4 Pull requests merged by 3 people

• nits for CI

  #1940 merged Mar 11, 2025

• ci: use HF cache

  #1958 merged Mar 11, 2025

• fix skip condition

  #1956 merged Mar 10, 2025

• handle wrapped thundermodules in generate

  #1955 merged Mar 10, 2025

4 Pull requests opened by 3 people

• QwQ-32B

  #1952 opened Mar 6, 2025

• bump: PT 2.6 + `bitsandbytes` & standalone tests

  #1959 opened Mar 11, 2025

• ci: split HF caching

  #1960 opened Mar 11, 2025

• thunder fsdp strategy fix

  #1961 opened Mar 11, 2025

5 Issues opened by 5 people

• Multiple redundant calls to generate_example() when using multiple GPUs

  #1957 opened Mar 10, 2025

• Falcon3-1B-Base has the model.safetensors.index.json file from Falcon3-3B-Base?

  #1954 opened Mar 9, 2025

• litgpt chat crash at first char that differs from english encodind with error : UnicodeDecodeError:

  #1953 opened Mar 8, 2025

• Lora training seems to be using the same single record for validation step

  #1951 opened Mar 6, 2025

• Getting probability distributions

  #1950 opened Mar 5, 2025

12 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.

• Support for mini-omni and mini-omni2 pre training, fine tuning on custom dataset.

  #1809 commented on Mar 7, 2025 • 0 new comments

• example for full finetuning with python code done!

  #1331 commented on Mar 11, 2025 • 0 new comments

• Do not wrap LoRA layers with FSDP

  #1538 commented on Mar 11, 2025 • 0 new comments

• OLMo 2

  #1897 commented on Mar 11, 2025 • 0 new comments

• Raise error if disk is full before downloading weights

  #1903 commented on Mar 11, 2025 • 0 new comments

• Support for KV caching and batched inference

  #1934 commented on Mar 5, 2025 • 0 new comments

• Speculative decoding: Base implementation

  #1938 commented on Mar 11, 2025 • 0 new comments

• Feature: Adds support for OpenAISpec in litgpt serve

  #1943 commented on Mar 11, 2025 • 0 new comments

• Add Multi-head Latent Attention (DeepSeekv2)

  #1945 commented on Mar 11, 2025 • 0 new comments

• fix n_query_groups for llama-3.1-405b

  #1946 commented on Mar 11, 2025 • 0 new comments

• Fix: incorrect gradient accumulation steps bug

  #1947 commented on Mar 12, 2025 • 0 new comments

• Phi4 mini

  #1949 commented on Mar 11, 2025 • 0 new comments