初始化项目,由ModelHub XC社区提供模型
Model: properly59/Jumini-Ko-1.2B Source: Original Platform
This commit is contained in:
30
hf_equivalence_atol2e-5.json
Normal file
30
hf_equivalence_atol2e-5.json
Normal file
@@ -0,0 +1,30 @@
|
||||
{
|
||||
"ok": true,
|
||||
"logits_shape": [
|
||||
2,
|
||||
8,
|
||||
128000
|
||||
],
|
||||
"max_abs_diff": 1.4781951904296875e-05,
|
||||
"mean_abs_diff": 1.073120984074194e-06,
|
||||
"rms_diff": 1.5380918512164499e-06,
|
||||
"max_rel_diff": 0.04641776159405708,
|
||||
"argmax_mismatch_rate": 0.0,
|
||||
"atol": 2e-05,
|
||||
"rtol": 2e-05,
|
||||
"config": "/raid2/chojm/hw2_llm_from_scratch/configs/jumini_1_2b.json",
|
||||
"checkpoint": "/tmp/chojm_hw2_runs/jumini-ko-1.2b-firsttok7000-argmax-boundary-layernormshock-v1-last16-4gpu0123-mb8-probe25-rerun20260608/step_00007020/model.pt",
|
||||
"hf_model_dir": "/raid2/chojm/hw2_llm_from_scratch/models/jumini-ko-1.2b-clean7020-best-20260609-hf-fp32",
|
||||
"hf_model_class": "LlamaForCausalLM",
|
||||
"device": "cpu",
|
||||
"comparison_dtype": "float32",
|
||||
"hf_export_torch_dtype": "float32",
|
||||
"round_native_to_export_dtype": false,
|
||||
"input_shape": [
|
||||
2,
|
||||
8
|
||||
],
|
||||
"input_ids_sha256": "b9225a84aaeeb49dbb7ac34866e25a8fdbf0823d9fe02530e3686d2e09f72f6f",
|
||||
"seed": 1234,
|
||||
"max_tokens": 64
|
||||
}
|
||||
Reference in New Issue
Block a user