Store matrices as numpy arrays instead of Python lists
f2e89c2
gary-boon
Claude Opus 4.5
commited on
Spaces:
Sleeping
Sleeping
Commit History
Add per-step memory cleanup for large model support
a94eb19
gary-boon
Claude Opus 4.5
commited on
Fix RAM exhaustion for large token generation
959074d
gary-boon
Claude Opus 4.5
commited on
feat: add auto_complete parameter for token generation
bb689ce
gary-boon
Claude Opus 4.5
commited on
fix: add QKV extraction support for Mistral/Devstral architecture
d1d37a8
gary-boon
Claude Opus 4.5
commited on
feat: implement lazy-loading for attention matrices
929ba88
gary-boon
Claude Opus 4.5
commited on
Add avg_entropy calculation for attention heads
66a46b6
gary-boon
Claude Opus 4.5
commited on
Revert QKV visualization fixes - need better approach for data streaming
d0b7e29
gary-boon
Claude Opus 4.5
commited on
Add safety checks for missing QKV keys
a79cb83
gary-boon
Claude Opus 4.5
commited on
Limit QKV matrices to top 5 heads per layer to reduce response size
decb5ab
gary-boon
Claude Opus 4.5
commited on
Fix QKV matrix extraction for Mistral/Devstral architecture
9056859
gary-boon
Claude Opus 4.5
commited on
Fix QKV visualization for Mistral/Devstral architecture
4ec134b
gary-boon
Claude Opus 4.5
commited on
Add future considerations doc for response size optimization
3e67ea2
gary-boon
Claude Opus 4.5
commited on
Fix: Import time module at top level for SSE events
15a862b
gary-boon
Claude Opus 4.5
commited on
Add SSE streaming endpoint for real-time analysis progress
172a186
gary-boon
Claude Opus 4.5
commited on
feat: Include token metadata in analysis response
ee0f6c9
gary-boon
Claude Opus 4.5
commited on
feat: Implement tier-based model filtering by device type
6bf9f5c
gary-boon
Claude Opus 4.5
commited on
Fix: Add attn_implementation="eager" to model switch function
f94a7ae
gary-boon
Claude Opus 4.5
commited on
Add Phase 5: Performance optimizations to phased plan
383a328
gary-boon
Claude Opus 4.5
commited on
Add tokenSections boundaries and update system prompt
c6f4cc5
gary-boon
Claude Opus 4.5
commited on
Fix: Handle MistralCommonTokenizer pad_token setter
e20ccaf
gary-boon
Claude Opus 4.5
commited on
Integrate mistral-common for correct Devstral tokenization
ed06dcb
gary-boon
Claude Opus 4.5
commited on
Remove mistral_common to fix dependency conflict
3d9d9ee
gary-boon
Claude Opus 4.5
commited on
Use mistral_common for proper Devstral prompt formatting
3e80769
gary-boon
Claude Opus 4.5
commited on
Add system prompt support for instruction-tuned models
2860768
gary-boon
Claude Opus 4.5
commited on
fix: Simpler prompt format and temperature=0 for Devstral
76020ee
gary-boon
Claude Opus 4.5
commited on
fix: Sanitize JSON response for NaN/Inf float values
99f6209
gary-boon
Claude Opus 4.5
commited on
fix: Check chat_template is set before using apply_chat_template
474927d
gary-boon
Claude Opus 4.5
commited on
fix: Add chat template support for Devstral instruct model
8d85da8
gary-boon
Claude Opus 4.5
commited on
fix: Convert bfloat16 to float32 for numpy compatibility
cb6f39c
gary-boon
Claude Opus 4.5
commited on
fix: Use eager attention for output_attentions support
5333b21
gary-boon
Claude Opus 4.5
commited on
fix: Skip heavy ML deps in CI security checks
ba27c0c
gary-boon
Claude Opus 4.5
commited on
fix: Update torch to 2.3+ for transformers compatibility
1b73605
gary-boon
Claude Opus 4.5
commited on
fix: Update transformers for Devstral support
b788304
gary-boon
Claude Opus 4.5
commited on
docs: Mark GPU HF Space Devstral deployment complete
65c6e2e
gary-boon
Claude Opus 4.5
commited on
docs: Update phased plan with Phase 2/2b/2c completion status
688efad
gary-boon
Claude Opus 4.5
commited on
Add vocabSize to modelInfo response
499afba
gary-boon
Claude Opus 4.5
commited on
Update .env.spark.example: TORCH_DTYPE now auto-detected
543454f
gary-boon
Claude Opus 4.5
commited on
Add recommended_dtype to model configs
62525b2
gary-boon
Claude Opus 4.5
commited on
Phase 2: Add Devstral backend support
9080f28
gary-boon
Claude Opus 4.5
commited on
Update plan: Phase 1 paused due to GB10 GPU support
e694533
gary-boon
Claude Opus 4.5
commited on
Add DEVICE env var to force CPU mode on DGX Spark
5f122aa
gary-boon
Claude Opus 4.5
commited on
Use NGC PyTorch 24.08 for Python 3.10 compatibility
a2875a2
gary-boon
Claude Opus 4.5
commited on
Use NVIDIA NGC PyTorch container for GB10 support
a4cfbff
gary-boon
Claude Opus 4.5
commited on
Try PyTorch nightly for GB10/sm_121 GPU support
a009a49
gary-boon
Claude Opus 4.5
commited on
Make zarr/numcodecs imports optional for ARM64 compatibility
6435a75
gary-boon
Claude Opus 4.5
commited on
Skip zarr/numcodecs in Spark build (ARM64 incompatible)
d129e37
gary-boon
Claude Opus 4.5
commited on
Fix numcodecs ARM64 compatibility in Dockerfile.spark
772fc80
gary-boon
Claude Opus 4.5
commited on
Fix Dockerfile.spark for CUDA 13.0 compatibility
a4927aa
gary-boon
Claude Opus 4.5
commited on
Fix Dockerfile.spark for ARM64 architecture (DGX Spark)
9d00d33
gary-boon
Claude Opus 4.5
commited on