8)
- [Block 9 Tensor Group - ~79M Elements](#blk_9)
- [Block 31 Tensor Group - ~79M Elements](#blk_31)

### <a name="base">Base Tensor Group : ~262M Elements</a>

| T_ID | Tensor Layer Name  | Human Friendly Tensor Layer Name | Elements          | Shape                 | Type |
|-----:|:-------------------|:---------------------------------|:------------------|:----------------------|:-----|
|    0 | token_embd.weight  | Token Embedding (W)              | (~131M) 131072000 |  2560 x 51200 x 1 x 1 | Q6_K |
|  303 | output.bias        | Output (B)                       | ( ~51K)     51200 | 51200 x     1 x 1 x 1 | F32  |
|  304 | output.weight      | Output (W)                       | (~131M) 131072000 |  2560 x 51200 x 1 x 1 | Q6_K |
|  305 | output_norm.bias   | Output Normalization (B)         | (  ~3K)      2560 |  2560 x     1 x 1 x 1 | F32  |
|  306 | output_norm.weight | Output Normalization (W)         | (  ~3K)      2560 |  2560 x     1 x 1 x 1 | F32  |

- Total elements in base: (~262M) 262200320
- Percentage of total elements: 9.43%


### <a name="blk_0">Block 0 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name        | Human Friendly Tensor Layer Name        | Elements        | Shape                 | Type |
|-----:|:-------------------------|:----------------------------------------|:----------------|:----------------------|:-----|
|    1 | blk.0.attn_norm.bias     | Block 0 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|    2 | blk.0.attn_norm.weight   | Block 0 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|    3 | blk.0.attn_qkv.bias      | Block 0 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|    4 | blk.0.attn_qkv.weight    | Block 0 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|    5 | blk.0.attn_output.bias   | Block 0 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|    6 | blk.0.attn_output.weight | Block 0 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|    7 | blk.0.ffn_up.bias        | Block 0 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|    8 | blk.0.ffn_up.weight      | Block 0 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|    9 | blk.0.ffn_down.bias      | Block 0 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   10 | blk.0.ffn_down.weight    | Block 0 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.0: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_1">Block 1 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name        | Human Friendly Tensor Layer Name        | Elements        | Shape                 | Type |
|-----:|:-------------------------|:----------------------------------------|:----------------|:----------------------|:-----|
|   11 | blk.1.attn_norm.bias     | Block 1 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   12 | blk.1.attn_norm.weight   | Block 1 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   13 | blk.1.attn_qkv.bias      | Block 1 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|   14 | blk.1.attn_qkv.weight    | Block 1 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|   15 | blk.1.attn_output.bias   | Block 1 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   16 | blk.1.attn_output.weight | Block 1 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|   17 | blk.1.ffn_up.bias        | Block 1 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|   18 | blk.1.ffn_up.weight      | Block 1 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|   19 | blk.1.ffn_down.bias      | Block 1 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   20 | blk.1.ffn_down.weight    | Block 1 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.1: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_10">Block 10 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|   21 | blk.10.attn_norm.bias     | Block 10 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   22 | blk.10.attn_norm.weight   | Block 10 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   23 | blk.10.attn_qkv.bias      | Block 10 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|   24 | blk.10.attn_qkv.weight    | Block 10 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|   25 | blk.10.attn_output.bias   | Block 10 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   26 | blk.10.attn_output.weight | Block 10 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|   27 | blk.10.ffn_up.bias        | Block 10 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|   28 | blk.10.ffn_up.weight      | Block 10 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|   29 | blk.10.ffn_down.bias      | Block 10 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   30 | blk.10.ffn_down.weight    | Block 10 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.10: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_11">Block 11 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|   31 | blk.11.attn_norm.bias     | Block 11 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   32 | blk.11.attn_norm.weight   | Block 11 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   33 | blk.11.attn_qkv.bias      | Block 11 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|   34 | blk.11.attn_qkv.weight    | Block 11 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|   35 | blk.11.attn_output.bias   | Block 11 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   36 | blk.11.attn_output.weight | Block 11 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|   37 | blk.11.ffn_up.bias        | Block 11 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|   38 | blk.11.ffn_up.weight      | Block 11 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|   39 | blk.11.ffn_down.bias      | Block 11 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   40 | blk.11.ffn_down.weight    | Block 11 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.11: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_12">Block 12 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|   41 | blk.12.attn_norm.bias     | Block 12 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   42 | blk.12.attn_norm.weight   | Block 12 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   43 | blk.12.attn_qkv.bias      | Block 12 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|   44 | blk.12.attn_qkv.weight    | Block 12 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|   45 | blk.12.attn_output.bias   | Block 12 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   46 | blk.12.attn_output.weight | Block 12 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|   47 | blk.12.ffn_up.bias        | Block 12 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|   48 | blk.12.ffn_up.weight      | Block 12 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|   49 | blk.12.ffn_down.bias      | Block 12 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   50 | blk.12.ffn_down.weight    | Block 12 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.12: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_13">Block 13 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|   51 | blk.13.attn_norm.bias     | Block 13 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   52 | blk.13.attn_norm.weight   | Block 13 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   53 | blk.13.attn_qkv.bias      | Block 13 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|   54 | blk.13.attn_qkv.weight    | Block 13 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|   55 | blk.13.attn_output.bias   | Block 13 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   56 | blk.13.attn_output.weight | Block 13 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|   57 | blk.13.ffn_up.bias        | Block 13 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|   58 | blk.13.ffn_up.weight      | Block 13 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|   59 | blk.13.ffn_down.bias      | Block 13 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   60 | blk.13.ffn_down.weight    | Block 13 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.13: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_14">Block 14 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|   61 | blk.14.attn_norm.bias     | Block 14 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   62 | blk.14.attn_norm.weight   | Block 14 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   63 | blk.14.attn_qkv.bias      | Block 14 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|   64 | blk.14.attn_qkv.weight    | Block 14 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|   65 | blk.14.attn_output.bias   | Block 14 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   66 | blk.14.attn_output.weight | Block 14 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|   67 | blk.14.ffn_up.bias        | Block 14 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|   68 | blk.14.ffn_up.weight      | Block 14 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|   69 | blk.14.ffn_down.bias      | Block 14 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   70 | blk.14.ffn_down.weight    | Block 14 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.14: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_15">Block 15 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|   71 | blk.15.attn_norm.bias     | Block 15 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   72 | blk.15.attn_norm.weight   | Block 15 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   73 | blk.15.attn_qkv.bias      | Block 15 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|   74 | blk.15.attn_qkv.weight    | Block 15 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|   75 | blk.15.attn_output.bias   | Block 15 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   76 | blk.15.attn_output.weight | Block 15 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|   77 | blk.15.ffn_up.bias        | Block 15 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|   78 | blk.15.ffn_up.weight      | Block 15 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|   79 | blk.15.ffn_down.bias      | Block 15 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   80 | blk.15.ffn_down.weight    | Block 15 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.15: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_16">Block 16 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|   81 | blk.16.attn_norm.bias     | Block 16 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   82 | blk.16.attn_norm.weight   | Block 16 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   83 | blk.16.attn_qkv.bias      | Block 16 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|   84 | blk.16.attn_qkv.weight    | Block 16 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|   85 | blk.16.attn_output.bias   | Block 16 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   86 | blk.16.attn_output.weight | Block 16 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|   87 | blk.16.ffn_up.bias        | Block 16 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|   88 | blk.16.ffn_up.weight      | Block 16 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|   89 | blk.16.ffn_down.bias      | Block 16 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   90 | blk.16.ffn_down.weight    | Block 16 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.16: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_17">Block 17 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|   91 | blk.17.attn_norm.bias     | Block 17 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   92 | blk.17.attn_norm.weight   | Block 17 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   93 | blk.17.attn_qkv.bias      | Block 17 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|   94 | blk.17.attn_qkv.weight    | Block 17 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|   95 | blk.17.attn_output.bias   | Block 17 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|   96 | blk.17.attn_output.weight | Block 17 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|   97 | blk.17.ffn_up.bias        | Block 17 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|   98 | blk.17.ffn_up.weight      | Block 17 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|   99 | blk.17.ffn_down.bias      | Block 17 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  100 | blk.17.ffn_down.weight    | Block 17 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.17: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_18">Block 18 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|  101 | blk.18.attn_norm.bias     | Block 18 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  102 | blk.18.attn_norm.weight   | Block 18 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  103 | blk.18.attn_qkv.bias      | Block 18 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  104 | blk.18.attn_qkv.weight    | Block 18 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  105 | blk.18.attn_output.bias   | Block 18 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  106 | blk.18.attn_output.weight | Block 18 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  107 | blk.18.ffn_up.bias        | Block 18 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  108 | blk.18.ffn_up.weight      | Block 18 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  109 | blk.18.ffn_down.bias      | Block 18 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  110 | blk.18.ffn_down.weight    | Block 18 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.18: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_19">Block 19 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|  111 | blk.19.attn_norm.bias     | Block 19 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  112 | blk.19.attn_norm.weight   | Block 19 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  113 | blk.19.attn_qkv.bias      | Block 19 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  114 | blk.19.attn_qkv.weight    | Block 19 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  115 | blk.19.attn_output.bias   | Block 19 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  116 | blk.19.attn_output.weight | Block 19 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  117 | blk.19.ffn_up.bias        | Block 19 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  118 | blk.19.ffn_up.weight      | Block 19 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  119 | blk.19.ffn_down.bias      | Block 19 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  120 | blk.19.ffn_down.weight    | Block 19 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.19: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_2">Block 2 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name        | Human Friendly Tensor Layer Name        | Elements        | Shape                 | Type |
|-----:|:-------------------------|:----------------------------------------|:----------------|:----------------------|:-----|
|  121 | blk.2.attn_norm.bias     | Block 2 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  122 | blk.2.attn_norm.weight   | Block 2 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  123 | blk.2.attn_qkv.bias      | Block 2 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  124 | blk.2.attn_qkv.weight    | Block 2 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  125 | blk.2.attn_output.bias   | Block 2 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  126 | blk.2.attn_output.weight | Block 2 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  127 | blk.2.ffn_up.bias        | Block 2 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  128 | blk.2.ffn_up.weight      | Block 2 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  129 | blk.2.ffn_down.bias      | Block 2 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  130 | blk.2.ffn_down.weight    | Block 2 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.2: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_20">Block 20 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|  131 | blk.20.attn_norm.bias     | Block 20 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  132 | blk.20.attn_norm.weight   | Block 20 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  133 | blk.20.attn_qkv.bias      | Block 20 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  134 | blk.20.attn_qkv.weight    | Block 20 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  135 | blk.20.attn_output.bias   | Block 20 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  136 | blk.20.attn_output.weight | Block 20 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  137 | blk.20.ffn_up.bias        | Block 20 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  138 | blk.20.ffn_up.weight      | Block 20 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  139 | blk.20.ffn_down.bias      | Block 20 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  140 | blk.20.ffn_down.weight    | Block 20 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.20: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_21">Block 21 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|  141 | blk.21.attn_norm.bias     | Block 21 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  142 | blk.21.attn_norm.weight   | Block 21 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  143 | blk.21.attn_qkv.bias      | Block 21 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  144 | blk.21.attn_qkv.weight    | Block 21 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  145 | blk.21.attn_output.bias   | Block 21 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  146 | blk.21.attn_output.weight | Block 21 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  147 | blk.21.ffn_up.bias        | Block 21 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  148 | blk.21.ffn_up.weight      | Block 21 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  149 | blk.21.ffn_down.bias      | Block 21 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  150 | blk.21.ffn_down.weight    | Block 21 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.21: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_22">Block 22 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|  151 | blk.22.attn_norm.bias     | Block 22 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  152 | blk.22.attn_norm.weight   | Block 22 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  153 | blk.22.attn_qkv.bias      | Block 22 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  154 | blk.22.attn_qkv.weight    | Block 22 