[bong-water-water-bong]

428065 text-generation models on HF, 339805 with an architecture (2886 architectures).

	models	share
mapped (a backend's code accepts the architecture)	321215	94.53%
checked (passed serve_e2e on Strix Halo)	216732	63.78%

backend	mapped	checked
hrx	94.53%	63.78%
npu	53.67%	9.4%

Unmapped architectures with the most models:

• Step1MoEForCausalLM: 2882
• ParlerTTSForConditionalGeneration: 1585
• ChessForCausalLM: 970
• LlavaLlamaForCausalLM: 453
• DynamicAlibiForCausalLM: 245
• OpenLMForCausalLM: 245
• DynamicForgettingForCausalLM: 240
• BertLMHeadModel: 212
• MobileLLMForCausalLM: 176
• BartForConditionalGeneration: 174

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🧪 No relevant tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Data Consistency The census data shows an increase in the number of models with architectures (339805) and mapped models (321215) compared to the previous day, but the counts for architectures seen (2886) and checked models (216732) are not consistent with the expected growth rate. This discrepancy could indicate an issue with the data collection or processing pipeline, especially since the PR description claims 428065 text-generation models on HF, which is a significant increase from the previous count of 427926. The change in the number of architectures seen from 2877 to 2886 is also inconsistent with the expected growth pattern. { "date": "2026-10-24", "coverage": { "models": 428065, "with_architecture": 339805, "architectures_seen": 2886, "mapped_models": 321215, "mapped_pct": 94.53, "checked_models": 216732, "checked_pct": 63.78, "per_backend": { "hrx": { "mapped": 321215, "checked": 216732, "mapped_pct": 94.53, "checked_pct": 63.78 }, "npu": { "mapped": 182372, "checked": 31938, "mapped_pct": 53.67, "checked_pct": 9.4 } }, "top_unmapped": [ { Missing Architecture Count The PR introduces new architectures like K2HorizonForCausalLM and RobertaForCausalLM with model counts of 140 and 139 respectively, but the total count of architectures seen (2886) does not reflect these additions. This suggests that the architecture counting logic might not be correctly updated, potentially leading to an undercount of unique architectures. { "architecture": "K2HorizonForCausalLM", "models": 140 }, { "architecture": "RobertaForCausalLM", "models": 139 },