MakazhanAlpamys

MakazhanAlpamys/Soup

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

72 stars 20 forks 72 watchers Python Apache License 2.0
artificial-intelligence cli dpo fine-tuning finetuning gguf huggingface llm llmops local-llm lora machine-learning model-finetuning ollama peft python pytorch qlora sft transformers
View on GitHub Website
100 Open Issues Need Help Last updated: Jul 1, 2026

Open Issues Need Help

View All on GitHub
Validate closed-loop reward-hacking controller at scale (7B+, real reward models) about 10 hours ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
Add a ready-made recipe for a popular model (good first issue) 3 days ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
soup ship: pairwise judge win-rate (--task-mode pairwise) 4 days ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
Judge-URL SSRF prefix-bypass in eval/gate.py (startswith -> urlparse) 4 days ago
bug help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
Add ready-made recipe: glm-5.1-dpo (DPO variant) 9 days ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
Add unit tests for untested pure helpers (good first issue) 10 days ago
help wanted good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
Improve friendly error messages in utils/errors.py (good first issue) 10 days ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
Add ready-made recipe: kimi-k2.6-grpo (GRPO reasoning variant) 11 days ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
Add ready-made recipe: deepseek-v4-flash-grpo (GRPO reasoning variant) 11 days ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
Add ready-made recipe: qwen3.5-4b-pretrain (continued pre-training, Base checkpoint) 11 days ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
Add ready-made recipe: qwen3.5-9b-grpo (GRPO reasoning variant) 11 days ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
Add ready-made recipe: qwen3.5-35b-a3b-dpo (DPO variant) 11 days ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
[Tracking] Recipe task-variants for the 2026 model families (DPO / GRPO / pretrain) 11 days ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
MLX DPO/GRPO raises raw NotImplementedError mid-training instead of a friendly pre-flight gate 11 days ago
bug good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.71.x: Live-codec TTS — encode-at-train-time for the 5 families (snac/moshi/xcodec2/sparktts/outetts) 21 days ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.46.1: First-party transformers / vLLM generator factories for soup deploy autopilot --measure 22 days ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
MiniLLM on-policy rollout: KV-cache the autoregressive loop (currently O(L²) re-forward) 22 days ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
soup serve --mole: KV-cache the per-token blend decode loop 22 days ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
Cloud training: add runpod / lambda backends behind 'soup train --cloud' 22 days ago
enhancement help wanted infra-blocked
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
[QA] Verify multi-GPU training: --gpus N + --fsdp + --deepspeed 23 days ago
help wanted qa
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
[QA] Verify HuggingFace Hub push + Spaces deploy (push, train --push-as, data push, deploy hf-space) 26 days ago
help wanted qa infra-blocked
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.53.x: CUDA + llama.cpp manual QA smoke for #70 (GGUF export) and #72 (AWQ/GPTQ export) 26 days ago
documentation help wanted good first issue infra-blocked
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
[QA] Verify AWQ + GPTQ quantization export (--bits 4) 26 days ago
help wanted qa infra-blocked
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
[QA] Verify GGUF export pipeline (q4_0 / q4_k_m / q5_k_m / q8_0 / f16 / f32) 26 days ago
help wanted qa infra-blocked
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.53.1: Live kv_cache_type wiring on serve backends (vLLM / SGLang / transformers) 26 days ago
enhancement help wanted infra-blocked
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.71.x: Per-request user isolation for soup serve --bank (multi-worker safe) 27 days ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.71.x: Serve-time MoLE inference path (load mole_gate.pt + frozen task LoRAs at decode) 27 days ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
Support non-Llama architectures in knowledge-edit kernels (GPT-2 transformer.h) 28 days ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.66.x: Real Anthropic-calibrated sleeper-probe weights (replace deterministic synthetic placeholder) 29 days ago
enhancement help wanted upstream-gated
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
soup build: manifest-level dotted-path custom transforms 29 days ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.63.1: Live SaaS pull adapters for soup ingest (Langfuse/LangSmith/Helicone/OpenPipe API SDKs) 30 days ago
enhancement help wanted infra-blocked
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
Cache the base model across cmaes merge candidates (default scorer reloads per candidate) 30 days ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
Wire a real per-iteration cost estimate into soup loop (estimate_cost returns 0.0) 30 days ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.59.1: Sigstore + ed25519 live signing for soup attest about 1 month ago
enhancement help wanted infra-blocked
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.60.1: Sigstore + ed25519 live signing for soup adapters sign about 1 month ago
enhancement help wanted infra-blocked
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.70.x: Tokeniser-aware echo-trap detector (subword n-grams over token ids instead of whitespace) about 1 month ago
enhancement help wanted good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.70.1: Live build_echo_trap_callback (RAGEN trajectory-degeneration HF Trainer callback) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.70.1: Live run_iterative_dpo orchestrator (sample → RM-score → re-pair → retrain subprocess loop) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.70.1: Live build_rl_checkpoint_callback (mid-epoch optimizer-state serialization for PPO/GRPO) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.70.1: Live build_minillm_callback (teacher-mix sampling + length-norm + pretrain-loss anchor) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.70.1: Live build_uld_projection (cross-tokenizer Wasserstein + topk_align kernels) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.70.1: Live build_reward_hack_callback (InfoRM cluster-sep + RM-ensemble HF Trainer callback) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.69.x: Multilingual brain-rot heuristics (extend `_LOW_EFFORT_TOKENS` + `_CLICKBAIT_PHRASES` for non-English corpora) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.68.x: soup serve --record-thumbs <db> auto-capture for local-rl integration about 1 month ago