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  155 | blk.22.attn_output.bias   | Block 22 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  156 | blk.22.attn_output.weight | Block 22 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  157 | blk.22.ffn_up.bias        | Block 22 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  158 | blk.22.ffn_up.weight      | Block 22 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  159 | blk.22.ffn_down.bias      | Block 22 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  160 | blk.22.ffn_down.weight    | Block 22 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.22: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_23">Block 23 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|  161 | blk.23.attn_norm.bias     | Block 23 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  162 | blk.23.attn_norm.weight   | Block 23 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  163 | blk.23.attn_qkv.bias      | Block 23 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  164 | blk.23.attn_qkv.weight    | Block 23 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  165 | blk.23.attn_output.bias   | Block 23 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  166 | blk.23.attn_output.weight | Block 23 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  167 | blk.23.ffn_up.bias        | Block 23 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  168 | blk.23.ffn_up.weight      | Block 23 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  169 | blk.23.ffn_down.bias      | Block 23 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  170 | blk.23.ffn_down.weight    | Block 23 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.23: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_24">Block 24 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|  171 | blk.24.attn_norm.bias     | Block 24 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  172 | blk.24.attn_norm.weight   | Block 24 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  173 | blk.24.attn_qkv.bias      | Block 24 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  174 | blk.24.attn_qkv.weight    | Block 24 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  175 | blk.24.attn_output.bias   | Block 24 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  176 | blk.24.attn_output.weight | Block 24 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  177 | blk.24.ffn_up.bias        | Block 24 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  178 | blk.24.ffn_up.weight      | Block 24 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  179 | blk.24.ffn_down.bias      | Block 24 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  180 | blk.24.ffn_down.weight    | Block 24 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.24: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_25">Block 25 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|  181 | blk.25.attn_norm.bias     | Block 25 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  182 | blk.25.attn_norm.weight   | Block 25 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  183 | blk.25.attn_qkv.bias      | Block 25 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  184 | blk.25.attn_qkv.weight    | Block 25 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  185 | blk.25.attn_output.bias   | Block 25 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  186 | blk.25.attn_output.weight | Block 25 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  187 | blk.25.ffn_up.bias        | Block 25 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  188 | blk.25.ffn_up.weight      | Block 25 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  189 | blk.25.ffn_down.bias      | Block 25 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  190 | blk.25.ffn_down.weight    | Block 25 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.25: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_26">Block 26 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|  191 | blk.26.attn_norm.bias     | Block 26 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  192 | blk.26.attn_norm.weight   | Block 26 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  193 | blk.26.attn_qkv.bias      | Block 26 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  194 | blk.26.attn_qkv.weight    | Block 26 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  195 | blk.26.attn_output.bias   | Block 26 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  196 | blk.26.attn_output.weight | Block 26 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  197 | blk.26.ffn_up.bias        | Block 26 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  198 | blk.26.ffn_up.weight      | Block 26 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  199 | blk.26.ffn_down.bias      | Block 26 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  200 | blk.26.ffn_down.weight    | Block 26 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.26: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_27">Block 27 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|  201 | blk.27.attn_norm.bias     | Block 27 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  202 | blk.27.attn_norm.weight   | Block 27 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  203 | blk.27.attn_qkv.bias      | Block 27 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  204 | blk.27.attn_qkv.weight    | Block 27 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  205 | blk.27.attn_output.bias   | Block 27 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  206 | blk.27.attn_output.weight | Block 27 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  207 | blk.27.ffn_up.bias        | Block 27 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  208 | blk.27.ffn_up.weight      | Block 27 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  209 | blk.27.ffn_down.bias      | Block 27 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  210 | blk.27.ffn_down.weight    | Block 27 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.27: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_28">Block 28 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|  211 | blk.28.attn_norm.bias     | Block 28 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  212 | blk.28.attn_norm.weight   | Block 28 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  213 | blk.28.attn_qkv.bias      | Block 28 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  214 | blk.28.attn_qkv.weight    | Block 28 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  215 | blk.28.attn_output.bias   | Block 28 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  216 | blk.28.attn_output.weight | Block 28 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  217 | blk.28.ffn_up.bias        | Block 28 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  218 | blk.28.ffn_up.weight      | Block 28 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  219 | blk.28.ffn_down.bias      | Block 28 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  220 | blk.28.ffn_down.weight    | Block 28 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.28: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_29">Block 29 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|  221 | blk.29.attn_norm.bias     | Block 29 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  222 | blk.29.attn_norm.weight   | Block 29 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  223 | blk.29.attn_qkv.bias      | Block 29 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  224 | blk.29.attn_qkv.weight    | Block 29 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  225 | blk.29.attn_output.bias   | Block 29 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  226 | blk.29.attn_output.weight | Block 29 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  227 | blk.29.ffn_up.bias        | Block 29 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  228 | blk.29.ffn_up.weight      | Block 29 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  229 | blk.29.ffn_down.bias      | Block 29 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  230 | blk.29.ffn_down.weight    | Block 29 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.29: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_3">Block 3 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name        | Human Friendly Tensor Layer Name        | Elements        | Shape                 | Type |
|-----:|:-------------------------|:----------------------------------------|:----------------|:----------------------|:-----|
