Stijn

[Feature Request] Support for QDoRA: Efficient quantized fine-tuning

2

> Today we’re releasing the next step: QDoRA. This is just as memory efficient and scalable as FSDP/QLoRA, and critically is also as accurate for continued pre-training as full weight...

Hi, because of it's multi-linguality and undertrainedness, I'd like to slice Falcon-11B. There are 60 of these layers. Supposedly, it would just be an easy layer name change? Or bumping...

Allow Sonnet 3.5 in main.js

1

## What this Pull Request (PR) does Add Claude 3.5 Sonnet

TPU v4-32 set-up not working

13

**Describe the bug** Can't train with multiple VM's; TPU v-4-32 It stops after loading the model, won't even load the data Been trying for two days, maybe my set-up is...

[MODEL EVALUATION REQUEST] mistralai/Mistral-Large-Instruct-2407

2

### Model ID mistralai/Mistral-Large-Instruct-2407 ### Model type Decoder model (e.g., GPT) ### Model languages - [X] Danish - [X] Swedish - [X] Norwegian (Bokmål or Nynorsk) - [X] Icelandic -...

[Feature Request] Custom Prompt

3

Hi, How would I add a custom prompt? with {{question}} or something to add code in between. I want to test this prompt ```markdown You are an expert AI programming...

``` [{ "resource": "/Users/air/Developer/Projects/../src/.../extract.py", "owner": "_generated_diagnostic_collection_name_#6", "code": { "value": "reportMissingModuleSource", "target": { "$mid": 1, "path": "/microsoft/pyright/blob/main/docs/configuration.md", "scheme": "https", "authority": "github.com", "fragment": "reportMissingModuleSource" } }, "severity": 4, "message": "Import \"pandas\" could...

DeepSpeed ZeRO-3 Parameter Loading Issue with ppo_vllm_thread_ray_gtrl.py

1

``` python open_instruct/ppo_vllm_thread_ray_gtrl.py --exp_name tulu-3-8b-rlvr --dataset_mixer_list allenai/RLVR-GSM-MATH-IF-Mixed-Constraints 1.0 --dataset_mixer_list_splits train --dataset_mixer_eval_list allenai/RLVR-GSM-MATH-IF-Mixed-Constraints 16 --dataset_mixer_eval_list_splits train --max_token_length 2048 --max_prompt_token_length 2048 --response_length 2048 --model_name_or_path allenai/Llama-3.1-Tulu-3-8B-DPO --reward_model_path allenai/Llama-3.1-Tulu-3-8B-RM --non_stop_penalty --stop_token eos --temperature 1.0...