Models / exaone

EXAONE-3.5-32B-Instruct-AWQ

Reconstructed from its own config.json with no weights read. 172K 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

23.03B

23,030,233,088 parameters

In the published checkpoint

32.00B

32,003,200,000 scalars · safetensors.total, read 2026-02-06

Delta

-28.0%

quantized This checkpoint is stored quantized (awq). The tensor count in the file counts stored elements under a packing scheme, not logical parameters, so the two numbers below are not measuring the same thing in either direction.

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

$40767.22

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 × 32768
2	Embedding	Embedding	1 × 32768 × 5120
3	RoPE	RoPE	1 × 32768 × 5120
4	LayerNorm_1_1	LayerNorm	1 × 32768 × 5120
5	Attention_1	Grouped Query Attn	1 × 32768 × 5120
6	Add_1_attn	Add	1 × 32768 × 5120
7	LayerNorm_1_2	LayerNorm	1 × 32768 × 5120
8	FFN_1	Feed Forward	1 × 32768 × 5120
9	Add_1_ffn	Add	1 × 32768 × 5120
10	LayerNorm_2_1	LayerNorm	1 × 32768 × 5120
11	Attention_2	Grouped Query Attn	1 × 32768 × 5120
12	Add_2_attn	Add	1 × 32768 × 5120
13	LayerNorm_2_2	LayerNorm	1 × 32768 × 5120
14	FFN_2	Feed Forward	1 × 32768 × 5120
15	Add_2_ffn	Add	1 × 32768 × 5120
16	LayerNorm_3_1	LayerNorm	1 × 32768 × 5120
17	Attention_3	Grouped Query Attn	1 × 32768 × 5120
18	Add_3_attn	Add	1 × 32768 × 5120
19	LayerNorm_3_2	LayerNorm	1 × 32768 × 5120
20	FFN_3	Feed Forward	1 × 32768 × 5120
21	Add_3_ffn	Add	1 × 32768 × 5120
22	LayerNorm_4_1	LayerNorm	1 × 32768 × 5120
23	Attention_4	Grouped Query Attn	1 × 32768 × 5120
24	Add_4_attn	Add	1 × 32768 × 5120
25	LayerNorm_4_2	LayerNorm	1 × 32768 × 5120
26	FFN_4	Feed Forward	1 × 32768 × 5120
27	Add_4_ffn	Add	1 × 32768 × 5120
28	LayerNorm_5_1	LayerNorm	1 × 32768 × 5120
29	Attention_5	Grouped Query Attn	1 × 32768 × 5120
30	Add_5_attn	Add	1 × 32768 × 5120
31	LayerNorm_5_2	LayerNorm	1 × 32768 × 5120
32	FFN_5	Feed Forward	1 × 32768 × 5120
33	Add_5_ffn	Add	1 × 32768 × 5120
34	LayerNorm_6_1	LayerNorm	1 × 32768 × 5120
35	Attention_6	Grouped Query Attn	1 × 32768 × 5120
36	Add_6_attn	Add	1 × 32768 × 5120
37	LayerNorm_6_2	LayerNorm	1 × 32768 × 5120
38	FFN_6	Feed Forward	1 × 32768 × 5120
39	Add_6_ffn	Add	1 × 32768 × 5120
40	LayerNorm_7_1	LayerNorm	1 × 32768 × 5120
41	Attention_7	Grouped Query Attn	1 × 32768 × 5120
42	Add_7_attn	Add	1 × 32768 × 5120
43	LayerNorm_7_2	LayerNorm	1 × 32768 × 5120
44	FFN_7	Feed Forward	1 × 32768 × 5120
45	Add_7_ffn	Add	1 × 32768 × 5120
46	LayerNorm_8_1	LayerNorm	1 × 32768 × 5120