|  231 | blk.3.attn_norm.bias     | Block 3 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  232 | blk.3.attn_norm.weight   | Block 3 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  233 | blk.3.attn_qkv.bias      | Block 3 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  234 | blk.3.attn_qkv.weight    | Block 3 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  235 | blk.3.attn_output.bias   | Block 3 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  236 | blk.3.attn_output.weight | Block 3 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  237 | blk.3.ffn_up.bias        | Block 3 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  238 | blk.3.ffn_up.weight      | Block 3 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  239 | blk.3.ffn_down.bias      | Block 3 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  240 | blk.3.ffn_down.weight    | Block 3 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.3: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_30">Block 30 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|  241 | blk.30.attn_norm.bias     | Block 30 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  242 | blk.30.attn_norm.weight   | Block 30 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  307 | blk.30.attn_qkv.bias      | Block 30 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  308 | blk.30.attn_qkv.weight    | Block 30 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  309 | blk.30.attn_output.bias   | Block 30 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  310 | blk.30.attn_output.weight | Block 30 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  311 | blk.30.ffn_up.bias        | Block 30 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  312 | blk.30.ffn_up.weight      | Block 30 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  313 | blk.30.ffn_down.bias      | Block 30 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  314 | blk.30.ffn_down.weight    | Block 30 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.30: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_4">Block 4 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name        | Human Friendly Tensor Layer Name        | Elements        | Shape                 | Type |
|-----:|:-------------------------|:----------------------------------------|:----------------|:----------------------|:-----|
|  243 | blk.4.attn_norm.bias     | Block 4 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  244 | blk.4.attn_norm.weight   | Block 4 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  245 | blk.4.attn_qkv.bias      | Block 4 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  246 | blk.4.attn_qkv.weight    | Block 4 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  247 | blk.4.attn_output.bias   | Block 4 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  248 | blk.4.attn_output.weight | Block 4 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  249 | blk.4.ffn_up.bias        | Block 4 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  250 | blk.4.ffn_up.weight      | Block 4 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  251 | blk.4.ffn_down.bias      | Block 4 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  252 | blk.4.ffn_down.weight    | Block 4 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.4: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_5">Block 5 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name        | Human Friendly Tensor Layer Name        | Elements        | Shape                 | Type |
|-----:|:-------------------------|:----------------------------------------|:----------------|:----------------------|:-----|
|  253 | blk.5.attn_norm.bias     | Block 5 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  254 | blk.5.attn_norm.weight   | Block 5 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  255 | blk.5.attn_qkv.bias      | Block 5 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  256 | blk.5.attn_qkv.weight    | Block 5 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  257 | blk.5.attn_output.bias   | Block 5 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  258 | blk.5.attn_output.weight | Block 5 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  259 | blk.5.ffn_up.bias        | Block 5 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  260 | blk.5.ffn_up.weight      | Block 5 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  261 | blk.5.ffn_down.bias      | Block 5 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  262 | blk.5.ffn_down.weight    | Block 5 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.5: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_6">Block 6 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name        | Human Friendly Tensor Layer Name        | Elements        | Shape                 | Type |
|-----:|:-------------------------|:----------------------------------------|:----------------|:----------------------|:-----|
|  263 | blk.6.attn_norm.bias     | Block 6 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  264 | blk.6.attn_norm.weight   | Block 6 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  265 | blk.6.attn_qkv.bias      | Block 6 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  266 | blk.6.attn_qkv.weight    | Block 6 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  267 | blk.6.attn_output.bias   | Block 6 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  268 | blk.6.attn_output.weight | Block 6 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  269 | blk.6.ffn_up.bias        | Block 6 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  270 | blk.6.ffn_up.weight      | Block 6 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  271 | blk.6.ffn_down.bias      | Block 6 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  272 | blk.6.ffn_down.weight    | Block 6 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.6: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_7">Block 7 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name        | Human Friendly Tensor Layer Name        | Elements        | Shape                 | Type |
|-----:|:-------------------------|:----------------------------------------|:----------------|:----------------------|:-----|
|  273 | blk.7.attn_norm.bias     | Block 7 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  274 | blk.7.attn_norm.weight   | Block 7 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  275 | blk.7.attn_qkv.bias      | Block 7 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  276 | blk.7.attn_qkv.weight    | Block 7 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  277 | blk.7.attn_output.bias   | Block 7 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  278 | blk.7.attn_output.weight | Block 7 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  279 | blk.7.ffn_up.bias        | Block 7 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  280 | blk.7.ffn_up.weight      | Block 7 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  281 | blk.7.ffn_down.bias      | Block 7 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  282 | blk.7.ffn_down.weight    | Block 7 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.7: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_8">Block 8 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name        | Human Friendly Tensor Layer Name        | Elements        | Shape                 | Type |
|-----:|:-------------------------|:----------------------------------------|:----------------|:----------------------|:-----|
|  283 | blk.8.attn_norm.bias     | Block 8 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  284 | blk.8.attn_norm.weight   | Block 8 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  285 | blk.8.attn_qkv.bias      | Block 8 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  286 | blk.8.attn_qkv.weight    | Block 8 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  287 | blk.8.attn_output.bias   | Block 8 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  288 | blk.8.attn_output.weight | Block 8 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  289 | blk.8.ffn_up.bias        | Block 8 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  290 | blk.8.ffn_up.weight      | Block 8 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  291 | blk.8.ffn_down.bias      | Block 8 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  292 | blk.8.ffn_down.weight    | Block 8 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.8: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_9">Block 9 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name        | Human Friendly Tensor Layer Name        | Elements        | Shape                 | Type |
|-----:|:-------------------------|:----------------------------------------|:----------------|:----------------------|:-----|
|  293 | blk.9.attn_norm.bias     | Block 9 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  294 | blk.9.attn_norm.weight   | Block 9 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  295 | blk.9.attn_qkv.bias      | Block 9 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  296 | blk.9.attn_qkv.weight    | Block 9 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  297 | blk.9.attn_output.bias   | Block 9 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  298 | blk.9.attn_output.weight | Block 9 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  299 | blk.9.ffn_up.bias        | Block 9 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  300 | blk.9.ffn_up.weight      | Block 9 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  301 | blk.9.ffn_down.bias      | Block 9 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  302 | blk.9.ffn_down.weight    | Block 9 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.9: (~79M) 78671360
- Percentage of total elements: 2.83%


