Qwen3-32B

Reconstructed from its own config.json with no weights read. 5.2M downloads on Hugging Face.

Our count against the checkpoint

The left number comes from the graph. The right one is the number of scalars in the published weight files. Nothing on this page was tuned to make them agree.

Derived from structure

32.76B

32,762,258,816 parameters

In the published checkpoint

32.76B

32,762,123,264 scalars · safetensors.total, read 2025-07-26

Delta

+0.00%

What it costs to run

Cost is a roofline estimate on the priced GPU for 10 epochs at batch 32 over 50,000 samples (assumed; no dataset attached). GPU fit is fp32 weights plus gradients plus two Adam moments (16 bytes per parameter) with 1.3x headroom; activations are not included and grow with batch size.

Layers

388

Will it forward-pass

Yes

Priced on

A10G (24GB)

Est. one run

$81956.66

Card	Memory
T4 (16GB)	weights + activations	does not fit
A100 (40GB)	weights + activations	does not fit
H100 (80GB)	weights + activations	does not fit

Structure

390 nodes. Output shapes are propagated from the input shape, batch dimension excluded.

	Layer	Type	Output shape
1	Input	Input	1 × 40960
2	Embedding	Embedding	1 × 40960 × 5120
3	RoPE	RoPE	1 × 40960 × 5120
4	RMSNorm_1_1	RMSNorm	1 × 40960 × 5120
5	Attention_1	Grouped Query Attn	1 × 40960 × 5120
6	Add_1_attn	Add	1 × 40960 × 5120
7	RMSNorm_1_2	RMSNorm	1 × 40960 × 5120
8	FFN_1	SwiGLU	1 × 40960 × 5120
9	Add_1_ffn	Add	1 × 40960 × 5120
10	RMSNorm_2_1	RMSNorm	1 × 40960 × 5120
11	Attention_2	Grouped Query Attn	1 × 40960 × 5120
12	Add_2_attn	Add	1 × 40960 × 5120
13	RMSNorm_2_2	RMSNorm	1 × 40960 × 5120
14	FFN_2	SwiGLU	1 × 40960 × 5120
15	Add_2_ffn	Add	1 × 40960 × 5120
16	RMSNorm_3_1	RMSNorm	1 × 40960 × 5120
17	Attention_3	Grouped Query Attn	1 × 40960 × 5120
18	Add_3_attn	Add	1 × 40960 × 5120
19	RMSNorm_3_2	RMSNorm	1 × 40960 × 5120
20	FFN_3	SwiGLU	1 × 40960 × 5120
21	Add_3_ffn	Add	1 × 40960 × 5120
22	RMSNorm_4_1	RMSNorm	1 × 40960 × 5120
23	Attention_4	Grouped Query Attn	1 × 40960 × 5120
24	Add_4_attn	Add	1 × 40960 × 5120
25	RMSNorm_4_2	RMSNorm	1 × 40960 × 5120
26	FFN_4	SwiGLU	1 × 40960 × 5120
27	Add_4_ffn	Add	1 × 40960 × 5120
28	RMSNorm_5_1	RMSNorm	1 × 40960 × 5120
29	Attention_5	Grouped Query Attn	1 × 40960 × 5120
30	Add_5_attn	Add	1 × 40960 × 5120
31	RMSNorm_5_2	RMSNorm	1 × 40960 × 5120
32	FFN_5	SwiGLU	1 × 40960 × 5120
33	Add_5_ffn	Add	1 × 40960 × 5120
34	RMSNorm_6_1	RMSNorm	1 × 40960 × 5120
35	Attention_6	Grouped Query Attn	1 × 40960 × 5120
36	Add_6_attn	Add	1 × 40960 × 5120