47	Attention_8	Grouped Query Attn	1 × 32768 × 5120
48	Add_8_attn	Add	1 × 32768 × 5120
49	LayerNorm_8_2	LayerNorm	1 × 32768 × 5120
50	FFN_8	Feed Forward	1 × 32768 × 5120
51	Add_8_ffn	Add	1 × 32768 × 5120
52	LayerNorm_9_1	LayerNorm	1 × 32768 × 5120
53	Attention_9	Grouped Query Attn	1 × 32768 × 5120
54	Add_9_attn	Add	1 × 32768 × 5120
55	LayerNorm_9_2	LayerNorm	1 × 32768 × 5120
56	FFN_9	Feed Forward	1 × 32768 × 5120
57	Add_9_ffn	Add	1 × 32768 × 5120
58	LayerNorm_10_1	LayerNorm	1 × 32768 × 5120
59	Attention_10	Grouped Query Attn	1 × 32768 × 5120
60	Add_10_attn	Add	1 × 32768 × 5120
61	LayerNorm_10_2	LayerNorm	1 × 32768 × 5120
62	FFN_10	Feed Forward	1 × 32768 × 5120
63	Add_10_ffn	Add	1 × 32768 × 5120
64	LayerNorm_11_1	LayerNorm	1 × 32768 × 5120
65	Attention_11	Grouped Query Attn	1 × 32768 × 5120
66	Add_11_attn	Add	1 × 32768 × 5120
67	LayerNorm_11_2	LayerNorm	1 × 32768 × 5120
68	FFN_11	Feed Forward	1 × 32768 × 5120
69	Add_11_ffn	Add	1 × 32768 × 5120
70	LayerNorm_12_1	LayerNorm	1 × 32768 × 5120
71	Attention_12	Grouped Query Attn	1 × 32768 × 5120
72	Add_12_attn	Add	1 × 32768 × 5120
73	LayerNorm_12_2	LayerNorm	1 × 32768 × 5120
74	FFN_12	Feed Forward	1 × 32768 × 5120
75	Add_12_ffn	Add	1 × 32768 × 5120
76	LayerNorm_13_1	LayerNorm	1 × 32768 × 5120
77	Attention_13	Grouped Query Attn	1 × 32768 × 5120
78	Add_13_attn	Add	1 × 32768 × 5120
79	LayerNorm_13_2	LayerNorm	1 × 32768 × 5120
80	FFN_13	Feed Forward	1 × 32768 × 5120
81	Add_13_ffn	Add	1 × 32768 × 5120
82	LayerNorm_14_1	LayerNorm	1 × 32768 × 5120
83	Attention_14	Grouped Query Attn	1 × 32768 × 5120
84	Add_14_attn	Add	1 × 32768 × 5120
85	LayerNorm_14_2	LayerNorm	1 × 32768 × 5120
86	FFN_14	Feed Forward	1 × 32768 × 5120
87	Add_14_ffn	Add	1 × 32768 × 5120
88	LayerNorm_15_1	LayerNorm	1 × 32768 × 5120
89	Attention_15	Grouped Query Attn	1 × 32768 × 5120
90	Add_15_attn	Add	1 × 32768 × 5120
91	LayerNorm_15_2	LayerNorm	1 × 32768 × 5120
92	FFN_15	Feed Forward	1 × 32768 × 5120
93	Add_15_ffn	Add	1 × 32768 × 5120
94	LayerNorm_16_1	LayerNorm	1 × 32768 × 5120
95	Attention_16	Grouped Query Attn	1 × 32768 × 5120
96	Add_16_attn	Add	1 × 32768 × 5120
97	LayerNorm_16_2	LayerNorm	1 × 32768 × 5120
98	FFN_16	Feed Forward	1 × 32768 × 5120
99	Add_16_ffn	Add	1 × 32768 × 5120
100	LayerNorm_17_1	LayerNorm	1 × 32768 × 5120
101	Attention_17	Grouped Query Attn	1 × 32768 × 5120
102	Add_17_attn	Add	1 × 32768 × 5120
103	LayerNorm_17_2	LayerNorm	1 × 32768 × 5120
104	FFN_17	Feed Forward	1 × 32768 × 5120
105	Add_17_ffn	Add	1 × 32768 × 5120
106	LayerNorm_18_1	LayerNorm	1 × 32768 × 5120