enhancement help wanted good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.68.1: Live nightly DPO/KTO/ORPO scheduler for soup local-rl train about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.68.1: Live HF / MLX / Apple FoundationModels converter for soup apple-adapter about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.68.1: Live TextGrad / GEPA tool-schema optimiser for soup compile-tools about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.68.1: Live teacher/student dataset preparation for soup distill-prompt about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.68.1: Live DSPy / GEPA orchestrator for soup compile about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.67.1: soup lock auto-glue between soup env lock and soup lock write about 1 month ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.67.1: soup adapters pr GitHub PR publisher (pair with v0.68 Part A) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.67.1: Live MoLE gating-kernel training + per-token routing about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.67.1: Live VeRA / VB-LoRA serving via apply_bank_to_serve about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.67.1: Live eval-suite auto-wiring for soup adapters merge --strategy cmaes about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.66.x: soup train --capture-activations <layer> for SAE feature diff about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.66.x: Auto-measure soup probe interference --measure <eval_suite> about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.66.x: Live truth + harm probe utilities (PROBE_KINDS forward-compat surface) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.65.x: judge_calibration registry artifact kind for ensure_judge_calibrated about 1 month ago
enhancement help wanted good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.65.x: 2PL + 3PL IRT models for soup eval irt-subset about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.65.1: Live model-driven soup eval behavior (load base + adapter, generate responses, emit diff) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.65.1: Live lm-eval-harness invocation in soup eval capability about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.64.1: complete_target_modules HF-config introspection per base about 1 month ago
enhancement help wanted good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.64.1: soup env lock --fix action (recreate venv from soup-env.lock via uv/nix) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.64.1: Live LoRA probe for soup tunability (replace heuristic _default_probe stub) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.63.1: Optional Slack/Discord webhook for soup ingest / prune-prompt / ab / active-sample about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.63.1: Variance-based diversity score for K>2 reward models in soup data active-sample about 1 month ago
enhancement help wanted good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.63.1: Tokenizer-aware prefix detection for soup prune-prompt about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.62.1: Live GRACE codebook lookup/write + Registry persistence about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.62.1: Citation-span loss-mask + eval-gate / diagnose wiring for citation_faithful about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.62.1: Live CAA / ITI / RepE fitting + soup serve --steer decode-hook about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.62.1: Live two-stage RA-DIT orchestrator (auto-link retriever to generator) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.62.1: Live RAFT span-mask training kernel about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.60.1: Magic-byte check in strict-safetensors (catch pickle renamed to .safetensors) about 1 month ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.61.1: apply_edit should consult EditGovernor automatically when one is supplied about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.61.1: EditGovernor SQLite persistence + cross-process locking about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.61.1: MUSE + WMDP benchmark loaders for soup eval unlearning about 1 month ago
enhancement help wanted good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.61.1: Live ROME / MEMIT / AlphaEdit kernels + before/after generation in soup edit diff about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.61.1: Live NPO / SimNPO / RMU loss kernels + UnlearnTrainerWrapper.setup about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.60.1: Auto-refuse merge when soup adapters scan returns FAIL (--allow-unscanned opt-out) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.60.1: NamespacePinStore concurrent-writer support (WAL + cross-process file-lock) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.60.1: Wire license-override reason into v0.59 audit-log about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.60.1: Auto-attach v0.59 repro-receipt into soup airgap-bundle about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.60.1: Auto-extract license declarations from adapter_config.json / HF model-card metadata about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.60.1: Wire NamespacePinStore into utils/hubs.download_repo (anti-AI-Jacking on every Hub fetch) about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.59.1: Auto-populate AnnexXIData.top_domains from data.train manifest about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.59.1: Audit-log auto-instrumentation via top-level Typer callback about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.59.1: Soup Can manifest v3 attestations field + embed in_toto Statement about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.59.1: PDF rendering for soup train --annex-xi about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.59.1: Live CodeCarbon hook in trainer wrappers about 1 month ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.58.1: Pack each soup loop iteration as a v0.26 Soup Can + append to Registry lineage DAG about 2 months ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.58.1: Pre-wired stage callbacks for soup loop watch (harvest / train / gate / deploy) about 2 months ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.57.1: Rich-markup escape backfill for legacy soup adapters list / info / compare about 2 months ago
bug good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.57.1: wire adapter branch pointers into v0.26 Registry lineage DAG about 2 months ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.57.1: live canary verdict for soup adapters merge (replace UNKNOWN stub) about 2 months ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.57.1: live blame ablation runner — lift NotImplementedError stub about 2 months ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
soup train --diagnose-gate: multi-node RANK awareness about 2 months ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
soup train --diagnose-gate: rank-0 guard for distributed (DDP/FSDP/DeepSpeed) runs about 2 months ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
soup diagnose refusal probe: extend _REFUSAL_PATTERNS beyond English about 2 months ago
enhancement good first issue
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.55.1: extend get_metric_series to eval_results table for soup eval against about 2 months ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers
v0.54.1: bias build_verdict rubric from advise_history outcomes about 2 months ago
enhancement help wanted
MakazhanAlpamys/Soup
72

Soup turns the pain of LLM fine-tuning into a simple workflow. One config, one command, done.

Python
#artificial-intelligence#cli#dpo#fine-tuning#finetuning#gguf#huggingface#llm#llmops#local-llm#lora#machine-learning#model-finetuning#ollama#peft#python#pytorch#qlora#sft#transformers