37	RMSNorm_6_2	RMSNorm	1 × 40960 × 5120
38	FFN_6	SwiGLU	1 × 40960 × 5120
39	Add_6_ffn	Add	1 × 40960 × 5120
40	RMSNorm_7_1	RMSNorm	1 × 40960 × 5120
41	Attention_7	Grouped Query Attn	1 × 40960 × 5120
42	Add_7_attn	Add	1 × 40960 × 5120
43	RMSNorm_7_2	RMSNorm	1 × 40960 × 5120
44	FFN_7	SwiGLU	1 × 40960 × 5120
45	Add_7_ffn	Add	1 × 40960 × 5120
46	RMSNorm_8_1	RMSNorm	1 × 40960 × 5120
47	Attention_8	Grouped Query Attn	1 × 40960 × 5120
48	Add_8_attn	Add	1 × 40960 × 5120
49	RMSNorm_8_2	RMSNorm	1 × 40960 × 5120
50	FFN_8	SwiGLU	1 × 40960 × 5120
51	Add_8_ffn	Add	1 × 40960 × 5120
52	RMSNorm_9_1	RMSNorm	1 × 40960 × 5120
53	Attention_9	Grouped Query Attn	1 × 40960 × 5120
54	Add_9_attn	Add	1 × 40960 × 5120
55	RMSNorm_9_2	RMSNorm	1 × 40960 × 5120
56	FFN_9	SwiGLU	1 × 40960 × 5120
57	Add_9_ffn	Add	1 × 40960 × 5120
58	RMSNorm_10_1	RMSNorm	1 × 40960 × 5120
59	Attention_10	Grouped Query Attn	1 × 40960 × 5120
60	Add_10_attn	Add	1 × 40960 × 5120
61	RMSNorm_10_2	RMSNorm	1 × 40960 × 5120
62	FFN_10	SwiGLU	1 × 40960 × 5120
63	Add_10_ffn	Add	1 × 40960 × 5120
64	RMSNorm_11_1	RMSNorm	1 × 40960 × 5120
65	Attention_11	Grouped Query Attn	1 × 40960 × 5120
66	Add_11_attn	Add	1 × 40960 × 5120
67	RMSNorm_11_2	RMSNorm	1 × 40960 × 5120
68	FFN_11	SwiGLU	1 × 40960 × 5120
69	Add_11_ffn	Add	1 × 40960 × 5120
70	RMSNorm_12_1	RMSNorm	1 × 40960 × 5120
71	Attention_12	Grouped Query Attn	1 × 40960 × 5120
72	Add_12_attn	Add	1 × 40960 × 5120
73	RMSNorm_12_2	RMSNorm	1 × 40960 × 5120
74	FFN_12	SwiGLU	1 × 40960 × 5120
75	Add_12_ffn	Add	1 × 40960 × 5120
76	RMSNorm_13_1	RMSNorm	1 × 40960 × 5120
77	Attention_13	Grouped Query Attn	1 × 40960 × 5120
78	Add_13_attn	Add	1 × 40960 × 5120
79	RMSNorm_13_2	RMSNorm	1 × 40960 × 5120
80	FFN_13	SwiGLU	1 × 40960 × 5120
81	Add_13_ffn	Add	1 × 40960 × 5120
82	RMSNorm_14_1	RMSNorm	1 × 40960 × 5120
83	Attention_14	Grouped Query Attn	1 × 40960 × 5120
84	Add_14_attn	Add	1 × 40960 × 5120
85	RMSNorm_14_2	RMSNorm	1 × 40960 × 5120
86	FFN_14	SwiGLU	1 × 40960 × 5120
87	Add_14_ffn	Add	1 × 40960 × 5120
88	RMSNorm_15_1	RMSNorm	1 × 40960 × 5120
89	Attention_15	Grouped Query Attn	1 × 40960 × 5120
90	Add_15_attn	Add	1 × 40960 × 5120
91	RMSNorm_15_2	RMSNorm	1 × 40960 × 5120
92	FFN_15	SwiGLU	1 × 40960 × 5120
93	Add_15_ffn	Add	1 × 40960 × 5120
94	RMSNorm_16_1	RMSNorm	1 × 40960 × 5120
95	Attention_16	Grouped Query Attn	1 × 40960 × 5120
96	Add_16_attn	Add	1 × 40960 × 5120
97	RMSNorm_16_2	RMSNorm	1 × 40960 × 5120
98	FFN_16	SwiGLU	1 × 40960 × 5120
99	Add_16_ffn	Add	1 × 40960 × 5120
100	RMSNorm_17_1	RMSNorm	1 × 40960 × 5120
101	Attention_17	Grouped Query Attn	1 × 40960 × 5120