### <a name="blk_31">Block 31 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name         | Human Friendly Tensor Layer Name         | Elements        | Shape                 | Type |
|-----:|:--------------------------|:-----------------------------------------|:----------------|:----------------------|:-----|
|  315 | blk.31.attn_norm.bias     | Block 31 Attention Normalization (B)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  316 | blk.31.attn_norm.weight   | Block 31 Attention Normalization (W)     | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  317 | blk.31.attn_qkv.bias      | Block 31 Attention Query-Key-Value (B)   | ( ~8K)     7680 |  7680 x     1 x 1 x 1 | F32  |
|  318 | blk.31.attn_qkv.weight    | Block 31 Attention Query-Key-Value (W)   | (~20M) 19660800 |  2560 x  7680 x 1 x 1 | Q6_K |
|  319 | blk.31.attn_output.bias   | Block 31 Attention Output (B)            | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  320 | blk.31.attn_output.weight | Block 31 Attention Output (W)            | ( ~7M)  6553600 |  2560 x  2560 x 1 x 1 | Q6_K |
|  321 | blk.31.ffn_up.bias        | Block 31 Feed-Forward Network "Up" (B)   | (~10K)    10240 | 10240 x     1 x 1 x 1 | F32  |
|  322 | blk.31.ffn_up.weight      | Block 31 Feed-Forward Network "Up" (W)   | (~26M) 26214400 |  2560 x 10240 x 1 x 1 | Q6_K |
|  323 | blk.31.ffn_down.bias      | Block 31 Feed-Forward Network "Down" (B) | ( ~3K)     2560 |  2560 x     1 x 1 x 1 | F32  |
|  324 | blk.31.ffn_down.weight    | Block 31 Feed-Forward Network "Down" (W) | (~26M) 26214400 | 10240 x  2560 x 1 x 1 | Q6_K |