107	Attention_18	Grouped Query Attn	1 × 32768 × 5120
108	Add_18_attn	Add	1 × 32768 × 5120
109	LayerNorm_18_2	LayerNorm	1 × 32768 × 5120
110	FFN_18	Feed Forward	1 × 32768 × 5120
111	Add_18_ffn	Add	1 × 32768 × 5120
112	LayerNorm_19_1	LayerNorm	1 × 32768 × 5120
113	Attention_19	Grouped Query Attn	1 × 32768 × 5120
114	Add_19_attn	Add	1 × 32768 × 5120
115	LayerNorm_19_2	LayerNorm	1 × 32768 × 5120
116	FFN_19	Feed Forward	1 × 32768 × 5120
117	Add_19_ffn	Add	1 × 32768 × 5120
118	LayerNorm_20_1	LayerNorm	1 × 32768 × 5120
119	Attention_20	Grouped Query Attn	1 × 32768 × 5120
120	Add_20_attn	Add	1 × 32768 × 5120
121	LayerNorm_20_2	LayerNorm	1 × 32768 × 5120
122	FFN_20	Feed Forward	1 × 32768 × 5120
123	Add_20_ffn	Add	1 × 32768 × 5120
124	LayerNorm_21_1	LayerNorm	1 × 32768 × 5120
125	Attention_21	Grouped Query Attn	1 × 32768 × 5120
126	Add_21_attn	Add	1 × 32768 × 5120
127	LayerNorm_21_2	LayerNorm	1 × 32768 × 5120
128	FFN_21	Feed Forward	1 × 32768 × 5120
129	Add_21_ffn	Add	1 × 32768 × 5120
130	LayerNorm_22_1	LayerNorm	1 × 32768 × 5120
131	Attention_22	Grouped Query Attn	1 × 32768 × 5120
132	Add_22_attn	Add	1 × 32768 × 5120
133	LayerNorm_22_2	LayerNorm	1 × 32768 × 5120
134	FFN_22	Feed Forward	1 × 32768 × 5120
135	Add_22_ffn	Add	1 × 32768 × 5120
136	LayerNorm_23_1	LayerNorm	1 × 32768 × 5120
137	Attention_23	Grouped Query Attn	1 × 32768 × 5120
138	Add_23_attn	Add	1 × 32768 × 5120
139	LayerNorm_23_2	LayerNorm	1 × 32768 × 5120
140	FFN_23	Feed Forward	1 × 32768 × 5120
141	Add_23_ffn	Add	1 × 32768 × 5120
142	LayerNorm_24_1	LayerNorm	1 × 32768 × 5120
143	Attention_24	Grouped Query Attn	1 × 32768 × 5120
144	Add_24_attn	Add	1 × 32768 × 5120
145	LayerNorm_24_2	LayerNorm	1 × 32768 × 5120
146	FFN_24	Feed Forward	1 × 32768 × 5120
147	Add_24_ffn	Add	1 × 32768 × 5120
148	LayerNorm_25_1	LayerNorm	1 × 32768 × 5120
149	Attention_25	Grouped Query Attn	1 × 32768 × 5120
150	Add_25_attn	Add	1 × 32768 × 5120
151	LayerNorm_25_2	LayerNorm	1 × 32768 × 5120
152	FFN_25	Feed Forward	1 × 32768 × 5120
153	Add_25_ffn	Add	1 × 32768 × 5120
154	LayerNorm_26_1	LayerNorm	1 × 32768 × 5120
155	Attention_26	Grouped Query Attn	1 × 32768 × 5120
156	Add_26_attn	Add	1 × 32768 × 5120
157	LayerNorm_26_2	LayerNorm	1 × 32768 × 5120
158	FFN_26	Feed Forward	1 × 32768 × 5120
159	Add_26_ffn	Add	1 × 32768 × 5120
160	LayerNorm_27_1	LayerNorm	1 × 32768 × 5120
161	Attention_27	Grouped Query Attn	1 × 32768 × 5120
162	Add_27_attn	Add	1 × 32768 × 5120
163	LayerNorm_27_2	LayerNorm	1 × 32768 × 5120
164	FFN_27	Feed Forward	1 × 32768 × 5120
