Skip to content

Commit 3a91a54

Browse files
xaduprexadupre
committed
cache config
1 parent c42e4e1 commit 3a91a54

File tree

2 files changed

+143
-1
lines changed

2 files changed

+143
-1
lines changed

‎onnx_diagnostic/torch_models/hghub/hub_data_cached_configs.py‎

Lines changed: 142 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -4687,3 +4687,145 @@ def _ccached_zai_glm_45():
46874687
},
46884688
}
46894689
)
4690+
4691+
4692+
def _ccached_microsoft_phi3_mini_128k_instruct():
4693+
"microsoft/Phi-3-mini-128k-instruct"
4694+
return transformers.Phi3Config(
4695+
**{
4696+
"_name_or_path": "Phi-3-mini-128k-instruct",
4697+
"architectures": ["Phi3ForCausalLM"],
4698+
"attention_dropout": 0.0,
4699+
"auto_map": {
4700+
"AutoConfig": "configuration_phi3.Phi3Config",
4701+
"AutoModelForCausalLM": "modeling_phi3.Phi3ForCausalLM",
4702+
},
4703+
"bos_token_id": 1,
4704+
"embd_pdrop": 0.0,
4705+
"eos_token_id": 32000,
4706+
"hidden_act": "silu",
4707+
"hidden_size": 3072,
4708+
"initializer_range": 0.02,
4709+
"intermediate_size": 8192,
4710+
"max_position_embeddings": 131072,
4711+
"model_type": "phi3",
4712+
"num_attention_heads": 32,
4713+
"num_hidden_layers": 32,
4714+
"num_key_value_heads": 32,
4715+
"original_max_position_embeddings": 4096,
4716+
"pad_token_id": 32000,
4717+
"resid_pdrop": 0.0,
4718+
"rms_norm_eps": 1e-05,
4719+
"rope_scaling": {
4720+
"long_factor": [
4721+
1.0700000524520874,
4722+
1.1200000047683716,
4723+
1.149999976158142,
4724+
1.4199999570846558,
4725+
1.5699999332427979,
4726+
1.7999999523162842,
4727+
2.129999876022339,
4728+
2.129999876022339,
4729+
3.009999990463257,
4730+
5.910000324249268,
4731+
6.950000286102295,
4732+
9.070000648498535,
4733+
9.930000305175781,
4734+
10.710000038146973,
4735+
11.130000114440918,
4736+
14.609999656677246,
4737+
15.409998893737793,
4738+
19.809999465942383,
4739+
37.279998779296875,
4740+
38.279998779296875,
4741+
38.599998474121094,
4742+
40.12000274658203,
4743+
46.20000457763672,
4744+
50.940006256103516,
4745+
53.66000747680664,
4746+
54.9373893737793,
4747+
56.89738845825195,
4748+
57.28738784790039,
4749+
59.98738479614258,
4750+
60.86738586425781,
4751+
60.887386322021484,
4752+
61.71739196777344,
4753+
62.91739273071289,
4754+
62.957393646240234,
4755+
63.41739273071289,
4756+
63.8173942565918,
4757+
63.83739471435547,
4758+
63.897396087646484,
4759+
63.93739700317383,
4760+
64.06739807128906,
4761+
64.11434936523438,
4762+
64.12435150146484,
4763+
64.15435028076172,
4764+
64.19435119628906,
4765+
64.24435424804688,
4766+
64.57435607910156,
4767+
64.69000244140625,
4768+
64.76000213623047,
4769+
],
4770+
"short_factor": [
4771+
1.1,
4772+
1.1,
4773+
1.1,
4774+
1.3000000000000003,
4775+
1.3500000000000003,
4776+
1.3500000000000003,
4777+
1.4000000000000004,
4778+
1.5500000000000005,
4779+
2.000000000000001,
4780+
2.000000000000001,
4781+
2.000000000000001,
4782+
2.000000000000001,
4783+
2.000000000000001,
4784+
2.000000000000001,
4785+
2.000000000000001,
4786+
2.000000000000001,
4787+
2.000000000000001,
4788+
2.000000000000001,
4789+
2.000000000000001,
4790+
2.000000000000001,
4791+
2.000000000000001,
4792+
2.000000000000001,
4793+
2.000000000000001,
4794+
2.000000000000001,
4795+
2.000000000000001,
4796+
2.0500000000000007,
4797+
2.0500000000000007,
4798+
2.0500000000000007,
4799+
2.0500000000000007,
4800+
2.0500000000000007,
4801+
2.0500000000000007,
4802+
2.1000000000000005,
4803+
2.1000000000000005,
4804+
2.1500000000000004,
4805+
2.25,
4806+
2.25,
4807+
2.25,
4808+
2.25,
4809+
2.25,
4810+
2.3999999999999995,
4811+
2.4499999999999993,
4812+
2.499999999999999,
4813+
2.6999999999999984,
4814+
2.6999999999999984,
4815+
2.7499999999999982,
4816+
2.799999999999998,
4817+
2.8999999999999977,
4818+
3.049999999999997,
4819+
],
4820+
"type": "longrope",
4821+
},
4822+
"rope_theta": 10000.0,
4823+
"sliding_window": 262144,
4824+
"tie_word_embeddings": false,
4825+
"torch_dtype": "bfloat16",
4826+
"transformers_version": "4.40.2",
4827+
"use_cache": true,
4828+
"attention_bias": false,
4829+
"vocab_size": 32064,
4830+
}
4831+
)

‎onnx_diagnostic/torch_models/validate.py‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -836,7 +836,7 @@ def node_iter(proto):
836836
else:
837837
raise NotImplementedError(f"Unexpected type={type(proto)}")
838838

839-
counts = {}
839+
counts: Dict[str, Union[float, int]] = {}
840840
for proto in node_iter(onx):
841841
if isinstance(proto, onnx.NodeProto):
842842
key = f"n_node_{proto.op_type}"

0 commit comments

Comments
 (0)