Issues

• any plan to support qwen3.5 and onwards variants

  snowflakedb/ArcticTraining#375 · Filter by author jamesunnc opened on Jun 15, 2026

• FA3 with SwiftKV

  snowflakedb/ArcticTraining#372 · Filter by author raywanb opened on Apr 28, 2026

• sft_factory bug

  snowflakedb/ArcticTraining#353 · Filter by author duanzhenyu001 opened on Feb 28, 2026

• [feature request] dataset blending

  snowflakedb/ArcticTraining#348 · Filter by author sfc-gh-sbekman opened on Jan 30, 2026
  5

• Design rationale for TiledMLP parameter registration (Hardcoded vs. Dynamic) & PEFT Support

  snowflakedb/ArcticTraining#334 · Filter by author akshatvishu opened on Dec 23, 2025
  1

• Print Metric Issue / Feature Request

  snowflakedb/ArcticTraining#305 · Filter by author sfc-gh-fjiang opened on Nov 11, 2025
  1

• Does this training support multi-modal models?

  snowflakedb/ArcticTraining#301 · Filter by author ggg-s opened on Nov 4, 2025
  1

• Will the Arctic-Text2SQL-R1 training code be released?

  snowflakedb/ArcticTraining#297 · Filter by author wuxhua opened on Oct 28, 2025

• FP8 models load on CPU

  snowflakedb/ArcticTraining#281 · Filter by author Cameron7195 opened on Sep 18, 2025

• [feature request] Support for MoE models

  snowflakedb/ArcticTraining#278 · Filter by author Cameron7195 opened on Sep 11, 2025
  7

  

• [feature request] Support for multiple models

  snowflakedb/ArcticTraining#277 · Filter by author Cameron7195 opened on Sep 9, 2025
  2

• How can I train a model like Arctic-Text2SQL-R1

  snowflakedb/ArcticTraining#273 · Filter by author xiaolong25 opened on Aug 31, 2025

• Liger-Kernel fails to compute loss when using train+eval and SP>1

  snowflakedb/ArcticTraining#266 · Filter by author sfc-gh-sbekman opened on Aug 20, 2025
  6

  

• CPU OOM issue

  snowflakedb/ArcticTraining#263 · Filter by author sfc-gh-jrasley opened on Aug 14, 2025

• Experiment artifact output

  snowflakedb/ArcticTraining#262 · Filter by author sfc-gh-jrasley opened on Aug 14, 2025

• OOM on QuickStart with 8xH100 & Llama

  snowflakedb/ArcticTraining#261 · Filter by author Cameron7195 opened on Aug 14, 2025
  3

• Support / guidance on training an LSTM-based speculator for Qwen2.5-VL-32B

  snowflakedb/ArcticTraining#255 · Filter by author lahmuller opened on Aug 7, 2025

• training loss does not converge

  snowflakedb/ArcticTraining#246 · Filter by author sufengniu opened on Jul 22, 2025

• Compatible with MoE block - tiled_mlp_forward_common?

  snowflakedb/ArcticTraining#231 · Filter by author airlsyn opened on Jul 10, 2025

• How can I obtain the training data for Arctic-Text2SQL-R1-7B?

  snowflakedb/ArcticTraining#224 · Filter by author xingyuma618 opened on Jun 30, 2025
  

• Does this repo support RL, or will it in the future?

  snowflakedb/ArcticTraining#200 · Filter by author ChinaStark opened on Jun 10, 2025
  2

• The performance of Snowflake / Arctic-Text2SQL-R1-7B

  snowflakedb/ArcticTraining#195 · Filter by author 652994331 opened on Jun 4, 2025
  1

  

• better yaml format / indentation reporting

  snowflakedb/ArcticTraining#158 · Filter by author sfc-gh-sbekman opened on Apr 14, 2025
  

• data packing sometimes misfires and fails to come close to the wanted max_length

  snowflakedb/ArcticTraining#155 · Filter by author sfc-gh-sbekman opened on Apr 14, 2025
  2

• metrics to support GAS

  snowflakedb/ArcticTraining#143 · Filter by author sfc-gh-sbekman opened on Apr 10, 2025
  1