165	Add_27_ffn	Add	1 × 32768 × 5120
166	LayerNorm_28_1	LayerNorm	1 × 32768 × 5120
167	Attention_28	Grouped Query Attn	1 × 32768 × 5120
168	Add_28_attn	Add	1 × 32768 × 5120
169	LayerNorm_28_2	LayerNorm	1 × 32768 × 5120
170	FFN_28	Feed Forward	1 × 32768 × 5120
171	Add_28_ffn	Add	1 × 32768 × 5120
172	LayerNorm_29_1	LayerNorm	1 × 32768 × 5120
173	Attention_29	Grouped Query Attn	1 × 32768 × 5120
174	Add_29_attn	Add	1 × 32768 × 5120
175	LayerNorm_29_2	LayerNorm	1 × 32768 × 5120
176	FFN_29	Feed Forward	1 × 32768 × 5120
177	Add_29_ffn	Add	1 × 32768 × 5120
178	LayerNorm_30_1	LayerNorm	1 × 32768 × 5120
179	Attention_30	Grouped Query Attn	1 × 32768 × 5120
180	Add_30_attn	Add	1 × 32768 × 5120
181	LayerNorm_30_2	LayerNorm	1 × 32768 × 5120
182	FFN_30	Feed Forward	1 × 32768 × 5120
183	Add_30_ffn	Add	1 × 32768 × 5120
184	LayerNorm_31_1	LayerNorm	1 × 32768 × 5120
185	Attention_31	Grouped Query Attn	1 × 32768 × 5120
186	Add_31_attn	Add	1 × 32768 × 5120
187	LayerNorm_31_2	LayerNorm	1 × 32768 × 5120
188	FFN_31	Feed Forward	1 × 32768 × 5120
189	Add_31_ffn	Add	1 × 32768 × 5120
190	LayerNorm_32_1	LayerNorm	1 × 32768 × 5120
191	Attention_32	Grouped Query Attn	1 × 32768 × 5120
192	Add_32_attn	Add	1 × 32768 × 5120
193	LayerNorm_32_2	LayerNorm	1 × 32768 × 5120
194	FFN_32	Feed Forward	1 × 32768 × 5120
195	Add_32_ffn	Add	1 × 32768 × 5120
196	LayerNorm_33_1	LayerNorm	1 × 32768 × 5120
197	Attention_33	Grouped Query Attn	1 × 32768 × 5120
198	Add_33_attn	Add	1 × 32768 × 5120
199	LayerNorm_33_2	LayerNorm	1 × 32768 × 5120
200	FFN_33	Feed Forward	1 × 32768 × 5120
201	Add_33_ffn	Add	1 × 32768 × 5120
202	LayerNorm_34_1	LayerNorm	1 × 32768 × 5120
203	Attention_34	Grouped Query Attn	1 × 32768 × 5120
204	Add_34_attn	Add	1 × 32768 × 5120
205	LayerNorm_34_2	LayerNorm	1 × 32768 × 5120
206	FFN_34	Feed Forward	1 × 32768 × 5120
207	Add_34_ffn	Add	1 × 32768 × 5120
208	LayerNorm_35_1	LayerNorm	1 × 32768 × 5120
209	Attention_35	Grouped Query Attn	1 × 32768 × 5120
210	Add_35_attn	Add	1 × 32768 × 5120
211	LayerNorm_35_2	LayerNorm	1 × 32768 × 5120
212	FFN_35	Feed Forward	1 × 32768 × 5120
213	Add_35_ffn	Add	1 × 32768 × 5120
214	LayerNorm_36_1	LayerNorm	1 × 32768 × 5120
215	Attention_36	Grouped Query Attn	1 × 32768 × 5120
216	Add_36_attn	Add	1 × 32768 × 5120
217	LayerNorm_36_2	LayerNorm	1 × 32768 × 5120
218	FFN_36	Feed Forward	1 × 32768 × 5120
219	Add_36_ffn	Add	1 × 32768 × 5120
220	LayerNorm_37_1	LayerNorm	1 × 32768 × 5120
221	Attention_37	Grouped Query Attn	1 × 32768 × 5120
222	Add_37_attn	Add	1 × 32768 × 5120
223	LayerNorm_37_2	LayerNorm	1 × 32768 × 5120
