Update README.md
Browse files
README.md
CHANGED
|
@@ -54,7 +54,33 @@ Rest of the remaining layers were quantized to *q3_k_l*
|
|
| 54 |
|
| 55 |
---
|
| 56 |
# Model Architect
|
| 57 |
-
<pre><code>```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 58 |
|
| 59 |
---
|
| 60 |
# Performance & Limitations
|
|
|
|
| 54 |
|
| 55 |
---
|
| 56 |
# Model Architect
|
| 57 |
+
<pre><code>```Qwen2ForCausalLM(
|
| 58 |
+
(model): Qwen2Model(
|
| 59 |
+
(embed_tokens): Embedding(151936, 896, padding_idx=151665)
|
| 60 |
+
(layers): ModuleList(
|
| 61 |
+
(0-23): 24 x Qwen2DecoderLayer(
|
| 62 |
+
(self_attn): Qwen2Attention(
|
| 63 |
+
(q_proj): Linear(in_features=896, out_features=896, bias=True)
|
| 64 |
+
(k_proj): Linear(in_features=896, out_features=128, bias=True)
|
| 65 |
+
(v_proj): Linear(in_features=896, out_features=128, bias=True)
|
| 66 |
+
(o_proj): Linear(in_features=896, out_features=896, bias=False)
|
| 67 |
+
(rotary_emb): LlamaRotaryEmbedding()
|
| 68 |
+
)
|
| 69 |
+
(mlp): Qwen2MLP(
|
| 70 |
+
(gate_proj): Linear(in_features=896, out_features=4864, bias=False)
|
| 71 |
+
(up_proj): Linear(in_features=896, out_features=4864, bias=False)
|
| 72 |
+
(down_proj): Linear(in_features=4864, out_features=896, bias=False)
|
| 73 |
+
(act_fn): SiLU()
|
| 74 |
+
)
|
| 75 |
+
(input_layernorm): Qwen2RMSNorm((896,), eps=1e-06)
|
| 76 |
+
(post_attention_layernorm): Qwen2RMSNorm((896,), eps=1e-06)
|
| 77 |
+
)
|
| 78 |
+
)
|
| 79 |
+
(norm): Qwen2RMSNorm((896,), eps=1e-06)
|
| 80 |
+
(rotary_emb): LlamaRotaryEmbedding()
|
| 81 |
+
)
|
| 82 |
+
(lm_head): Linear(in_features=896, out_features=151936, bias=False)
|
| 83 |
+
) ```</code></pre>
|
| 84 |
|
| 85 |
---
|
| 86 |
# Performance & Limitations
|