102	Add_17_attn	Add	1 × 40960 × 5120
103	RMSNorm_17_2	RMSNorm	1 × 40960 × 5120
104	FFN_17	SwiGLU	1 × 40960 × 5120
105	Add_17_ffn	Add	1 × 40960 × 5120
106	RMSNorm_18_1	RMSNorm	1 × 40960 × 5120
107	Attention_18	Grouped Query Attn	1 × 40960 × 5120
108	Add_18_attn	Add	1 × 40960 × 5120
109	RMSNorm_18_2	RMSNorm	1 × 40960 × 5120
110	FFN_18	SwiGLU	1 × 40960 × 5120
111	Add_18_ffn	Add	1 × 40960 × 5120
112	RMSNorm_19_1	RMSNorm	1 × 40960 × 5120
113	Attention_19	Grouped Query Attn	1 × 40960 × 5120
114	Add_19_attn	Add	1 × 40960 × 5120
115	RMSNorm_19_2	RMSNorm	1 × 40960 × 5120
116	FFN_19	SwiGLU	1 × 40960 × 5120
117	Add_19_ffn	Add	1 × 40960 × 5120
118	RMSNorm_20_1	RMSNorm	1 × 40960 × 5120
119	Attention_20	Grouped Query Attn	1 × 40960 × 5120
120	Add_20_attn	Add	1 × 40960 × 5120
121	RMSNorm_20_2	RMSNorm	1 × 40960 × 5120
122	FFN_20	SwiGLU	1 × 40960 × 5120
123	Add_20_ffn	Add	1 × 40960 × 5120
124	RMSNorm_21_1	RMSNorm	1 × 40960 × 5120
125	Attention_21	Grouped Query Attn	1 × 40960 × 5120
126	Add_21_attn	Add	1 × 40960 × 5120
127	RMSNorm_21_2	RMSNorm	1 × 40960 × 5120
128	FFN_21	SwiGLU	1 × 40960 × 5120
129	Add_21_ffn	Add	1 × 40960 × 5120
130	RMSNorm_22_1	RMSNorm	1 × 40960 × 5120
131	Attention_22	Grouped Query Attn	1 × 40960 × 5120
132	Add_22_attn	Add	1 × 40960 × 5120
133	RMSNorm_22_2	RMSNorm	1 × 40960 × 5120
134	FFN_22	SwiGLU	1 × 40960 × 5120
135	Add_22_ffn	Add	1 × 40960 × 5120
136	RMSNorm_23_1	RMSNorm	1 × 40960 × 5120
137	Attention_23	Grouped Query Attn	1 × 40960 × 5120
138	Add_23_attn	Add	1 × 40960 × 5120
139	RMSNorm_23_2	RMSNorm	1 × 40960 × 5120
140	FFN_23	SwiGLU	1 × 40960 × 5120
141	Add_23_ffn	Add	1 × 40960 × 5120
142	RMSNorm_24_1	RMSNorm	1 × 40960 × 5120
143	Attention_24	Grouped Query Attn	1 × 40960 × 5120
144	Add_24_attn	Add	1 × 40960 × 5120
145	RMSNorm_24_2	RMSNorm	1 × 40960 × 5120
146	FFN_24	SwiGLU	1 × 40960 × 5120
147	Add_24_ffn	Add	1 × 40960 × 5120
148	RMSNorm_25_1	RMSNorm	1 × 40960 × 5120
149	Attention_25	Grouped Query Attn	1 × 40960 × 5120
150	Add_25_attn	Add	1 × 40960 × 5120
151	RMSNorm_25_2	RMSNorm	1 × 40960 × 5120
152	FFN_25	SwiGLU	1 × 40960 × 5120
153	Add_25_ffn	Add	1 × 40960 × 5120
154	RMSNorm_26_1	RMSNorm	1 × 40960 × 5120
155	Attention_26	Grouped Query Attn	1 × 40960 × 5120
156	Add_26_attn	Add	1 × 40960 × 5120
157	RMSNorm_26_2	RMSNorm	1 × 40960 × 5120
158	FFN_26	SwiGLU	1 × 40960 × 5120
159	Add_26_ffn	Add	1 × 40960 × 5120
160	RMSNorm_27_1	RMSNorm	1 × 40960 × 5120
161	Attention_27	Grouped Query Attn	1 × 40960 × 5120
162	Add_27_attn	Add	1 × 40960 × 5120
163	RMSNorm_27_2	RMSNorm	1 × 40960 × 5120
