Skip to content

Pull requests: FlashML-org/FreeToken

New pull request New
143 Open 41 Closed
143 Open 41 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

feat(qwen3_5_moe): keep a natively-fp8 lm_head native instead of failing to load
#416 opened Sep 8, 2026 by salekseev Loading…
1
fix(qwen3_5_moe): accept a per-channel fp8 weight_scale in the dense loader
#415 opened Sep 8, 2026 by salekseev Loading…
fix(kernel): order the batch-memcpy probe against the current stream
#414 opened Sep 8, 2026 by tspeaks Loading…
fix(models): support llm-compressor NVFP4 MoE export variants
#413 opened Sep 8, 2026 by Romeriz Loading…
1
fix(bench): show how to use custom bandwidth profiles
#412 opened Sep 7, 2026 by earlvanze Loading…
1
fix(server): honor configured output default across APIs
#411 opened Sep 7, 2026 by earlvanze Loading…
1
feat(kvcache): add nvfp4 kv quantization
#408 opened Sep 7, 2026 by ArqAlice Loading…
docs(models): qualify full GLM-5.3 on H200 Serverless
#406 opened Sep 7, 2026 by earlvanze Loading…
fix(models): serve qwen4_exp FTW checkpoints with PLE streamed from source
#405 opened Sep 7, 2026 by Eng-Ahmd Loading…
fix(models): support nvidia/Qwen3.8-Flash-Next-NVFP4 in qwen4_exp
#404 opened Sep 7, 2026 by Eng-Ahmd Loading…
Help/pr 132
#400 opened Sep 6, 2026 by samuelishida • Draft
fix(cpu-moe): honor padded fp8 scale strides, add avx512f tier and float64 parity (builds on #36)
#399 opened Sep 6, 2026 by ChenyuHeee Loading…
feat(kvcache): add ISO3/ISO4 KV-cache quantization (--kv-cache-iso)
#398 opened Sep 5, 2026 by AsmanovLev Loading…
2
feat(server): OpenAI protocol coverage -- sampling extras, n, echo/suffix, usage details, tokenize/detokenize/metrics
#393 opened Sep 5, 2026 by gdevenyi • Draft
4 tasks done
qwen4_exp: serve the block-FP8 dense projections natively (+25% decode)
#392 opened Sep 5, 2026 by gberasmus87 Loading…
3
fix(tokenizer): accept the OpenAI developer role (map to system for templates without it)
#391 opened Sep 5, 2026 by gdevenyi • Draft
fix(models): better support for mixed-precision compressed-tensors NVFP4
#390 opened Sep 5, 2026 by Sam-Izdat Loading…
7
feat(qwen4_exp): load-time per-tensor FP8 dense projections (W8A8 via _scaled_mm)
#389 opened Sep 4, 2026 by gdevenyi • Draft
2
test(qwen4_exp): compare chunked prefill to one-shot at bf16 tolerance
#388 opened Sep 4, 2026 by gdevenyi • Draft
feat(qwen4_exp): image input (vision tower + mRoPE) for Qwen3.8-Flash-Next
#386 opened Sep 4, 2026 by gdevenyi • Draft
5
feat(qwen4_exp): tensor parallelism for Qwen3.8-Flash-Next (offload backend)
#385 opened Sep 4, 2026 by gdevenyi • Draft
1
ci(container): настроить сборку образа
#380 opened Sep 4, 2026 by Fgeeha Loading…
1
fix(rocm): make CPU/Hybrid MoE graph replay safe
#378 opened Sep 4, 2026 by zihaomu • Draft
1
feat(gguf): implement Q8_0 dequantization
#368 opened Sep 3, 2026 by Cyber-Marty Loading…
1
fix(scheduler): reserve paged KV at allocation granularity
#367 opened Sep 3, 2026 by taking-lying-flat Contributor Loading…
ProTip! Filter pull requests by the default branch with base:main.