- Total elements in blk.31: (~79M) 78671360
- Percentage of total elements: 2.83%

@mofosyne mofosyne added Review Complexity : Low Trivial changes to code that most beginner devs (or those who want a break) can tackle. e.g. UI fix python python script changes labels Jun 10, 2024
@mofosyne

mofosyne commented Jun 10, 2024 •
edited
Loading

Copy link
Copy Markdown
Collaborator Author

FYI the context of this, was I was trying to understand the GGUF file standard a bit more. I also attempted to get the dump to print as a graphviz dotfile... but it was quite a pain to get working. However markdown is quite easy so this seems like a good compromise.

The main benefit I can see however is in automatically grouping all the blocks together.

@mofosyne
mofosyne force-pushed the gguf-dump-markdown branch from aff4168 to f15ce9f Compare June 10, 2024 10:47
@mofosyne
mofosyne requested a review from KerfuffleV2 June 10, 2024 13:19
@mofosyne
mofosyne force-pushed the gguf-dump-markdown branch from c0439a4 to e38b649 Compare June 10, 2024 16:04
@mofosyne
mofosyne requested a review from compilade June 10, 2024 23:37
compilade
compilade reviewed Jun 13, 2024
View reviewed changes
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
@mofosyne

mofosyne commented Jun 13, 2024

Copy link
Copy Markdown
Collaborator Author

@compilade updated the script with your suggestion also updated the example output with example output of a 'phi-2.Q6_K.gguf - GGUF Internal File Dump' dump so you can see how it looks now.

compilade
compilade reviewed Jun 13, 2024
View reviewed changes
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
@mofosyne

mofosyne commented Jun 13, 2024

Copy link
Copy Markdown
Collaborator Author

@compilade looks good thanks

compilade
compilade approved these changes Jun 14, 2024
View reviewed changes

@compilade compilade left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There are still things about sub-column alignment which could be improved, although at least now the column bars are always aligned across rows.

Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
@mofosyne

mofosyne commented Jun 14, 2024 •
edited
Loading

Copy link
Copy Markdown
Collaborator Author

Adjusted

### <a name="blk_1">Block 1 Tensor Group : ~79M Elements</a>

| T_ID | Tensor Layer Name        | Human Friendly Tensor Layer Name        | Elements         | Shape                 | Type |
|-----:|:-------------------------|:----------------------------------------|:-----------------|:----------------------|:-----|
|   11 | blk.1.attn_norm.bias     | Block 1 Attention Normalization (B)     | (  ~3K)     2560 | 2560  x   1   x 1 x 1 | F32  |
|   12 | blk.1.attn_norm.weight   | Block 1 Attention Normalization (W)     | (  ~3K)     2560 | 2560  x   1   x 1 x 1 | F32  |
|   13 | blk.1.attn_qkv.bias      | Block 1 Attention Query-Key-Value (B)   | (  ~8K)     7680 | 7680  x   1   x 1 x 1 | F32  |
|   14 | blk.1.attn_qkv.weight    | Block 1 Attention Query-Key-Value (W)   | ( ~20M) 19660800 | 2560  x 7680  x 1 x 1 | Q6_K |
|   15 | blk.1.attn_output.bias   | Block 1 Attention Output (B)            | (  ~3K)     2560 | 2560  x   1   x 1 x 1 | F32  |
|   16 | blk.1.attn_output.weight | Block 1 Attention Output (W)            | (  ~7M)  6553600 | 2560  x 2560  x 1 x 1 | Q6_K |
|   17 | blk.1.ffn_up.bias        | Block 1 Feed-Forward Network "Up" (B)   | ( ~10K)    10240 | 10240 x   1   x 1 x 1 | F32  |
|   18 | blk.1.ffn_up.weight      | Block 1 Feed-Forward Network "Up" (W)   | ( ~26M) 26214400 | 2560  x 10240 x 1 x 1 | Q6_K |
|   19 | blk.1.ffn_down.bias      | Block 1 Feed-Forward Network "Down" (B) | (  ~3K)     2560 | 2560  x   1   x 1 x 1 | F32  |
|   20 | blk.1.ffn_down.weight    | Block 1 Feed-Forward Network "Down" (W) | ( ~26M) 26214400 | 10240 x 2560  x 1 x 1 | Q6_K |