164	FFN_27	SwiGLU	1 × 40960 × 5120
165	Add_27_ffn	Add	1 × 40960 × 5120
166	RMSNorm_28_1	RMSNorm	1 × 40960 × 5120
167	Attention_28	Grouped Query Attn	1 × 40960 × 5120
168	Add_28_attn	Add	1 × 40960 × 5120
169	RMSNorm_28_2	RMSNorm	1 × 40960 × 5120
170	FFN_28	SwiGLU	1 × 40960 × 5120
171	Add_28_ffn	Add	1 × 40960 × 5120
172	RMSNorm_29_1	RMSNorm	1 × 40960 × 5120
173	Attention_29	Grouped Query Attn	1 × 40960 × 5120
174	Add_29_attn	Add	1 × 40960 × 5120
175	RMSNorm_29_2	RMSNorm	1 × 40960 × 5120
176	FFN_29	SwiGLU	1 × 40960 × 5120
177	Add_29_ffn	Add	1 × 40960 × 5120
178	RMSNorm_30_1	RMSNorm	1 × 40960 × 5120
179	Attention_30	Grouped Query Attn	1 × 40960 × 5120
180	Add_30_attn	Add	1 × 40960 × 5120
181	RMSNorm_30_2	RMSNorm	1 × 40960 × 5120
182	FFN_30	SwiGLU	1 × 40960 × 5120
183	Add_30_ffn	Add	1 × 40960 × 5120
184	RMSNorm_31_1	RMSNorm	1 × 40960 × 5120
185	Attention_31	Grouped Query Attn	1 × 40960 × 5120
186	Add_31_attn	Add	1 × 40960 × 5120
187	RMSNorm_31_2	RMSNorm	1 × 40960 × 5120
188	FFN_31	SwiGLU	1 × 40960 × 5120
189	Add_31_ffn	Add	1 × 40960 × 5120
190	RMSNorm_32_1	RMSNorm	1 × 40960 × 5120
191	Attention_32	Grouped Query Attn	1 × 40960 × 5120
192	Add_32_attn	Add	1 × 40960 × 5120
193	RMSNorm_32_2	RMSNorm	1 × 40960 × 5120
194	FFN_32	SwiGLU	1 × 40960 × 5120
195	Add_32_ffn	Add	1 × 40960 × 5120
196	RMSNorm_33_1	RMSNorm	1 × 40960 × 5120
197	Attention_33	Grouped Query Attn	1 × 40960 × 5120
198	Add_33_attn	Add	1 × 40960 × 5120
199	RMSNorm_33_2	RMSNorm	1 × 40960 × 5120
200	FFN_33	SwiGLU	1 × 40960 × 5120
201	Add_33_ffn	Add	1 × 40960 × 5120
202	RMSNorm_34_1	RMSNorm	1 × 40960 × 5120
203	Attention_34	Grouped Query Attn	1 × 40960 × 5120
204	Add_34_attn	Add	1 × 40960 × 5120
205	RMSNorm_34_2	RMSNorm	1 × 40960 × 5120
206	FFN_34	SwiGLU	1 × 40960 × 5120
207	Add_34_ffn	Add	1 × 40960 × 5120
208	RMSNorm_35_1	RMSNorm	1 × 40960 × 5120
209	Attention_35	Grouped Query Attn	1 × 40960 × 5120
210	Add_35_attn	Add	1 × 40960 × 5120
211	RMSNorm_35_2	RMSNorm	1 × 40960 × 5120
212	FFN_35	SwiGLU	1 × 40960 × 5120
213	Add_35_ffn	Add	1 × 40960 × 5120
214	RMSNorm_36_1	RMSNorm	1 × 40960 × 5120
215	Attention_36	Grouped Query Attn	1 × 40960 × 5120
216	Add_36_attn	Add	1 × 40960 × 5120
217	RMSNorm_36_2	RMSNorm	1 × 40960 × 5120
218	FFN_36	SwiGLU	1 × 40960 × 5120
219	Add_36_ffn	Add	1 × 40960 × 5120
220	RMSNorm_37_1	RMSNorm	1 × 40960 × 5120
221	Attention_37	Grouped Query Attn	1 × 40960 × 5120
222	Add_37_attn	Add	1 × 40960 × 5120
223	RMSNorm_37_2	RMSNorm	1 × 40960 × 5120
224	FFN_37	SwiGLU	1 × 40960 × 5120
225	Add_37_ffn	Add	1 × 40960 × 5120