224	FFN_37	Feed Forward	1 × 32768 × 5120
225	Add_37_ffn	Add	1 × 32768 × 5120
226	LayerNorm_38_1	LayerNorm	1 × 32768 × 5120
227	Attention_38	Grouped Query Attn	1 × 32768 × 5120
228	Add_38_attn	Add	1 × 32768 × 5120
229	LayerNorm_38_2	LayerNorm	1 × 32768 × 5120
230	FFN_38	Feed Forward	1 × 32768 × 5120
231	Add_38_ffn	Add	1 × 32768 × 5120
232	LayerNorm_39_1	LayerNorm	1 × 32768 × 5120
233	Attention_39	Grouped Query Attn	1 × 32768 × 5120
234	Add_39_attn	Add	1 × 32768 × 5120
235	LayerNorm_39_2	LayerNorm	1 × 32768 × 5120
236	FFN_39	Feed Forward	1 × 32768 × 5120
237	Add_39_ffn	Add	1 × 32768 × 5120
238	LayerNorm_40_1	LayerNorm	1 × 32768 × 5120
239	Attention_40	Grouped Query Attn	1 × 32768 × 5120
240	Add_40_attn	Add	1 × 32768 × 5120
241	LayerNorm_40_2	LayerNorm	1 × 32768 × 5120
242	FFN_40	Feed Forward	1 × 32768 × 5120
243	Add_40_ffn	Add	1 × 32768 × 5120
244	LayerNorm_41_1	LayerNorm	1 × 32768 × 5120
245	Attention_41	Grouped Query Attn	1 × 32768 × 5120
246	Add_41_attn	Add	1 × 32768 × 5120
247	LayerNorm_41_2	LayerNorm	1 × 32768 × 5120
248	FFN_41	Feed Forward	1 × 32768 × 5120
249	Add_41_ffn	Add	1 × 32768 × 5120
250	LayerNorm_42_1	LayerNorm	1 × 32768 × 5120
251	Attention_42	Grouped Query Attn	1 × 32768 × 5120
252	Add_42_attn	Add	1 × 32768 × 5120
253	LayerNorm_42_2	LayerNorm	1 × 32768 × 5120
254	FFN_42	Feed Forward	1 × 32768 × 5120
255	Add_42_ffn	Add	1 × 32768 × 5120
256	LayerNorm_43_1	LayerNorm	1 × 32768 × 5120
257	Attention_43	Grouped Query Attn	1 × 32768 × 5120
258	Add_43_attn	Add	1 × 32768 × 5120
259	LayerNorm_43_2	LayerNorm	1 × 32768 × 5120
260	FFN_43	Feed Forward	1 × 32768 × 5120
261	Add_43_ffn	Add	1 × 32768 × 5120
262	LayerNorm_44_1	LayerNorm	1 × 32768 × 5120
263	Attention_44	Grouped Query Attn	1 × 32768 × 5120
264	Add_44_attn	Add	1 × 32768 × 5120
265	LayerNorm_44_2	LayerNorm	1 × 32768 × 5120
266	FFN_44	Feed Forward	1 × 32768 × 5120
267	Add_44_ffn	Add	1 × 32768 × 5120
268	LayerNorm_45_1	LayerNorm	1 × 32768 × 5120
269	Attention_45	Grouped Query Attn	1 × 32768 × 5120
270	Add_45_attn	Add	1 × 32768 × 5120
271	LayerNorm_45_2	LayerNorm	1 × 32768 × 5120
272	FFN_45	Feed Forward	1 × 32768 × 5120
273	Add_45_ffn	Add	1 × 32768 × 5120
274	LayerNorm_46_1	LayerNorm	1 × 32768 × 5120
275	Attention_46	Grouped Query Attn	1 × 32768 × 5120
276	Add_46_attn	Add	1 × 32768 × 5120
277	LayerNorm_46_2	LayerNorm	1 × 32768 × 5120
278	FFN_46	Feed Forward	1 × 32768 × 5120
279	Add_46_ffn	Add	1 × 32768 × 5120
280	LayerNorm_47_1	LayerNorm	1 × 32768 × 5120
281	Attention_47	Grouped Query Attn	1 × 32768 × 5120