- Total elements in blk.1: (~79M) 78671360
- Percentage of total elements: 2.83%

If you have no other thoughts @compilade feel free to press merge.

compilade
compilade reviewed Jun 15, 2024
View reviewed changes

@compilade compilade left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Still think a generalized approach to sub-column alignment might be appropriate. At least now it's only the Elements column which has problematic alignment.

Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
@mofosyne

mofosyne commented Jun 15, 2024 •
edited
Loading

Copy link
Copy Markdown
Collaborator Author

@compilade included your recommendation. How is it now?

@mofosyne
mofosyne force-pushed the gguf-dump-markdown branch from 7dc405b to 51032b1 Compare June 15, 2024 13:41
@mofosyne

mofosyne commented Jun 15, 2024

Copy link
Copy Markdown
Collaborator Author

rebased with no modification to sync against latest master to deal with ci issue

@mofosyne

mofosyne commented Jun 16, 2024

Copy link
Copy Markdown
Collaborator Author

Will merge soon as I've addressed all of @compilade outstanding points and ci has passed.

@mofosyne mofosyne added the merge ready A maintainer can use this label to indicate that they consider the changes final and ready to merge. label Jun 16, 2024
compilade
compilade reviewed Jun 17, 2024
View reviewed changes

@compilade compilade left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Table formatting seems good in more cases. Some minor things to fix regarding consistency of the printed text, then this will be good to merge.

Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
Comment thread gguf-py/scripts/gguf-dump.py Resolved
@mofosyne @compilade
Apply suggestions from code review
6200b43 
Co-authored-by: compilade <git@compilade.net>
@mofosyne
mofosyne merged commit 006167a into ggml-org:master Jun 17, 2024
@mofosyne
mofosyne deleted the gguf-dump-markdown branch June 17, 2024 05:25
@compilade compilade mentioned this pull request Jun 24, 2024
Merged
2 of 4 tasks
Seunghhon pushed a commit to Seunghhon/llama.cpp that referenced this pull request Apr 26, 2026
@mofosyne @compilade
gguf-dump.py: add --markdown dump output (ggml-org#7853)
374f3ea 
* gguf-dump.py: add --markdown dump output

* gguf-dump.py: Add toc

* gguf-dump.py: use standard tensor name lookup. Also add tensor ID field

* gguf-dump.py: Add tensor overview count

* gguf-dump.py: fix array preview

* gguf-dump.py: markdownTableWithAlignmentSupport() added

* Add type hints and spacing

Co-authored-by: compilade <git@compilade.net>

* gguf-dump.py: prettyfy dimention

* gguf-dump: right align element count

* gguf-dump.py: element count autosizing

* Apply suggestions from code review

Co-authored-by: compilade <git@compilade.net>

---------

Co-authored-by: compilade <git@compilade.net>
phuongncn pushed a commit to phuongncn/llama.cpp-gx10-dgx-sparks-deepseekv4 that referenced this pull request Apr 28, 2026
@mofosyne @compilade
gguf-dump.py: add --markdown dump output (ggml-org#7853)
69a2e73 
* gguf-dump.py: add --markdown dump output

* gguf-dump.py: Add toc

* gguf-dump.py: use standard tensor name lookup. Also add tensor ID field

* gguf-dump.py: Add tensor overview count

* gguf-dump.py: fix array preview

* gguf-dump.py: markdownTableWithAlignmentSupport() added

* Add type hints and spacing

Co-authored-by: compilade <git@compilade.net>

* gguf-dump.py: prettyfy dimention

* gguf-dump: right align element count

* gguf-dump.py: element count autosizing

* Apply suggestions from code review

Co-authored-by: compilade <git@compilade.net>

---------

Co-authored-by: compilade <git@compilade.net>
ljubomirj pushed a commit to ljubomirj/llama.cpp that referenced this pull request May 6, 2026
@mofosyne @compilade
gguf-dump.py: add --markdown dump output (ggml-org#7853)
8070aaf 
* gguf-dump.py: add --markdown dump output