226	RMSNorm_38_1	RMSNorm	1 × 40960 × 5120
227	Attention_38	Grouped Query Attn	1 × 40960 × 5120
228	Add_38_attn	Add	1 × 40960 × 5120
229	RMSNorm_38_2	RMSNorm	1 × 40960 × 5120
230	FFN_38	SwiGLU	1 × 40960 × 5120
231	Add_38_ffn	Add	1 × 40960 × 5120
232	RMSNorm_39_1	RMSNorm	1 × 40960 × 5120
233	Attention_39	Grouped Query Attn	1 × 40960 × 5120
234	Add_39_attn	Add	1 × 40960 × 5120
235	RMSNorm_39_2	RMSNorm	1 × 40960 × 5120
236	FFN_39	SwiGLU	1 × 40960 × 5120
237	Add_39_ffn	Add	1 × 40960 × 5120
238	RMSNorm_40_1	RMSNorm	1 × 40960 × 5120
239	Attention_40	Grouped Query Attn	1 × 40960 × 5120
240	Add_40_attn	Add	1 × 40960 × 5120
241	RMSNorm_40_2	RMSNorm	1 × 40960 × 5120
242	FFN_40	SwiGLU	1 × 40960 × 5120
243	Add_40_ffn	Add	1 × 40960 × 5120
244	RMSNorm_41_1	RMSNorm	1 × 40960 × 5120
245	Attention_41	Grouped Query Attn	1 × 40960 × 5120
246	Add_41_attn	Add	1 × 40960 × 5120
247	RMSNorm_41_2	RMSNorm	1 × 40960 × 5120
248	FFN_41	SwiGLU	1 × 40960 × 5120
249	Add_41_ffn	Add	1 × 40960 × 5120
250	RMSNorm_42_1	RMSNorm	1 × 40960 × 5120
251	Attention_42	Grouped Query Attn	1 × 40960 × 5120
252	Add_42_attn	Add	1 × 40960 × 5120
253	RMSNorm_42_2	RMSNorm	1 × 40960 × 5120
254	FFN_42	SwiGLU	1 × 40960 × 5120
255	Add_42_ffn	Add	1 × 40960 × 5120
256	RMSNorm_43_1	RMSNorm	1 × 40960 × 5120
257	Attention_43	Grouped Query Attn	1 × 40960 × 5120
258	Add_43_attn	Add	1 × 40960 × 5120
259	RMSNorm_43_2	RMSNorm	1 × 40960 × 5120
260	FFN_43	SwiGLU	1 × 40960 × 5120
261	Add_43_ffn	Add	1 × 40960 × 5120
262	RMSNorm_44_1	RMSNorm	1 × 40960 × 5120
263	Attention_44	Grouped Query Attn	1 × 40960 × 5120
264	Add_44_attn	Add	1 × 40960 × 5120
265	RMSNorm_44_2	RMSNorm	1 × 40960 × 5120
266	FFN_44	SwiGLU	1 × 40960 × 5120
267	Add_44_ffn	Add	1 × 40960 × 5120
268	RMSNorm_45_1	RMSNorm	1 × 40960 × 5120
269	Attention_45	Grouped Query Attn	1 × 40960 × 5120
270	Add_45_attn	Add	1 × 40960 × 5120
271	RMSNorm_45_2	RMSNorm	1 × 40960 × 5120
272	FFN_45	SwiGLU	1 × 40960 × 5120
273	Add_45_ffn	Add	1 × 40960 × 5120
274	RMSNorm_46_1	RMSNorm	1 × 40960 × 5120
275	Attention_46	Grouped Query Attn	1 × 40960 × 5120
276	Add_46_attn	Add	1 × 40960 × 5120
277	RMSNorm_46_2	RMSNorm	1 × 40960 × 5120
278	FFN_46	SwiGLU	1 × 40960 × 5120
279	Add_46_ffn	Add	1 × 40960 × 5120
280	RMSNorm_47_1	RMSNorm	1 × 40960 × 5120
281	Attention_47	Grouped Query Attn	1 × 40960 × 5120
282	Add_47_attn	Add	1 × 40960 × 5120
283	RMSNorm_47_2	RMSNorm	1 × 40960 × 5120
284	FFN_47	SwiGLU	1 × 40960 × 5120
285	Add_47_ffn	Add	1 × 40960 × 5120
286	RMSNorm_48_1	RMSNorm	1 × 40960 × 5120
287	Attention_48	Grouped Query Attn	1 × 40960 × 5120
