-
Notifications
You must be signed in to change notification settings - Fork 82
Insights: mirage-project/mirage
Overview
Could not load contribution data
Please try again later
11 Pull requests merged by 7 people
-
Fix some hardcode
#338 merged
Jun 27, 2025 -
fix tmp file permission issue
#363 merged
Jun 26, 2025 -
fix build errors on some systems
#359 merged
Jun 26, 2025 -
update cmake to use the newest version
#350 merged
Jun 26, 2025 -
Remove cudnn/cublas dependency
#347 merged
Jun 25, 2025 -
Add a Qwen3 chat demo, increase output length cap to 4096, change tem…
#355 merged
Jun 25, 2025 -
[MPK] Remove mpi4py dependency for running the Qwen3 demo on a single GPU
#349 merged
Jun 23, 2025 -
Use symbolic tensor dimensions to reduce search space
#161 merged
Jun 23, 2025 -
Add issue template
#341 merged
Jun 23, 2025 -
Support other qwen3 model
#337 merged
Jun 22, 2025 -
Clear buffers in a unified way
#334 merged
Jun 22, 2025
4 Pull requests opened by 4 people
-
[MPK] Bug fixes for running MPK on multiple GPUs
#364 opened
Jun 27, 2025 -
implement paged attention
#367 opened
Jun 28, 2025 -
Support Prompt Lookup Decoding
#368 opened
Jun 28, 2025 -
probe gpu and set worker and scheduler according to it
#369 opened
Jun 28, 2025
4 Issues closed by 2 people
-
Prompt-Lookup Decoding
#366 closed
Jun 28, 2025 -
[Feature Request] - Update cmake to use the newest version
#348 closed
Jun 26, 2025 -
Remove cudnn/cublas dependencies in building pipeline
#333 closed
Jun 25, 2025 -
[MPK] Remove the flashinfer dependencies in the demo
#320 closed
Jun 23, 2025
16 Issues opened by 12 people
-
[Roadmap] - Speculative Decoding Support
#365 opened
Jun 28, 2025 -
[Expected output] Qwen3 demo
#362 opened
Jun 26, 2025 -
[Parallelization] - How to compile and run on multiple gpus?
#361 opened
Jun 26, 2025 -
[Feature Request] - [MPK] Support LLAMA-3 model family
#360 opened
Jun 26, 2025 -
[Feature Request] - qwen3-8B demo; larger batch size
#358 opened
Jun 25, 2025 -
[Bug] - Qwen3 demo hung on NV-A10
#357 opened
Jun 25, 2025 -
[Bug] - Qwen2.5-0.5B core dump
#356 opened
Jun 25, 2025 -
[Comment] - Default Persistent Kernel Configuration for Different GPU types
#354 opened
Jun 24, 2025 -
[Bug] - qwen2.5 coredump on H20
#351 opened
Jun 24, 2025 -
USE_NVSHMEM how to set this when build mirage?
#346 opened
Jun 23, 2025 -
Exception: data did not match any variant of untagged enum ModelWrapper at line 757479 column 3
#345 opened
Jun 23, 2025 -
core dump after migrage with Assertion `tensor.dim[dim_idx] % dim_div == 0' failed
#344 opened
Jun 23, 2025 -
performance
#343 opened
Jun 23, 2025 -
vLLM integration
#342 opened
Jun 23, 2025 -
[MPK] Adding instructions for profiling MPK megakernel through perfetto trace
#340 opened
Jun 22, 2025 -
tg4perfetto Additional instructions for Quick Installation / Quickstart
#339 opened
Jun 22, 2025
7 Unresolved conversations
Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.
-
Blackwell Support
#263 commented on
Jun 24, 2025 • 4 new comments -
build failed
#335 commented on
Jun 22, 2025 • 0 new comments -
Support Qwen3 demo with other Qwen3 model and other GPUs
#336 commented on
Jun 24, 2025 • 0 new comments -
[MPK] Development Roadmap for Mirage Persistent Kernel
#325 commented on
Jun 26, 2025 • 0 new comments -
[MPK] Supporting MoE models in MPK
#332 commented on
Jun 29, 2025 • 0 new comments -
[MPK] Supporting Prompt Lookup Decoding in MPK
#328 commented on
Jun 29, 2025 • 0 new comments -
Fingerprints without GPU
#298 commented on
Jun 26, 2025 • 0 new comments