* gguf-dump.py: Add toc

* gguf-dump.py: use standard tensor name lookup. Also add tensor ID field

* gguf-dump.py: Add tensor overview count

* gguf-dump.py: fix array preview

* gguf-dump.py: markdownTableWithAlignmentSupport() added

* Add type hints and spacing

Co-authored-by: compilade <git@compilade.net>

* gguf-dump.py: prettyfy dimention

* gguf-dump: right align element count

* gguf-dump.py: element count autosizing

* Apply suggestions from code review

Co-authored-by: compilade <git@compilade.net>

---------

Co-authored-by: compilade <git@compilade.net>
my-other-github-account pushed a commit to my-other-github-account/llama.cpp that referenced this pull request May 15, 2026
@mofosyne @compilade
gguf-dump.py: add --markdown dump output (ggml-org#7853)
295a59c 
* gguf-dump.py: add --markdown dump output

* gguf-dump.py: Add toc

* gguf-dump.py: use standard tensor name lookup. Also add tensor ID field

* gguf-dump.py: Add tensor overview count

* gguf-dump.py: fix array preview

* gguf-dump.py: markdownTableWithAlignmentSupport() added

* Add type hints and spacing

Co-authored-by: compilade <git@compilade.net>

* gguf-dump.py: prettyfy dimention

* gguf-dump: right align element count

* gguf-dump.py: element count autosizing

* Apply suggestions from code review

Co-authored-by: compilade <git@compilade.net>

---------

Co-authored-by: compilade <git@compilade.net>
AlexiAlp pushed a commit to minghaop/llama.cpp that referenced this pull request Jun 2, 2026
@mofosyne @compilade
gguf-dump.py: add --markdown dump output (ggml-org#7853)
8ec7515 
* gguf-dump.py: add --markdown dump output

* gguf-dump.py: Add toc

* gguf-dump.py: use standard tensor name lookup. Also add tensor ID field

* gguf-dump.py: Add tensor overview count

* gguf-dump.py: fix array preview

* gguf-dump.py: markdownTableWithAlignmentSupport() added

* Add type hints and spacing

Co-authored-by: compilade <git@compilade.net>

* gguf-dump.py: prettyfy dimention

* gguf-dump: right align element count

* gguf-dump.py: element count autosizing

* Apply suggestions from code review

Co-authored-by: compilade <git@compilade.net>

---------

Co-authored-by: compilade <git@compilade.net>
AlexiAlp pushed a commit to minghaop/llama.cpp that referenced this pull request Jun 2, 2026
@mofosyne @compilade
gguf-dump.py: add --markdown dump output (ggml-org#7853)
4c4b342 
* gguf-dump.py: add --markdown dump output

* gguf-dump.py: Add toc

* gguf-dump.py: use standard tensor name lookup. Also add tensor ID field

* gguf-dump.py: Add tensor overview count

* gguf-dump.py: fix array preview

* gguf-dump.py: markdownTableWithAlignmentSupport() added

* Add type hints and spacing

Co-authored-by: compilade <git@compilade.net>

* gguf-dump.py: prettyfy dimention

* gguf-dump: right align element count

* gguf-dump.py: element count autosizing

* Apply suggestions from code review

Co-authored-by: compilade <git@compilade.net>

---------

Co-authored-by: compilade <git@compilade.net>
fukuro-kun pushed a commit to fukuro-kun/fukuro-llama-cpp-turboquant that referenced this pull request Jul 5, 2026
@mofosyne @compilade
gguf-dump.py: add --markdown dump output (ggml-org#7853)
6ed4007 
* gguf-dump.py: add --markdown dump output

* gguf-dump.py: Add toc

* gguf-dump.py: use standard tensor name lookup. Also add tensor ID field

* gguf-dump.py: Add tensor overview count

* gguf-dump.py: fix array preview

* gguf-dump.py: markdownTableWithAlignmentSupport() added

* Add type hints and spacing

Co-authored-by: compilade <git@compilade.net>

* gguf-dump.py: prettyfy dimention

* gguf-dump: right align element count

* gguf-dump.py: element count autosizing

* Apply suggestions from code review

Co-authored-by: compilade <git@compilade.net>

---------

Co-authored-by: compilade <git@compilade.net>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@KerfuffleV2 KerfuffleV2 Awaiting requested review from KerfuffleV2
1 more reviewer
@compilade compilade compilade approved these changes
Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

merge ready A maintainer can use this label to indicate that they consider the changes final and ready to merge. python python script changes Review Complexity : Low Trivial changes to code that most beginner devs (or those who want a break) can tackle. e.g. UI fix

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@mofosyne @compilade