288	Add_48_attn	Add	1 × 40960 × 5120
289	RMSNorm_48_2	RMSNorm	1 × 40960 × 5120
290	FFN_48	SwiGLU	1 × 40960 × 5120
291	Add_48_ffn	Add	1 × 40960 × 5120
292	RMSNorm_49_1	RMSNorm	1 × 40960 × 5120
293	Attention_49	Grouped Query Attn	1 × 40960 × 5120
294	Add_49_attn	Add	1 × 40960 × 5120
295	RMSNorm_49_2	RMSNorm	1 × 40960 × 5120
296	FFN_49	SwiGLU	1 × 40960 × 5120
297	Add_49_ffn	Add	1 × 40960 × 5120
298	RMSNorm_50_1	RMSNorm	1 × 40960 × 5120
299	Attention_50	Grouped Query Attn	1 × 40960 × 5120
300	Add_50_attn	Add	1 × 40960 × 5120
301	RMSNorm_50_2	RMSNorm	1 × 40960 × 5120
302	FFN_50	SwiGLU	1 × 40960 × 5120
303	Add_50_ffn	Add	1 × 40960 × 5120
304	RMSNorm_51_1	RMSNorm	1 × 40960 × 5120
305	Attention_51	Grouped Query Attn	1 × 40960 × 5120
306	Add_51_attn	Add	1 × 40960 × 5120
307	RMSNorm_51_2	RMSNorm	1 × 40960 × 5120
308	FFN_51	SwiGLU	1 × 40960 × 5120
309	Add_51_ffn	Add	1 × 40960 × 5120
310	RMSNorm_52_1	RMSNorm	1 × 40960 × 5120
311	Attention_52	Grouped Query Attn	1 × 40960 × 5120
312	Add_52_attn	Add	1 × 40960 × 5120
313	RMSNorm_52_2	RMSNorm	1 × 40960 × 5120
314	FFN_52	SwiGLU	1 × 40960 × 5120
315	Add_52_ffn	Add	1 × 40960 × 5120
316	RMSNorm_53_1	RMSNorm	1 × 40960 × 5120
317	Attention_53	Grouped Query Attn	1 × 40960 × 5120
318	Add_53_attn	Add	1 × 40960 × 5120
319	RMSNorm_53_2	RMSNorm	1 × 40960 × 5120
320	FFN_53	SwiGLU	1 × 40960 × 5120
321	Add_53_ffn	Add	1 × 40960 × 5120
322	RMSNorm_54_1	RMSNorm	1 × 40960 × 5120
323	Attention_54	Grouped Query Attn	1 × 40960 × 5120
324	Add_54_attn	Add	1 × 40960 × 5120
325	RMSNorm_54_2	RMSNorm	1 × 40960 × 5120
326	FFN_54	SwiGLU	1 × 40960 × 5120
327	Add_54_ffn	Add	1 × 40960 × 5120
328	RMSNorm_55_1	RMSNorm	1 × 40960 × 5120
329	Attention_55	Grouped Query Attn	1 × 40960 × 5120
330	Add_55_attn	Add	1 × 40960 × 5120
331	RMSNorm_55_2	RMSNorm	1 × 40960 × 5120
332	FFN_55	SwiGLU	1 × 40960 × 5120
333	Add_55_ffn	Add	1 × 40960 × 5120
334	RMSNorm_56_1	RMSNorm	1 × 40960 × 5120
335	Attention_56	Grouped Query Attn	1 × 40960 × 5120
336	Add_56_attn	Add	1 × 40960 × 5120
337	RMSNorm_56_2	RMSNorm	1 × 40960 × 5120
338	FFN_56	SwiGLU	1 × 40960 × 5120
339	Add_56_ffn	Add	1 × 40960 × 5120
340	RMSNorm_57_1	RMSNorm	1 × 40960 × 5120
341	Attention_57	Grouped Query Attn	1 × 40960 × 5120
342	Add_57_attn	Add	1 × 40960 × 5120
343	RMSNorm_57_2	RMSNorm	1 × 40960 × 5120
344	FFN_57	SwiGLU	1 × 40960 × 5120
345	Add_57_ffn	Add	1 × 40960 × 5120
346	RMSNorm_58_1	RMSNorm	1 × 40960 × 5120
347	Attention_58	Grouped Query Attn	1 × 40960 × 5120
348	Add_58_attn	Add	1 × 40960 × 5120
349	RMSNorm_58_2	RMSNorm	1 × 40960 × 5120