282	Add_47_attn	Add	1 × 32768 × 5120
283	LayerNorm_47_2	LayerNorm	1 × 32768 × 5120
284	FFN_47	Feed Forward	1 × 32768 × 5120
285	Add_47_ffn	Add	1 × 32768 × 5120
286	LayerNorm_48_1	LayerNorm	1 × 32768 × 5120
287	Attention_48	Grouped Query Attn	1 × 32768 × 5120
288	Add_48_attn	Add	1 × 32768 × 5120
289	LayerNorm_48_2	LayerNorm	1 × 32768 × 5120
290	FFN_48	Feed Forward	1 × 32768 × 5120
291	Add_48_ffn	Add	1 × 32768 × 5120
292	LayerNorm_49_1	LayerNorm	1 × 32768 × 5120
293	Attention_49	Grouped Query Attn	1 × 32768 × 5120
294	Add_49_attn	Add	1 × 32768 × 5120
295	LayerNorm_49_2	LayerNorm	1 × 32768 × 5120
296	FFN_49	Feed Forward	1 × 32768 × 5120
297	Add_49_ffn	Add	1 × 32768 × 5120
298	LayerNorm_50_1	LayerNorm	1 × 32768 × 5120
299	Attention_50	Grouped Query Attn	1 × 32768 × 5120
300	Add_50_attn	Add	1 × 32768 × 5120
301	LayerNorm_50_2	LayerNorm	1 × 32768 × 5120
302	FFN_50	Feed Forward	1 × 32768 × 5120
303	Add_50_ffn	Add	1 × 32768 × 5120
304	LayerNorm_51_1	LayerNorm	1 × 32768 × 5120
305	Attention_51	Grouped Query Attn	1 × 32768 × 5120
306	Add_51_attn	Add	1 × 32768 × 5120
307	LayerNorm_51_2	LayerNorm	1 × 32768 × 5120
308	FFN_51	Feed Forward	1 × 32768 × 5120
309	Add_51_ffn	Add	1 × 32768 × 5120
310	LayerNorm_52_1	LayerNorm	1 × 32768 × 5120
311	Attention_52	Grouped Query Attn	1 × 32768 × 5120
312	Add_52_attn	Add	1 × 32768 × 5120
313	LayerNorm_52_2	LayerNorm	1 × 32768 × 5120
314	FFN_52	Feed Forward	1 × 32768 × 5120
315	Add_52_ffn	Add	1 × 32768 × 5120
316	LayerNorm_53_1	LayerNorm	1 × 32768 × 5120
317	Attention_53	Grouped Query Attn	1 × 32768 × 5120
318	Add_53_attn	Add	1 × 32768 × 5120
319	LayerNorm_53_2	LayerNorm	1 × 32768 × 5120
320	FFN_53	Feed Forward	1 × 32768 × 5120
321	Add_53_ffn	Add	1 × 32768 × 5120
322	LayerNorm_54_1	LayerNorm	1 × 32768 × 5120
323	Attention_54	Grouped Query Attn	1 × 32768 × 5120
324	Add_54_attn	Add	1 × 32768 × 5120
325	LayerNorm_54_2	LayerNorm	1 × 32768 × 5120
326	FFN_54	Feed Forward	1 × 32768 × 5120
327	Add_54_ffn	Add	1 × 32768 × 5120
328	LayerNorm_55_1	LayerNorm	1 × 32768 × 5120
329	Attention_55	Grouped Query Attn	1 × 32768 × 5120
330	Add_55_attn	Add	1 × 32768 × 5120
331	LayerNorm_55_2	LayerNorm	1 × 32768 × 5120
332	FFN_55	Feed Forward	1 × 32768 × 5120
333	Add_55_ffn	Add	1 × 32768 × 5120
334	LayerNorm_56_1	LayerNorm	1 × 32768 × 5120
335	Attention_56	Grouped Query Attn	1 × 32768 × 5120
336	Add_56_attn	Add	1 × 32768 × 5120
337	LayerNorm_56_2	LayerNorm	1 × 32768 × 5120
338	FFN_56	Feed Forward	1 × 32768 × 5120
339	Add_56_ffn	Add	1 × 32768 × 5120