350	FFN_58	SwiGLU	1 × 40960 × 5120
351	Add_58_ffn	Add	1 × 40960 × 5120
352	RMSNorm_59_1	RMSNorm	1 × 40960 × 5120
353	Attention_59	Grouped Query Attn	1 × 40960 × 5120
354	Add_59_attn	Add	1 × 40960 × 5120
355	RMSNorm_59_2	RMSNorm	1 × 40960 × 5120
356	FFN_59	SwiGLU	1 × 40960 × 5120
357	Add_59_ffn	Add	1 × 40960 × 5120
358	RMSNorm_60_1	RMSNorm	1 × 40960 × 5120
359	Attention_60	Grouped Query Attn	1 × 40960 × 5120
360	Add_60_attn	Add	1 × 40960 × 5120
361	RMSNorm_60_2	RMSNorm	1 × 40960 × 5120
362	FFN_60	SwiGLU	1 × 40960 × 5120
363	Add_60_ffn	Add	1 × 40960 × 5120
364	RMSNorm_61_1	RMSNorm	1 × 40960 × 5120
365	Attention_61	Grouped Query Attn	1 × 40960 × 5120
366	Add_61_attn	Add	1 × 40960 × 5120
367	RMSNorm_61_2	RMSNorm	1 × 40960 × 5120
368	FFN_61	SwiGLU	1 × 40960 × 5120
369	Add_61_ffn	Add	1 × 40960 × 5120
370	RMSNorm_62_1	RMSNorm	1 × 40960 × 5120
371	Attention_62	Grouped Query Attn	1 × 40960 × 5120
372	Add_62_attn	Add	1 × 40960 × 5120
373	RMSNorm_62_2	RMSNorm	1 × 40960 × 5120
374	FFN_62	SwiGLU	1 × 40960 × 5120
375	Add_62_ffn	Add	1 × 40960 × 5120
376	RMSNorm_63_1	RMSNorm	1 × 40960 × 5120
377	Attention_63	Grouped Query Attn	1 × 40960 × 5120
378	Add_63_attn	Add	1 × 40960 × 5120
379	RMSNorm_63_2	RMSNorm	1 × 40960 × 5120
380	FFN_63	SwiGLU	1 × 40960 × 5120
381	Add_63_ffn	Add	1 × 40960 × 5120
382	RMSNorm_64_1	RMSNorm	1 × 40960 × 5120
383	Attention_64	Grouped Query Attn	1 × 40960 × 5120
384	Add_64_attn	Add	1 × 40960 × 5120
385	RMSNorm_64_2	RMSNorm	1 × 40960 × 5120
386	FFN_64	SwiGLU	1 × 40960 × 5120
387	Add_64_ffn	Add	1 × 40960 × 5120
388	Final_RMSNorm	RMSNorm	1 × 40960 × 5120
389	LM_Head	Linear	1 × 40960 × 151936
390	Output	Output	1 × 40960 × 151936

What the verifier says

infoAt 64 stacked attention layers, residual-branch outputs add up; unscaled init lets activation variance grow with depth. GPT-2/LLaMA-family models scale the residual projections by depth (N(0, 0.02 / √(2L))). Fix: Scale residual output projections by depth: nn.init.normal_(w, std=0.02 / math.sqrt(2 * n_layers))
deep-attention-default-init

Do this to your own model

Same numbers, on a model in your repo, in one command. No account.

pip install neurarch-trace
neurarch-trace Qwen/Qwen3-32B --plan --share

Paste a model instead Markdown twin Index JSON POST /api/v1/plan

Other qwen3 checkpoints

jina-reranker-v3

596M derived · -0.13% against the checkpoint

Kimi-K3-DSpark

4.26B derived · +89.3% against the checkpoint

Qwen3-0.6B

596M derived · -20.7% against the checkpoint

Qwen3-1.7B

1.72B derived · -15.3% against the checkpoint