340	LayerNorm_57_1	LayerNorm	1 × 32768 × 5120
341	Attention_57	Grouped Query Attn	1 × 32768 × 5120
342	Add_57_attn	Add	1 × 32768 × 5120
343	LayerNorm_57_2	LayerNorm	1 × 32768 × 5120
344	FFN_57	Feed Forward	1 × 32768 × 5120
345	Add_57_ffn	Add	1 × 32768 × 5120
346	LayerNorm_58_1	LayerNorm	1 × 32768 × 5120
347	Attention_58	Grouped Query Attn	1 × 32768 × 5120
348	Add_58_attn	Add	1 × 32768 × 5120
349	LayerNorm_58_2	LayerNorm	1 × 32768 × 5120
350	FFN_58	Feed Forward	1 × 32768 × 5120
351	Add_58_ffn	Add	1 × 32768 × 5120
352	LayerNorm_59_1	LayerNorm	1 × 32768 × 5120
353	Attention_59	Grouped Query Attn	1 × 32768 × 5120
354	Add_59_attn	Add	1 × 32768 × 5120
355	LayerNorm_59_2	LayerNorm	1 × 32768 × 5120
356	FFN_59	Feed Forward	1 × 32768 × 5120
357	Add_59_ffn	Add	1 × 32768 × 5120
358	LayerNorm_60_1	LayerNorm	1 × 32768 × 5120
359	Attention_60	Grouped Query Attn	1 × 32768 × 5120
360	Add_60_attn	Add	1 × 32768 × 5120
361	LayerNorm_60_2	LayerNorm	1 × 32768 × 5120
362	FFN_60	Feed Forward	1 × 32768 × 5120
363	Add_60_ffn	Add	1 × 32768 × 5120
364	LayerNorm_61_1	LayerNorm	1 × 32768 × 5120
365	Attention_61	Grouped Query Attn	1 × 32768 × 5120
366	Add_61_attn	Add	1 × 32768 × 5120
367	LayerNorm_61_2	LayerNorm	1 × 32768 × 5120
368	FFN_61	Feed Forward	1 × 32768 × 5120
369	Add_61_ffn	Add	1 × 32768 × 5120
370	LayerNorm_62_1	LayerNorm	1 × 32768 × 5120
371	Attention_62	Grouped Query Attn	1 × 32768 × 5120
372	Add_62_attn	Add	1 × 32768 × 5120
373	LayerNorm_62_2	LayerNorm	1 × 32768 × 5120
374	FFN_62	Feed Forward	1 × 32768 × 5120
375	Add_62_ffn	Add	1 × 32768 × 5120
376	LayerNorm_63_1	LayerNorm	1 × 32768 × 5120
377	Attention_63	Grouped Query Attn	1 × 32768 × 5120
378	Add_63_attn	Add	1 × 32768 × 5120
379	LayerNorm_63_2	LayerNorm	1 × 32768 × 5120
380	FFN_63	Feed Forward	1 × 32768 × 5120
381	Add_63_ffn	Add	1 × 32768 × 5120
382	LayerNorm_64_1	LayerNorm	1 × 32768 × 5120
383	Attention_64	Grouped Query Attn	1 × 32768 × 5120
384	Add_64_attn	Add	1 × 32768 × 5120
385	LayerNorm_64_2	LayerNorm	1 × 32768 × 5120
386	FFN_64	Feed Forward	1 × 32768 × 5120
387	Add_64_ffn	Add	1 × 32768 × 5120
388	Final_LayerNorm	LayerNorm	1 × 32768 × 5120
389	LM_Head	Linear	1 × 32768 × 102400
390	Output	Output	1 × 32768 × 102400

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 LGAI-EXAONE/EXAONE-3.5-32B-Instruct-AWQ --plan --share

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

Other exaone checkpoints

EXAONE-3.5-7.8B-Instruct

5.94B derived · -24.0% against the checkpoint

EXAONE-3.5-7.8B-Instruct-AWQ

5.94B derived · -24.0% against the checkpoint